VMware ESX 3.x Server and VirtualCenter 2.

x (GA Build Eval) Service Console Guide Document Version 1.1

By Mike Laverick RTFM Education For Errors/Corrections please contact: [email protected]

Table of Contents
Introduction .............................................................................................3 Module 1: Users and Rights ......................................................................6 Creating a New User ................................................................................. 6 Elevating yourself to ROOT ........................................................................ 6 Disabling Auditing on ROOT (Not Recommended) ......................................... 6 Miscellaneous User Managements Tasks ...................................................... 7 Module 2: Getting around & Getting help .................................................8 Module 3: File & Folder Management......................................................10 Module 4: Networking ............................................................................15 Viewing your Switches & Service Console Networking ................................. 15 Creating a vSwitch (Internal) ................................................................... 15 Creating a vSwitch (Single NIC) ............................................................... 16 Creating a vSwitch (Multiple NICs) ........................................................... 17 Deleting a Switch from vSwitch ................................................................ 18 Creating PortGroups for VLAN Networking NIC Team .................................. 19 Creating Vmkernel Switches .................................................................... 20 Changing your Service Console IP Settings ................................................ 22 Setting the Speed and Duplex of NICs ...................................................... 23 Recreating your vswif0 Interface .............................................................. 23 Removing a NIC from vSwitch.................................................................. 24 Deleting a PortGroup from vSwitch ........................................................... 24 Managing the ESX Firewall....................................................................... 24 Fixing ESX Networking after an Upgrade ................................................... 27 Typical vSwitch Errors ............................................................................. 29 Module 5: Configuring Physical Storage .................................................30 Configuring SAN Storage ......................................................................... 30 Configuring NAS Storage ......................................................................... 30 Connecting to iSCSI Storage (Software) .................................................... 31 Creating a VMFS Volume ......................................................................... 33 Changing the Volume Label ..................................................................... 35 Viewing VMFS Volumes/Partition Information ............................................. 36 Viewing Available Physical Disk Space ....................................................... 37 Mounting and U-mounting VMFSs ............................................................ 38 Managing Storage Devices....................................................................... 38 Using USB storage Devices ...................................................................... 39 Module 6: Managing Virtual Machines ....................................................43 Creating Virtual Disks ............................................................................. 43 Creating and Managing RDMs .................................................................. 44 Exporting & Importing Virtual Disks .......................................................... 46 Copying Files from one ESX host to another............................................... 48 Renaming Virtual Disks ........................................................................... 49 Resizing Virtual Disks.............................................................................. 50 Deleting Virtual Disks.............................................................................. 53 Using VMware-cmd ................................................................................. 53 Changing Parameters within a VM with /proc ............................................. 57 Module 7: Performance Monitoring ........................................................58 Using ESXTOP to Monitor Virtual Machines................................................. 58 Tools to Monitor the Service Console......................................................... 64 Module 8: Upgrading/Patching ESX........................................................65 Pre-Upgrade Checks ............................................................................... 65 Custom Upgrading with a Tar-Ball ............................................................ 65 Upgrading to VMFS-3 from the Service Console .......................................... 67

Introduction
Purpose of this guide This guide is designed for people who already know ESX 3.x and VC 2.x quite well. Although it starts as a beginners guide initially, it pretty rapidly starts to assume very good knowledge of the system. I would recommend you get to grips with the GUI first, and feel comfortable with Vi-3 before attempting this guide. That said you might be wanting to carry out a discrete procedure from the command-line. So feel free to dip in find what your looking for and then dip out again! It is not a comprehensive guide to ALL the commands just the primary ones. I hope to make this guide gradually more comprehensive, and cover all new commands that useful. Ive deliberately not covered every single esxcfg command because not all of them are terrifically useful There are some big topics that I have yet to add to this guide this includes setting NTP and Active Directory authentication for the Service Console Please email at the email address at the beginning of this document if spot any errors Where possible I dont use the VI Client. I only use the VI client if there is no other way even if the VI client is easier. The reasoning behind this force the use of the command-line. Perhaps I should state why using the command-line might be useful But I Automation of tasks? Er, because everything else is broken and login in at the ESX host at the Service Console is your only option Because sometimes its quicker (sometimes its quicker using the mouse too!) Because like Everest, it is there!

prefer the mouse? Then this guide isnt for you Choose File, and close But seriously, what are commands only words that carry out instructions Who in Windows hasnt used the net use or net servicename stop/start command? I grew up on DOS/Windows3.x/Novell a mix of GUI and commands. For which I am very grateful. I am happy in both environments. Although I do find myself swearing when commands refuse to work, but that said I also swear at Windows just as much perhaps like Pete from Big Brother have Tourettes Syndrome. God knows when ever I watch Big Brother I feel uncontrollable the urge/need/desire to shout wankers at the top of my voice Hopefully this guide will reduce the amount of swearing you do at computers. You must remember they cant hear or understand your profanities and talking to inanimate objects is usually the first stage of madness

Anyway, I digress - Ive noticed as the GUI grew to dominate our environments my command-line skills took a hit. If you dont use it, you loose it as the saying goes. I dont intend my skills to be eroded by a little mouse. Of course they could be equally eroded by the keyboard.

Hardware I use a 192.168.2.x range on my network with 192.168.2.101 for esx1, 192.168.2.102 for esx2. My DNS server is 192.168.2.200. All IPs here are just example from my network range replace with your own This guide assumes you have at least 2 servers with 2CPUs at 1.4mgz each 2GB of RAM 2x36GB Hard Drives (SCSI Internal) 4x Network Cards (production in a bond, vmotion and Service Console eth0) This is the specification of my server an old Dell 1650 PowerEdge using PIII processors! My hardware isnt officially supported by VMware anymore. Anyway it still runs. But at some point I am going to buy two DL380, and re-use this hardware as my VC box and a NAS box I have VC set-up with a SQL Database. The layout of my VC looks like this before the upgrade:

Two Servers, Two Virtual Machine Groups (for me and my mate, Trevor) The domain name is rtfm-ed.co.uk lavericm-admin is the VC Administrator set at Server Farms instructor is a Virtual Machine Admin from the RTFM Farm (used when I teach with my hardware) baverstockt is a Virtual Machine User with rights only for Trevors VMs Group Software In my case VC Server 2.x and SQL 2000 (SP3) all runs on Windows 2003 with Service Pack 1. This was the most current release of Windows at the time of writing this document Warning: As ever my documents are released as is and without warranty

Hardware & Software As we at the Service Console you dont really need to have exactly what I have but it would help to have what I have or better Conventions in this Guide I use PuTTy to get Service Console sessions I use nano rather than vi to edit and save files mainly because Im no Linux guru, and I find nano more user friendly. Vi text editor is popular in the Linux community because it pretty much standard amongst every Linux distribution. Dont let anyone make you feel an idiot or small because you use nano. As long as you can successful manage the system thats all that counts in my book. Linux people, do NOT flame me on the merits of using VI. I will absolute NOT respond!

Any Major titles marked with Red indicates this section

is broken, released as is, and I intend to return to fix it. But got bored banging my head against a brick wall.

IMPORTANT:
To have your command-line changes reflected in the VI Client you must restart the hostd service on the ESX host. You can do this by using service mgmt-vmware restart

Change Log from 1.1 to 1.2 None

Module 1: Users and Rights

Note: In most case your rights will be governed by VirtualCenter and Active Directory However, at the Service Console the account database thats used is local to the Service Console After an upgrade of ESX from 2.x to 3.x root does have rights to connect remotely using SSH. After a clean installation the default is that root has no remote access via SSH. This to enforce traceability. User must logon as underprivileged users and elevate the rights to root. This change in rights is logged in /var/log/messages The easiest way to create a normal user for use with PuTTy would be the VI Client. You can point it at the ESX host and use the User & Groups tab to create a new user

Creating a New User

1. Logon at the Service Console as ROOT (physical or ILO) 2. Type: useradd lavericm-admin p ********

Elevating yourself to ROOT

1. Logon at the Service Console as your user account 2. Type su Note: This Switches User and assumes root, unless otherwise specified. The takes roots environmental settings (very important if you want to run any commands properly) 3. Type the password for root Note: You can also use a tool called sudo. This allows you to logon as an underprivileged user, and run commands with the ID of a privileged user

Disabling Auditing on ROOT (Not Recommended)

Note: Some applications do not support levitation to a higher plain - for example WinSCP. Sure you could use WinSCP to gain access as an ordinary user, but then you might lack permission to copy the files you need. If you try to logon as root, WinSCP will give you access denied. If you wish to disable the restriction on ROOT not being allowed direct access using SSH then carry out the following task. I wouldnt recommend doing this as you will loose enforcement of your audit trail. 1. 2. 3. 4. nano w /etc/ssh/sshd_config Locate: PermitRootLogin no Place a # in front of PermitRootLogin no like so: #PermitRootLogin no Exit Nano & Save the file

5. Restart sshd with service sshd restart

Miscellaneous User Managements Tasks

To change your password: passwd root List connected users w Note: This gives a result like so:

Module 2: Getting around & Getting help

Note: This module is for novices/newbies and people who have never used the Service Console before (sometimes referred to in the community as vmnix or the COS) The modules are NOT prescriptive so if you would like skip and move to something more interesting please feel free to do so This section covers all the basics of using SSH and puTTy Oh, it occurs to me you might not know what these are SSH is secure protocol (22) which allows you connect remote to an ESX host giving a command-line interface. PuTTy is a free SSH Client which is very popular in the community. This saves you using an ILO to get to the command-line environment The command-line environment is actually something called the BASH (Bourne Again Shell) Get help on a command man reboot or reboot -- help or reboot -? history | tail history c This works with both Linux & ESX commands. Q to quit Gives you short explanation. With | more this pauses the output Warning: Beware of h this actually triggers a reboot of your ESX host! History normally list your past commands, used with tail it will filter to show only the last 10 Bash has a history, not unlike doskey (except its in a file, not memory). Use this to clear your history of commands previously typed Same as CLS in a cmd prompt It is regarded as good practice to login with a VM Administrator account and then switch to being ROOT if required. The minus sign switches bash to ROOT and ROOTs environment. Without you would have the rights of root, but command might fail due to being in the incorrect path for the executable I find it especially important to know who I am in all situations. Failure to know who you are can result in surprising results. Id shows your group membership as well. The u switch indicates the UTC format is being used. Numbers equate to mm/dd/hh/mm/yy

View Last 10 commands Clear Bash History

Clear the screen Switch to being root

clear su -

Currently logged in user Change the Date/Time

whoami or id date u 0701180504

Page 8 RTFM Education

Display ESX Version Number Search Man pages by index Reboot ESX Server Shutdown Shutdown & Halt

vmware -v makewhatis man k partition Reboot shutdown now Server shutting down for maintenance shutdown h now server shutting down for maintenance

Should display information along this kind VMware ESX Server 2.5.0 build-11343 Creates an index of the man pages Searches manual pages for string of partition Reboots the server, does not ask are you sure are you really sure? Shuts down the VMs and stops remote access physical console is still accessible Shuts down the VMs; Stops Remote Access and does a fault halt of the system and volumes un-mounted

Page 9 RTFM Education

Module 3: File & Folder Management

Purpose List files List files with a pause List hidden files Full Path location Return to the root Return one directory up Go to home directory Type the contents of a file Search for string inside a file Use a command together with grep Edit the contents of a file Create/Delete a directory Delete a file Delete many files without prompts Syntax Example/Sample ls l ls l | more ls a l pwd cd / cd .. cd cat instructor.vmx less instructor.vmx grep lavericm /etc/passwd ls /etc -l | grep vmware nano w /etc/fstab mkdir /root/scripts rmdir /root/scripts rm filename rm test*.txt f rm instructor*.vmdk Notes In PuTTy this colour codes the files, and directories and shows Type, Permissions, Group, User, Size (b) Date, and filename The | is the pipe symbol commonly found where the \ is on a UK keyboard Shows files that are hidden files are made hidden if prefixed with a period, such as ./install Like what would see in a DOS command prompt You have to put a space between d and / Again, you need a space between cd and .. CD on its own returns you to the home directory of the current users Same as the TYPE command in DOS Works better with longer files use the keystroke [Q] to quit like man Here we are searching for a piece of text called lavericm in the file called passwd Here we are listing all the files in /etc that contain the string vmware -w disables word wrap and stops unwanted carriage returns Control+X, [ENTER] [ENTER] exits nano and saves a file Without file path, directory made relative to your path If your deleting lots of files rm f *.txt will delete all the txt files but will NOT prompt you The f forces a removal off all the files that begin test and have the txt extension without a prompt. DANGEROUS COMMAND! The second example would delete a virtual disk both its metadata file and its flat file

Page 10 RTFM Education

Delete contents of directory and all files and subdirectories Mount a CD at the Service Console Mount an ISO at the Service Console

Rm /root/vm-support.3630 r f

mount /dev/cdrom ls l /mnt/cdrom mkdir /mnt/isocd mount -o loop -t iso9660 -r /vmimages/esx3.iso /mnt/isocd

Here I am deleting the extracted contents of an untard vm-support script. r does a recursive delete of all files and folders within vmsupport.3630 without a prompt (-f) DANGEROUS COMMAND! Root only! Create directory for the mount point -o loop means mount the device as a block device -t iso9660 means its using the iso file system (as opposed to say Joliet) r means to mount read-only Be careful with the use of wildcards if you try to copy every file with *.* as you would with DOS/Windows this would miss out files that DID NOT have extension. So w2k3.iso would be copied but w2k3 would not This uses secure copy. You will prompted with some security warnings on the first copy for the first time. You will have manually type the password of the remote machine. Warning: You will need to enable firewall the SSH Client on the ESX where your files are located (source) but NOT where they are being copied to (destination) Move because effectively, any rename is move procedure. If you know the name of a file but cant remember where it is stored the whereis command is dead easy to use Find is much more powerful but can take longer based on your search criteria. Search here begins at the root / and is caseinsensitive search by using iname and *.conf would find every conf file note are required these are single quotes not double-quotes / is the search point, -mount volumes mounted, -mtime is the duration and print is the format for output to the screen This would find files off / taking up more than 10MG Works best with single files such a vmdk file Caution, automatically deletes original and adds a gz extension to

Copy a file

cp /vmimages/w2k3.iso /vmfs/volumes/nas_isos

Securely copy a file from one ESX server to another

scp /vmimages/w2kadvsrv-sp4.iso [email protected]=ed.co.uk:/vmimages/

Renaming a file/folder Find a file Find a file

mv w2k3.iso cdw2k3.iso whereis vmsnap.pl find / -iname '*.conf'

Find new files Find files of N size Compress a single file

find / -mount -mtime -1 -print find / -mount size +10240k gzip /vmfs/volumes/local-esx1-esx1/instructor/*.vmdk best

Page 11 RTFM Education

Uncompress a single file

gunzip /vmfs/volumes/local-esx1-esx1/instructor/*.vmdk

Compress Multiple files Uncompress Multiple files Change Ownership Change Group Membership

tar -czvf /backup/allfiles.tgz /vmimages

tar xzvf /backup/allfiles.tgz /vmimages

chown lavericm:lavericm instructor.vmdk chgrp lavericm instructor.vmdk

the file name As above but in reverse! There is also a utility called bzip2 and bunzip2 which use newer algorithms which offer better compression ratios and better performance This would backup all the files in the vmimages directory to a backup folder (c create, z compress, v- verbosely listed files being tard, use a file) This would restore all the files in the vmimages directory to a backup (x Extract, z uncompress, v- verbosely listed files being tard, use a file)folder Change the Owner of instructor.dsk to be lavericm, also changes user group to be lavericm as well Change the Group rights to be the User Group, in this case lavericm

Page 12 RTFM Education

Change Permissions

chmod 774 /home/lavericm R chmod 754 /home/lavericm R

Changing permissions on files shouldnt be needed very often as all files are created and owned by root (despite being created with VI Client with a Windows Logon) Change permissions on files using decimals to represent RWX. Used with the R switch it sets these permission recursively. 77 is short for RWX for the user and group, and 4 for R for others. Although the Numbers are expressed in decimal - they are effectively binary where 1 is Execute, 2 is Write and 4 is R. They are represented not unlike jumper settings on the back of a SCSI device so 000 would mean NOTHING 0 001 would mean X 1 010 would mean W 2 011 would mean WX 3 100 would mean R 4 101 would mean RX 5 110 would mean RW 6 111 would mean RWX 7 Sum is less good than md5sum. You need to compare copied file to known good original usually from a website Most frequently done to check integrity of downloads checkout WinMD5Sum if you have download to Windows VMDK are just extensions. During import/export process vmkfstools e or i does not rename. Head examines the contents of the file a report true format VMDK: standard input: x86 boot sector (Monolithic) VMDK: standard input: ASCII Text (Sparse or Metadata File) For older Vmware products look for COW in the printed string

Check integrity of a file As Above Interrogating File format

sum /vmfs/volumes/local-esx1-esx1/instructor.vmdk md5sum /vmfs/local-esx1-esx1/instructor.vmdk

head /vmfs/volumes/local-esx1/instructor/instructor.vmdk | file or head /vmfs/volumes/local-esx1/instructor/instructor-flat.vmdk | file -

Page 13 RTFM Education

To create a ISO file from mounted CD at the Service Console To List the last ten lines of a log file

dd if=/dev/cdrom of=/vmimages/w2ksp4.iso bs=32k

tail /var/log/messages

Not recommended, as there is no check done on the integrity of the ISO file WinImage - http://www.winimage.com Tail can take parameters to show more or less information with long files like logs it can be easier to handle than cat

Page 14 RTFM Education

Module 4: Networking
Note: Networking involves the use of command esxcfg-vswitch, esxcfg-vswif0 and esxcfg-vmknic. These are quite involved commands used in a particular order to achieve the results your are looking for. So Ive decide to write more of a step-by-step guide than just a command-list with a brief explanation esxcfg-vswitch is the main command and it has a mix of parameter in lower and upper-case. Lower-case parameter manipulate the switch, where as upper-case switches manipulate the portgroup. So to add a switch its a and add a upper-case A adds a portgroup In one way this is nice but its incredibly easy to create a switch rather than portgroup attached to a switch and I have done this a few times

Viewing your Switches & Service Console Networking

1. To view your switches type the command: esxcfg-vswitch l

Note: This shows me I have one vSwitch (vSwitch0) using one NIC (vmnic0) with one portgroup called Service Console which is not using VLAN. 2. To view your Service Console network settings esxcfg-vswif l

Note: Nothing to state here but I think its interesting that it doesnt show me my all important default gateway settings which would have been nice 3. To View your network cards vmnic, pci (b:s:f), driver, link, speed, duplex and description: esxcfg-nics l

Creating a vSwitch (Internal)

1. To create a new switch type:

Page 15 RTFM Education

esxcfg-vswitch a vSwitch1 2. Then add a portgroup esxcfg-vswitch A internal vSwitch1 Note: Lower-case -a for adding a switch, and upper-case -A for adding a portgroup 3. If you run the command esxcfg-vswitch l you will see this information

4. If you wish to see this reflected in the VI Client then type: service mgmt-vmware restart Note: This is how my switch configuration looks now

Creating a vSwitch (Single NIC)

1. In this case we will create a new switch and port group in one line with: esxcfg-vswitch a vSwitch2 A production 2. Next patch a NIC to the vSwitch with esxcfg-vswitch -L vmnic1 vSwitch2 Note: Again this is a case-sensitive option. l lists switches, whereas L links Page 16 RTFM Education

nics to switches 3. If you run the command esxcfg-vswitch l you will see this information

4. If you wish to see this reflected in the VI Client then type: service mgmt-vmware restart Note: This is how my switch configuration looks now

Creating a vSwitch (Multiple NICs)

Note: It is very easy to create a NIC-Team just re-run the previous command, with a different NIC with

Page 17 RTFM Education

1. esxcfg-vswitch -L vmnic2 vSwtch1 2. If you run the command esxcfg-vswitch l you will see this information

3. If you wish to see this reflected in the VI Client then type: service mgmt-vmware restart Note: This is how my switch configuration looks now

Deleting a Switch from vSwitch

Note: Im now running out NICS for the next part of this guide. So I am going to blow away my vSwitch2 to free up my NICs 1. Type the command esxcfg-vswitch d vswitch2

Page 18 RTFM Education

Note: Notice how it doesnt ask are you sure, but then again neither does the VI client.

Creating PortGroups for VLAN Networking NIC Team

1. We have to create a switch, a portgroup for each VLAN, allocate our NICs and then set the VLAN ID esxcfg-vswitch esxcfg-vswitch esxcfg-vswitch esxcfg-vswitch esxcfg-vswitch esxcfg-vswitch a vSwitch2 A accounts vSwitch2 A rnd vSwitch2 A sales vSwitch2 L vmnic1 vSwitch2 L vmnic2 vSwitch2

2. The next part is to set the VLAN id for each network (account, rnd and sales) esxcfg-vswitch v 10 p accounts vSwitch2 esxcfg-vswitch v 20 p rnd vSwitch2 esxcfg-vswitch v 30 p sales vSwitch2 3. If you run the command esxcfg-vswitch l you will see this information

4. If you wish to see this reflected in the VI Client then type: service mgmt-vmware restart Note: This is how my switch configuration looks now

Page 19 RTFM Education

Creating Vmkernel Switches

Note: This does seem a bit limited I cant see away of enabling the VMKernel switch for Vmotion Nor can I see how I would set the default gateway for the switch either You need a good name for the portgroup because the esxcfg-vswitch command just shows as any ordinary switch You need to use esxcfg-vmknic l to see it

1. First create the Switch, Portgroup and Assign a NIC esxcfg-vswitch a vSwitch3 esxcfg-vswitch A VM Kernel vSwitch3 esxcfg-vswitch L vmnic3 vSwitch3 2. Next use the esxcfg-vmknic command to add in a VM Kernel NIC and set the IP and Subnet Mask esxcfg-vmknic a VM Kernel i 192.168.2.202 n 255.255.255.0 3. Then set the vmkernel default gateway with esxcfg-route 192.168.2.1 4. If you run the command esxcfg-vswitch l you will see this information

Page 20 RTFM Education

If you wish to see this reflected in the VI Client then type: service mgmt-vmware restart Note: This is how my switch configuration looks now

Page 21 RTFM Education

Changing your Service Console IP Settings

Note: You can do this through a putty session but clearly unless you have more than one vswif interface (with a different IP address) you will get disconnected It is entirely possible to have two vswif interface on two separate IP address and use one connection to change the IP address of the other You do run the risk of losing SSH connectivity if you screw up so perhaps doing this through your ILO is a safer-bet and less hard work You might also wish to include a new DNS entry for your ESX host before you make this change 1. To view your current IP and Netmask type: esxcfg-vswif l

2. To change your IP and Subnet Mask type: esxcfg-vswif i 192.168.2.203 n 255.255.255.0 vswf0 Page 22 RTFM Education

Note: You change your default gateway by editing nano w /etc/sysconfig/network and then restart your networking with services network restart Your DNS settings are located in nano w /etc/resolv.conf

Setting the Speed and Duplex of NICs

Note: We have always been able to change the speed & duplex of the vmkernel NICs (ESX 2.x MUI) but in the past the only way to change the Service Console speed and duplex was by editing the /etc/modules.conf file You can now change both the Service Console and the vmkernel nics through the GUI There could be a chicken-egg/catch22 situation here though. If your Service Console NIC mis-negs its speed/duplex you maybe unable to connect with the VI client to change it. Thats when knowing the Service Console commands come in handy. This can also be useful to reassign a 100Mps card to a Service Console and then fix its speed/duplex 1. Workout which NIC has been assigned to the portgroup Service Console with esxcfg-vswitch l

2. To view you current speed & duplex: esxcfg-nics l

3. To set the speed/duplex of vmnic0 to 100Mps/Half-Duplex type: esxcfg-nics -s 100 -d half vmnic0 4. To reset to auto-neg esxcfg-nics -a vmnic0

Recreating your vswif0 Interface

Note: OK, suppose something horrible goes wrong and you loose your vswif0 interface this is how you would recreate it

Page 23 RTFM Education

1. Logon locally to the ESX host or use your ILO card (do you have choice!) 2. Create a new switch esxcfg-vswitch -a vSwitch0 3. Create a new portgroup esxcfg-vswitch -p "Service Console" vSwitch0 4. Assign a NIC esxcfg-vswitch L vmnic0 vSwitch0 5. Assign a vswif interface and set its ip/sn: esxcfg-vswif a vswif0 -p "Service Console" i 192.168.2.102 n 255.255.255.0

Removing a NIC from vSwitch

esxcfg-vswitch U vmnic vSwitch2

Deleting a PortGroup from vSwitch

esxcfg-vswitch D production vSwitch

Managing the ESX Firewall

Note: Managing the firewall by the VI Client is real easy its a tick of box style interface. If you use the GUI interface and then query with the commandline tool you get friendly information about what enabled. If you purely use the command-line tool you will just get TCP port numbers and directions (outgoing and incoming They GUI tool also has some handy built-in friendly name for popular applications vendor specific agents like CommVault Dynamic/Static and many others By default 902 (VirtualCenter, uses TCP and UDP!), 80 (Web Access welcome page), 443 (Web Access Login page) and 22 (SSH) are enabled by default In additions there are some ports enabled for EMC AAM (Automated Availability Manager) and CIM (Common Information Model) developed by the DMTF, is a very broad approach to the management of systems and networks Other Ports worth knowing about: 903 2049 2050-5000 8042-8045 3260 8000 2700 Console sessions from Workstations to VMs NFS Storage (VM Kernel interfaces only) HA and EMC Autostart Manager (Between ESX Hosts iSCSI (VM Kernel interfaces only) VMotion Requests (VM Kernel Interfaces only License Server communications Page 24 RTFM Education

A bit like in ESX 2.x it is possible to set 3-level of security (high, medium and low) High Medium Low incoming/outgoing blocked incoming blocked, outgoing not blocked firewall off, no blocking of incoming/outgoing traffic

Viewing your Firewall Settings 1. Type the command esxcfg-firewall q outgoing esxcfg-firewall q incoming Note: You can also use esxcfg-firewall q on its own. This gives you lots of stuff most useful at the bottom:

Changing Your Security Level Note: If you wanted to weaken your security to medium you could use 1. Type the command esxcfg-firewall --allowOutgoing blockIncoming Note: You should get warnings like so: 2006-07-06 14:39:53 (1965) WARN : Setting firewall default /firewall/blockOutgoing to 0 2006-07-06 14:39:53 (1965) WARN : Setting firewall default /firewall/blockIncoming to 1 Enabling a Single Service/Client/Agent Note: If you want to SSH from an ESX host to another ESX host or SCP from an ESX host to another you need to enable the SSH Client on port 22 If you do this via the GUI and then do a esxcfg-firewall q you will see friendly information like so:

Page 25 RTFM Education

Incoming and outgoing ports blocked by default. Enabled services: AAMClient CIMSLP LicenseClient sshServer CIMHttpsServer CIMHttpServer sshClient vpxHeartbeats dgfdfgdsdsdsd Opened ports: To do the same from the command-line you would type the command: esxcfg-firewall -e sshClient Note: To disable esxcfg-firewall -d sshClient Enabling non-Standard Ports Note: Perhaps there is an application or service which is not listed for use with e or d Or the port numbers have been changed to non-standard ports It is possible to open a specific port by number, transport (udp/tcp) and direction (in/out) To enable port 22 outbound from the server type the command: esxcfg-firewall -o 22,tcp,out,ssh Note: The ssh at the end is a friendly label. If I run esxcfg-firewall q again at the bottom it states: Incoming and outgoing ports blocked by default. Enabled services: AAMClient CIMSLP LicenseClient sshServer Page 26 RTFM Education

CIMHttpsServer CIMHttpServer vpxHeartbeats dfsdfdfdfsdfsdf Opened ports: ssh : port 22 tcp.out

Fixing ESX Networking after an Upgrade

Note: If you have read my upgrade guide to ESX 3.x you will know that networking looks significantly different after an upgrade rather than a clean install The following shows how to clean up networking using the Service Console commands This is how my upgraded ESX server networking looks like

So what am I going to do with this?

Page 27 RTFM Education

What I am going to do is the following: Remove the VMotion Switch0 altogether thus freeing up vmnic0 Remove the Legacy vmnet_0 portgroup from vSwitch1 Remove the Legacy vmnic2 portgroup from vSwitch2 Remove vSwitch3 Create a new vswif interface on a new switch called vSwitch0 using vmnic0 Add vmnic1 to vSwitch2 Patch all my VMs to the Production on vSwitch2 Create with a VM Kernel Port (vSwitch3)

Removing VMotion Switch0 esxcfg-vswitch d vSwitch0

Remove Legacy vmnet_0 portgroup from vSwitch1 esxcfg-vswitch D Legacy vmnet_0 vSwitch1

Remove Legacy vmnic2 portgroup from vSwitch2 esxcfg-vswitch -D "Legacy vmnic2" vSwitch2

Remove Old vswif Interface and Remove vSwitch3 Disable vswif0 interface, and remove from the system You must disable the vswif interface before you delete the switch it is attached to Do this at the physical console via an ILO esxcfg-vswif d vswif0 esxcfg-vswitch d vSwitch3

Create a new vswif1 interface esxcfg-vswitch -a vSwitch0 esxcfg-vswitch -A "Service Console" vSwitch0 esxcfg-vswitch L vmnic0 vSwitch0 esxcfg-vswif a vswif0 -p "Service Console" -i 192.168.2.x n 255.255.255.0 Note: You will get this message [DateVnic warning] Generated New MAC Address, 00:50:56:XX:XX:XX for vswif0 Nothing to flush Note: I found my new vswif0 interface was already enabled. But if doesnt for you might like to try esxcfg-vswif d vswif0 Create a VM Kernel Port esxcfg-vswitch -a vSwitch3 esxcfg-vswitch -A "VM Kernel" vSwitch3 esxcfg-vswitch -L vmnic3 vSwitch3 esxcfg-vmknic -a "VM Kernel" -i 192.168.2.x -n 255.255.255.0 esxcfg-route 192.168.2.1

Page 28 RTFM Education

Typical vSwitch Errors

Getting your case-sensitivity muddled up! o -a add vSwitch o -A add portgroup o -l list vSwitches o -L link vmnic o -d delete vSwitch o -D delete portgroup Trying to change things that are in use Failed to remove vswitch: vSwitch3, Error: PortGroup "Legacy eth0" on VirtualSwitch "vSwitch3" is still in use: 1 active ports, vswif0 or Legacy vmnic2, Error: Unable to delete portgroup "Legacy vmnic2", for the following reasons: 1 active ports

Page 29 RTFM Education

Module 5: Configuring Physical Storage

Note: This guide assumes you already have set-up NAS and iSCSI storage correctly and that you have created a VM Kernel vSwitch/Port Group outlined in the previous module and now you wish to connect to it This module repeats some of the vmkfstools command listed earlier sorry about this repetition but this gives more detail than simple explanation of the commands

Configuring SAN Storage

Note: I dont have a SAN Someone give me an MSA1000 to play with? ;-) I promise to write some excellent documentation!!!

Configuring NAS Storage

Note: Remember you must have a VM Kernel Switch to do this and as authentication is driven through the Service Console NIC there must connectivity to the NAS via the Service Console NIC My NFS server is called nfs1.rtfm-ed.co.uk It has 3 exports on it /isos /templates/ /nas-vms an Mounting NAS Exports/Shares 1. Confirm you can ping your nfs1 by its name or ip address 2. Type the command: esxcfg-nas -a isos -o nfs1.rtfm-ed.co.uk -s isos Note: The command should respond with: Connecting to NAS volume: isos isos created and connected. Note: -a to add followed by the friendly label, -o to specify the nfs server, and s to specify the export/share I then continued to add in /template and nas-vms like so: esxcfg-nas -a templates -o nfs1.rtfm-ed.co.uk -s templates esxcfg-nas -a nas-vms -o nfs1.rtfm-ed.co.uk -s nas-vm Listing NAS Exports/Shares 1. Type the command: esxcfg-nas l Note: You should receive a response like so: Page 30 RTFM Education

isos is isos from nfs1.rtfm-ed.co.uk mounted nas-vms is nas-vms from nfs1.rtfm-ed.co.uk mounted templates is templates from nfs1.rtfm-ed.co.uk mounted Restoring NAS Connections Note: If the NAS become unavailable you may have to force a restore the mount This sometimes happens to me if start-up my ESX hosts before the NAS is up 1. Type the command esxcfg-nas r 2. Followed by esxcfg-nas -l

Connecting to iSCSI Storage (Software)

Note: My iSCSI box is Fedora Core 5 with iSCSI Enterprise Target installed and configured, its name is iscs1.rtfm-ed.co.uk iSCISi has 2 LUNs available, it supports dynamic discovery and its IP is 192.168.2.210. Use the iqn of iqn.2006-06.uk.co.rtfm-ed:storage.lvm on this iSCSI Target The esxcfg-swiscsi is more limited than the esxcfg-nas command. Although you can do many things you cannot configure the IP address and port number of the iSCSI server o o o o o e enable d disable q query if the adapter is enabled or disabled s force a scan k forcibly remove iscsi sw stack

Enabling the iSCSI Adapter 1. Type the command: esxcfg-swiscsi e Allowing software iSCSI traffic through firewall... Enabling software iSCSI... /usr/sbin/vmkload_mod /usr/lib/vmware/vmkmod/iscsi_mod.o Using /usr/lib/vmware/vmkmod/iscsi_mod.o Module load of iscsi_mod succeeded. Note: This will enable the iSCSI Software Adapter like so:

Page 31 RTFM Education

Setting the iSCSI Target IP with Discovery Mode Note: Remember the Software iSCSI adapter does NOT support static discovery 1. Type the command: vmkiscsi-tool -D -a 192.168.2.210 vmhba40 Note: -D sets the mode of Discovery, rather than static. a is used to add an iSCSI Target, followed by IP address of iSCSI Device and the HBA that will be used. vmkiscsi-tool is a powerful little utility with lots of options although documentation is a little bit thin on the ground. I intend to return to this section when I have more time to put some more flesh around it 2. To list the targets configured use: vmkiscsi-tool -l -T vmhba40 Note: This should report:

Forcing a Rescan of the iSCSI Adapter 1. You might wish to list the current LUNs/Disk currently visible with ls l /vmfs/devices/disks

Page 32 RTFM Education

Note: I have two internal disks on an internal SCSI Controller (vmhba0:0 and vmhba0:1) I also a have JBOD with six disks inside (vmhba1:1-6) 2. Force the rescan with esxcfg-swiscsi s Note: This should produce this result Scanning vmhba40... Rescanning vmhba40...done. On scsi2, removing:. On scsi2, adding: 0:0 0:1. 3. If we run the ls command again we can see the 20GB LUNs from my iSCSI box

Creating a VMFS Volume

Using FDISK to Create Partition Note: Generally we create one VMFS partition per disk/LUN 1. Logon to the Service Console as ROOT type: fdisk /vmfs/device/disks/vmhba40:0:0:0 Note: Under ESX 2.x we would have need to know how Linux address this disk/LUN with /dev/sd? Syntax. This is no longer required. 2. 3. 4. 5. Type N, to create a new partition Type P, for a Primary partition Choose 1 for the Partition Number Accept the defaults for First Cylinder and Last Cylinder Note: This creates a partition which uses all of the free space on the disk adjust the last cylinder value to adjust size of partition - also fdisk defaults to being a Linux type which in this case is not good for our VMFS partition

Page 33 RTFM Education

6. Type P, to print out the partition table (make a mental note of the partition number here) 7. Type T, to change the File System Type 8. Type the Hex Code of fb: (the code for a VMFS partition) Note: This should give the result of Changed system type of partition 1 to fb (Unknown) Hex codes tell the system what file system partition will support 07 NTFS, 82 Linux Swap, 83 Linux File System, FB for VMFS and FC for VMware Core Dump. At this stage the VI Client would identify the partition as VMFS unformatted. 9. Type W, to write your partition table changes to the hard drive it will give you this status information: Note: The partition table has been altered! Calling ioctl() to re-read partition table. Syncing disks. Note: If you re-run fdisk /vmfs/device/disks/vmhba40:0:0:0 and choose P to display the partition table

Format & Label the VMFS Partition 1. Logon to the Service Console as ROOT 2. To format the new partition with the VMFS file system type: vmkfstools C vmfs2 vmhbaA:T:L:V In my case the hard disk partition is located on the Adapter 40, TARGET 0, LUN 0 on Volume/Partition 1. So I would type: vmkfstools -C vmfs3 S iscsi-lun0 /vmfs/devices/disks/vmhba40:0:0:1 Warning: Please note it is an UPPERCASE C you type here. A lowercase c creates a vmdk file. You can specify b flag to set a block-size. You would need to do this if you thought any one of your virtual disks were going to be greater than 256GB in size. From the PDF Guide 1MB, 2MB, 4MB, or 8MB. When entering a size, indicate the unit type by adding a suffix of m or M. The unit type is not case sensitive. vmkfstools interprets either m or M to mean megabytes Page 34 RTFM Education

2m allows 512GB max file size 4m allows 1024GB max file size 8m allows 2048GB max file size Note: This should give the result like so: Creating file system on "vmhba40:0:0:1" with blockSize 1048576 and volume label "none". Successfully created new volume: 44a7bcf0-8b87cb86-940300065bec0eb6 Note: I repeated this for my other iSCSI LUN 1 fdisk /vmfs/devices/disks/vmhba40:0:1:0 vmkfstools -C vmfs3 S iscsi-lun1 /vmfs/devices/disks/vmhba40:0:1:1 Note: If we re-run ls l /vmfs/volumes will see this information

and in the VI client

Changing the Volume Label

Note: To do this you need to know the UUID of your VMFS volume 1. Type the command ls l /vmfs/volumes

Page 35 RTFM Education

Note: The UUID is the value in blue. So the try and unchanging volume id is actually something like: /vmfs/volumes/44a38c72-156b2590-be15-00065bec0eb7 2. Type the command: ln -sf /vmfs/volumes/44a38c72-156b2590-be15-00065bec0eb7 /vmfs/volumes/esx1-local Note: The LN makes symbolic links (a bit like shortcuts if you like) s makes/change a symbolic link, -f over-writes the existing symbolic link

Viewing VMFS Volumes/Partition Information

1. Type the command: vmkfstools -P /vmfs/volumes/local-esx1 Note: This returns information like so: VMFS-3.21 file system spanning 1 partitions. File system label (if any): local Mode: public Capacity 73282879488 (69888 file blocks * 1048576), 41889562624 (39949 blocks) avail UUID: 44a6c956-66056236-c671-00065bec0eb6 Partitions spanned: vmhba1:4:0:1 Note: For NAS data store it would look like this: NFS-1.00 file system spanning 1 partitions. File system label (if any): templates Mode: public Capacity 80023715840 (19537040 file blocks * 4096), 25076690944 (6122239 blocks) avail UUID: ab184e34-6f68911d-0000-000000000000 Partitions spanned: nfs:templates

Page 36 RTFM Education

Viewing Available Physical Disk Space

Note: 1. Logon to the Service Console as ROOT 2. Type: df h (the linux method) or vdf h (the VM Kernel Method)

Note: The partition table for /sda does NOT follow the recommendations for ESX 3.x this partition table was taken for an ESX 2.x server which was upgraded Note: If you wish to see how the vmhba syntax relates to /dev/sd? You will find that you can still use old ESX 2.x command vmkpcidivy -q vmhba_devs. However, it has been hugely depreciated and you should instead esxcfg-vmhbadevs q vmhba0:0:0 /dev/sda vmhba0:1:0 /dev/sdb vmhba1:2:0 /dev/sdc vmhba1:3:0 /dev/sdd vmhba1:4:0 /dev/sde vmhba1:5:0 /dev/sdf vmhba1:6:0 /dev/sdg vmhba40:0:0 /dev/sdh vmhba40:0:1 /dev/sdi

Page 37 RTFM Education

Note If you want to see all of the information together vmhba ID, /dev name, LVM id and VMFS volume label you can use this perl script #!/usr/bin/perl @array = `/usr/sbin/esxcfg-vmhbadevs -m`; foreach (@array) { ($vmk, $cos, $uuid) = split; ($tmp, $label) = split (/:/, `/usr/sbin/vmkfstools -P /vmfs/volumes/${uuid} | grep "File system label"`); print "$vmk \t $cos \t $uuid \t $label\n"; } This produces an output like so:

Note: This only shows me my VMFS, not my NAS datastores

Mounting and U-mounting VMFSs

Note: By default all VMFS volumes are mounted at start-up under the /VMFS mount point. Should you accidentally or deliberately umount these partitions there is a method for umount and mounting them manually. Remember there can be no running VMs on the VMFS because SCSI Reservation would stop access 1. Logon to the Service Console as ROOT 2. To umount the VMFSs type: umount /vmfs 3. To remount the vmfss type: mount t vmfs vmfs /vmfs

Managing Storage Devices

Note: You can use a couple commands to handle storage devices. Some of these have already been mentioned like the command esxcfg-swiscsi -s to rescan the iSCSI Adapter(s). For those of you who come from ESX 2.x background it looks like cosrescan.sh has been discontinued, its functionality has probably been integrated into the s switch Forcing a rescan of Fibre Channel Device 1. Type the command vmkfstools s vmhba1 Note: Page 38 RTFM Education

You can also use esxcfg-rescan vmhbaN to be rescanned as well which give this kind of output: Rescanning vmhba1...done. On scsi1, removing: 2:0 3:0 4:0 5:0 6:0. On scsi1, adding: 2:0 3:0 4:0 5:0 6:0. Managing SCSI Reservations of LUNs Note: Occasionally, things go wrong. Such is life. One example is when through a configuration error a SCSI reservation (lock) is put on a LUN, which subsequently doesnt get released. Until a LUN is released you wont be able manage the file system and other ESX host will not be very happy bunnies. Typically, these kind of reservation problems happen in clustering scenarios This locking is controlled by the L switch on vmkfstools From the Server Configuration guide: -L reserve . Reserves the specified LUN. After the reservation, only the server that reserved that LUN can access it. If other servers attempt to access that LUN, they will get a reservation error! -L release . Releases the reservation on the specified LUN. Any other server can access the LUN again. -L lunreset Resets the specified LUN by clearing any reservation on the LUN and making the LUN available to all servers again. The reset does not affect any of the other LUNs on the device. If another LUN on the device is reserved, it remains reserved. -L targetreset . Resets the entire target. The reset clears any reservations on all the LUNs associated with that target and makes the LUNs available to all servers again. -L busreset . Resets all accessible targets on the bus. The reset clears any reservation on all the LUNs accessible through the bus and makes them available to all servers again This command use the device parameter so uses the /vmfs/devices/disks/vmhbaA:T:L:P syntax

Using USB storage Devices

Note: By default USB device drivers are loaded automatically This can be disabled by modifying /etc/modules.conf and remarking out with # the line which begins alias usb-controller usb-ohci If someone has done this then you can load the modules (drivers) manually and access the device as needed using the insmod command You can read but not write to a NTFS partition on a removable device

Page 39 RTFM Education

FAT32 is supported for read/write. There are some free EXT3 drivers for Windows (the one I use currently only works on NT4, W2K and WXP, not W2K3) You might prefer to use EXT3 if you wish to export virtual disks in their monolithic format (on VMFS partition) into a the COW format (copy-onwrite) which will take up less physical space on the disk In my case it was dead easy I plugged the USB device into the server. It came up with a message and said it allocated the id of /dev/sdc to the disk. So all I had to do was mount it with mkdir /mnt/usbdisk mount /dev/sdc1 /mnt/usbdisk Note: If this doesnt happen to you then you MIGHT find the stuff below useful.

Systems using the USB-UHCI device driver with the USB 2.0 interface can cause ESX Server to show a false warning message during the boot sequence. The error looks like this on boot-up
Mar 29 11:04:10 vmserver1 rc.sysinit: Initializing USB controller (usb-uhci): succeeded Mar 29 11:04:10 vmserver1 modprobe: Hint: insmod errors can be caused by incorrect module parameters, including invalid IO or IRQ parameters Mar 29 11:04:10 vmserver1 modprobe: /lib/modules/2.4.9vmnix2/kernel/drivers/usb/usb-ohci.o: init_module: No such device Mar 29 11:04:10 vmserver1 modprobe: /lib/modules/2.4.9vmnix2/kernel/drivers/usb/usb-ohci.o: insmod /lib/modules/2.4.9mnix2/kernel/drivers/usb/usb-ohci.o failed Mar 29 11:04:10 vmserver1 modprobe: /lib/modules/2.4.9vmnix2/kernel/drivers/usb/usb-ohci.o: insmod usb-ohci failed Mar 29 11:04:10 vmserver1 rc.sysinit: Initializing USB controller (usb-ohci): failed

Note: This was taken from ESX 2.x server It is safe to ignore this message, but if you want to configure your system so that this warning does not appear the next time you boot your ESX Server machine, follow these steps Link: Answer ID 1659 You might find these other article related to USB and IRQ sharing which you might find interesting 1651 - Using USB Storage Devices on the ESX Server Service Console 1290 - IRQ Sharing May Impact Performance 1326 - Using USB When ESX Server Disables It If you want to use a USB hard-drive to copy files from the ESX server to it and you have these errors -then a general work around is to boot to a Knoppix Boot CD and do it that way. Remember if you are copying virtual disks these will have to be exported to the COW format first. If you have no errors like me then this how you can go about using USB

Page 40 RTFM Education

hard-drives and devices To list what type of USB host controller interface you have. There are 3 of USB host controller interface types (OHCI, UHCI and EHCI) You can use the command below to find out yours: lspci -v | grep HCI Note: My server returned: 00:0f.2 USB Controller: ServerWorks OSB4/CSB5 OHCI USB Controller (rev 05) (prog-if 10 [OHCI]) Subsystem: ServerWorks OSB4/CSB5 OHCI USB Controller so I know my server I know I need to load the OHCI driver, not the UHCI or EHCI driver 1. Logon to the Service Console as ROOT 2. [Optionally] Load up the USB Device Drivers insmod usbcore insmod usb-storage modprobe usb-ochi or modprobe usb-uhci or modprobe ehci 3. Use the dmesg to print a out a list of active devices and scroll up to locate Initializing USB Mass Storage Driver.

Note: The system assigns a SCSI device ID to the USB device (even though its like to be IDE Laptop Disk if its a portable hard-drive). The critical bits are reference to the SDG. This tells me the USB device has been added to the end of all my other SCSI disks (sda, sdb, sdc). This will help me in the next stage which is creating a mount point and mounting the partition on the USB disk. Adding additional hard-drives to the system can upset this allocation of sdn. 4. Create a mount point with

Page 41 RTFM Education

mkdir /mnt/usb 5. Mount the first partition on the disk with mount /dev/sdg1 /mnt/usb Note: If you unsure about the partition scheme on the disk you can use fdisk l /dev/sdg to print to the console the partition table 6. List Files and Start using the disk with ls l /mnt/usb 7. Umount the USB device when finished with cd / umount /mnt/usb Note: Even after unmounting the drive. If I should down the ESX server with the USB Storage attached I get this worrying message which could be benign
scsi: device set offline not ready or command retry failed after bus reset : host 2 channel 0 id 0 lun 0 i/o error: dev 08:62, sector 4168 i/o error: dev 08:62, sector 4168 i/o error: dev 08:62, sector 0

Page 42 RTFM Education

Module 6: Managing Virtual Machines

IMPORTANT: Even changes made inside a virtual machine whilst it is powered off, will require a restart of the hostd service with service mgmt-vmware restart

Creating Virtual Disks

1. Power off your target virtual machine 2. To create a virtual disk use the vmkfstools command: vmkfstools a lsilogic -c 10240m /vmfs/volumes/localesx1/instructor/instructor_1.vmdk Note: This is lower-case c which creates a file, upper-case C creates a vmfs file system. Here I am following the naming convention that VC would normally be applied. The first disk would instructor.vmdk, and subsequent disks would be serialised with instructor_1.vmdk and instructor_2.vmdk This actually creates two files instructor_1.vmdk (metadata, of a couple of KB) and an instructor_1-flat.vmdk which is 10GB in size. I could have used 1g instead of 10240m on the command-line The a option allows users to indicate which device driver should be used to communicate with the virtual disk. Failure to set this could cause a question in the virtual machine on power-up depending on what SCSI Adapter controller is used in the VM. Personally, I always use LSILogic as the driver unless it is unsupported in the guest OS I am working with. The default option when creating new virtual disks is zeroedthick. A zeroed thick disk has all space allocated at creation time, and this space is wiped clean of any previous contents on the physical media. There are other options under the d switch you may wish to investigate. The only disk formats that can be used for NFS are thin and 2gbsparse. By default, files and virtual disks are stored in thin format on NFS servers, with blocks allocated on demand as needed. No other options, including RDMs, are supported on NFS. 3. Next we need to add this into the virtual machine, we can do this by editing the VMX file nano w /vmfs/volumes/local-esx1/instructor/instructor.vmx and the lines: scsi0:1.present = "true" scsi0:1.fileName = " instructor_1.vmdk" scsi0:1.deviceType = "scsi-hardDisk" Note: scsi0:1 means it is the next disk after the boot disk, on the first virtual SCSI adapter. True means the device is connected, filename the virtual disk (metadata file only) and the device type indicates this virtual SCSI

Page 43 RTFM Education

disk, not a virtual IDE drive. 4. Save you VMX file and Exit nano 5. Power on your virtual machine with vmware-cmd l Note: Thats a lower-case L for lima. This produces a list of registered VMs on the ESX host vmware-cmd /vmfs/volumes/44a38c72-156b2590-be1500065bec0eb7/instructor/instructor.vmx start Note: In the past we used to be able to use the friendly volume label. This is no longer supported as outline in KB 2122

Creating and Managing RDMs

Note There are two types of RDM virtual compatibility mode and physical compatibility mode These are normally set by radio button option in the VI Client On the command-line the way these are specified are with two different switches Main uses of RDMs are for: Accessing existing data on a SAN/iSCSI Required for physical to virtual clustering Required for clustering-across-boxes RDM for Virtual Compatibility 1. You may wish to check before you begin on which LUNs your ESX server has access to with: ls -l /vmfs/devices/disks 2. Type the command: vmkfstools a lsilogic -r /vmfs/devices/disks/vmhba40:0:0:0 /vmfs/volumes/local-esx1/instructor/instructor_1.vmdk Note: This creates a RDM metadata file in /vmfs/volumes/local-esx1 using my iSCSI Lun on vmhba40:0:0:0. It creates two files as in the c example a metadata file called instructor_2.vmdk and a media_1-rdm.vmdk file. VMware automatically append rdm to the end of the file for us! Note: The metadata file MUST be stored on a VMFS partition it cannot reside on a NAS DataStore Note: You do have to specify the last 0 in vmhba40:0:0:0. This last 0 indicates you wish to use the entire LUN. You cannot create a RDM file to a specific

Page 44 RTFM Education

partition WITHIN a LUN 3. Next we need to add this into the virtual machine, we can do this by editing the VMX file nano -w /vmfs/volumes/local-esx1/instructor/instructor.vmx and the lines: scsi0:2.present = "true" scsi0:2.fileName = " instructor_1.vmdk" RDM for Physical Compatibility Note: Physical compatibility (required for physical to virtual clustering) is set up in a very similar way. With physical compatibility the SCSI Reservations/Filtering normally imposed by the vmkernel are loosened such that other systems that also want to impose SCSI reservations will work (such as clustering systems) This is sometime referred to as a Pass-Through RAW Device Mapping In this case the switch is z vmkfstools a lsilogic -z /vmfs/devices/disks/vmhba40:0:1:0 /vmfs/volumes/local-esx1/instructor/instructor_2.vmdk Note: In this case the system labels the corresponding file instructor_2rdmp.vmdk. So we know this is a RAW Disk Mapping file with physical compatibility Viewing RDM Information Note: In the past it wasnt easy (impossible?) to view the contents of an RDM file (I guess you could have used vmkfstools l to list files and see which were virtual disks and which were RDMs). There is now a switch to query the metadata file which then reports information 1. vmkfstools q /vmfs/volumes/local-esx1/instructor_1.vmdk Disk instructor_1.vmdk is a Non-passthrough Raw Device Mapping Disk Id: vml.010000000020202020564952545541 Maps to: vmhba40:0:0:0 2. vmkfstools q /vmfs/volumes/local-esx1/instructor_2.vmdk Disk instructor_2.vmdk is a Passthrough Raw Device Mapping Disk Id: vml.010001000020202020564952545541 Maps to: vmhba40:0:1:0 Note: You can also cat the contents of the metadata file as well but the VM MUST be powered off first! cat /vmfs/volumes/local-esx1/instructor_2.vmdk

Page 45 RTFM Education

# Disk DescriptorFile version=1 CID=0a8fee69 parentCID=ffffffff createType="vmfsRawDeviceMap" # Extent description RW 8388608 VMFSRDM "instructor_2-rdm.vmdk" # The Disk Data Base #DDB ddb.toolsVersion = "7172" ddb.adapterType = "lsilogic" ddb.geometry.sectors = "63" ddb.geometry.heads = "255" ddb.geometry.cylinders = "522" ddb.virtualHWVersion = "4" cat /vmfs/volumes/local-esx1/instructor_3vmdk # Disk DescriptorFile version=1 CID=3cfc3778 parentCID=ffffffff createType="vmfsPassthroughRawDeviceMap" # Extent description RW 8388608 VMFSRDM "instructor_3-rdmp.vmdk" # The Disk Data Base #DDB ddb.toolsVersion = "7172" ddb.adapterType = "lsilogic" ddb.geometry.sectors = "63" ddb.geometry.heads = "255" ddb.geometry.cylinders = "522" ddb.virtualHWVersion = "4"

Exporting & Importing Virtual Disks

Note: Those of you with some experience will know importing (into a VMFS volume) and exporting (out of a VMFS volume) virtual disks has two functions. First it serves as a safe method of converting disk into a format that can be taken to/from foreign file systems such as NTFS and ESX3. Historically, these file systems have tools that could corrupt files bigger than 2GB in size These two formats have in the past been referred to as Monolithic and COW (copy-on-write). COW is no longer a favoured term as it is technically inaccurate instead we refer to it as the Sparse format That is more or less still the same however there is a switch (-d) that allow you control the format of the disk. The d switch is not always required it depends on if your exporting or importing.

Page 46 RTFM Education

This is what the MAN pages say about the d switch: zeroedthick This is the default option when creating new virtual disks. A zeroed thick disk has all space allocated at creation time, and this space is wiped clean of any previous contents on the physical media. eagerzeroedthick An eager zeroed thick disk has all space allocated and zeroed out at creation time. Such disks may take longer time during creation compared to other disk formats. thick A thick disk has all space allocated at creation time. This space may contain stale data as it exists on the physical media. thin Space required for thin-provisioned virtual disk is allocated and zeroed on demand as opposed to upon creation. Another purpose is for backup the sparse format holds the true data minus any free space so generally more efficient for backup purposes.

Related to this there is also the switch -j Convert a thin virtual disk to thick with the additional guarantee that any data on thin disk is preserved and any blocks that were not allocated get allocated and zeroed out.

Another purpose is from transporting virtual disks safely to/from other vPlatforms such as Workstation on Linux or Windows or VMware Server (nee GSX) on Linux or Windows Perhaps you dont have VC you still have a stand-alone ESX servers and you still use templates and import/export method of ESX 2.x days For those of you from a ESX 2.x background you should know the old e switch is depreciated and the way do a e is with i and some additional switches in vmkfstools

Exporting Zeroedthick Virtual Disks (nee Monolithic) into the Sparse (nee COW) Format Note: In my case I am using /vmimages partition. But you could use any EXT3 location or even export directly to a mount NFS or SMB share 1. Type the command: vmkfstools -i /vmfs/volumes/local-esx1/instructor/instructor.vmdk -d 2gbsparse /vmimages/instructor.vmdk Note: You should get response like this Destination disk format: sparse with 2GB maximum extent size Cloning disk '/vmfs/volumes/local-esx1/instructor/instructor.vmdk'... Clone: 3% done. Page 47 RTFM Education

Note: The disk is now in a portable format you could tar it up and copy somewhere else for backup purposes connect with WinSCP on your workstation and bring the files down to your PC. If you have VMware Workstation on your PC you could configure a virtual machine to use this disk Importing Virtual Disk from the Sparse Format Note: For those you that a familiar with ESX 2.x this process hasnt changed much at all. What is different is because VMFS now stores directories, and the VMs configuration files are held in a directory is where do you restore the disk too The easiest way to do this I think is, to first create a VM without a virtual disk this will set-up up the directory the configuration files. Then trigger the vmkfstools import to the vmfs volume and directory and then edit the VMX file to tell the VM where its new disk is (as I showed just previously with c) 1. Type the command: vmkfstools -i /vmimages/workstationdisk.vmdk /vmfs/volumes/localesx1/workstation1/workstation1.vmdk Destination disk format: VMFS thick Cloning disk '/vmimages/media.vmdk'... Clone: 10% done.

Copying Files from one ESX host to another

Note: SCP (Secure Copy) the utility to use. The security is not one of reliability of the copy just stopping on the interception of the data through the network. This example copies a vmdk its sparse format. Remember to use ESX as SSH client this must be enabled in the ESX firewall first 1. Type: scp /vmimages/instructor*.vmdk [email protected]:/vmimages Note: Dont forget the colon: after the name of your server Note: It will warn you that the first time you copy the destination host The authenticity of host esx1.rtfm-ed.co.uk (192.168.2.102)' can't be established. RSA key fingerprint is 92:4b:ba:b5:ca:31:6f:e7:8c:2d:00:6e:cf:c6:b6:ea. This is because the certificate of the corresponding machine is un-trusted and cannot be verified. To get rid of these first time messages you need to assign your own certificates or acquire them from a 3rd party. 2. Choose Yes to continue Note: The system will then warn you that: Page 48 RTFM Education

 Warning: Permanently added esx1.rtfm-ed.co.uk, 192.168.2.102' (RSA) to the list of known hosts. 3. At the prompt which reads [email protected] password: type the other servers root password, in my case ******** Note: In this screen grab I copied the tar file for upgrade of esx 2.x to 3.x The system will give you a status bar like this

Renaming Virtual Disks

Note; Renaming virtual disks used to be dead easy. You would just the mv command, and edit your VMX file accordingly. Now virtual disks have their companion vmdk metadata file and the flat.vmdk. We need a special command to rename both files and keep the data in synch. I guess you could manually edit the metadata file but why bother when vmkfstools does it all for you. 1. Before my rename the metadata in the instructor.vmdk looked like this: # Disk DescriptorFile version=1 CID=a896726b parentCID=ffffffff createType="vmfs" xdgkdkfdlkfjldkjflkjsdf # Extent description RW 4194304 VMFS "instructor-flat.vmdk" xdgkdkfdlkfjldkjflkjsdf # The Disk Data Base #DDB ddb.adapterType = "lsilogic" ddb.geometry.sectors = "63" ddb.geometry.heads = "255" ddb.geometry.cylinders = "261" ddb.thinProvisioned = "1" ddb.virtualHWVersion = "4" ddb.toolsVersion = "7172" 2. Type the command: vmkfstools -E /vmfs/volumes/local-esx1/instructor/instructor.vmdk instructorOS.vmdk 3. After the rename the metadata looks like this # Disk DescriptorFile version=1 CID=a896726b parentCID=ffffffff createType="vmfs" xdgkdkfdlkfjldkjflkjsdf Page 49 RTFM Education

# Extent description RW 4194304 VMFS "instructorOS-flat.vmdk" xdgkdkfdlkfjldkjflkjsdf # The Disk Data Base #DDB ddb.adapterType = "lsilogic" ddb.geometry.sectors = "63" ddb.geometry.heads = "255" ddb.geometry.cylinders = "261" ddb.thinProvisioned = "1" ddb.virtualHWVersion = "4" ddb.toolsVersion = "7172" Note: Remember, you still have to manually edit the VMX file to reflect your changes

Resizing Virtual Disks

Note: According to VMware employees on the forum, the x switch is only supported for making virtual disks larger not smaller. This has been the case in ESX 2.x. This surprised me because in the past I have successfully used x to make a virtual disk smaller. This however, was way back in the midst of time, circa ESX 2.0 (around 2004). So perhaps they withdrew it as function as is very easy to corrupt files when you do a shrink If you want to shrink a virtual disk perhaps you use ghost to clone a big virtual disks to a small one Note: There a couple of ways approaching this issue while it is relatively easier to increase and decrease the VMDK file, Windows does not give you an easy way to reduce or increase the size of a partition. Therefore 3rd Party tools maybe required to complete the procedure especially it the partition is the boot partition There are five main methods: o Partition Tools you can use Power Quest Partition Magic on Client based OS, to resize the partition or Power Quests Volume Manager on Server based OS o These are commercial available tools and they cost money alternatively you can use the Knoppix Boot CD (version 3 and higher) and tool on the CD called qparted which does the same thing o If you have W2K3 (or W2K with the Resource Kit) you can go through a procedure using MS Diskpart utility. Its not a very friendly process and is quite convoluted. Its main advantage is that is cheap as you do not need 3rd party tools achieve it. It disadvantage is that as far as I am aware its is not available in W2K or NT4. For how to do this more information about this visit vmwareprofessional.com o Ghost Method Create a new disk of the appropriate size and restore the data using a disk to disk restore method o DR method Backup OS System State to another location create a new OS disk file, and restore the data Its worth knowing the alternatives in case you cannot get your preferred method to work

Page 50 RTFM Education

Expect a number of reboots for the one asked for by the Partitioning tool, and the one called for by Windows. This can be somewhat annoying!

Caution: Copy/Backup/Gzip/Export the VMDK in case the next procedure fails Note: I noticed that Knoppix is susceptible to disconnections from the Remote Console especially when you are shutting down the VM after booting from the ISO 1. Shutdown the VM 2. Type the following command: vmkfstools X 6144m force /vmfs/volumes/local/instructor.vmdk Note: If you follow this command with ls l h you will see it is bigger. Again I could have specified 6g instead of 6144m. Its flat file that changes size the metadata file stays the same but has new geometry information 3. Start-up the VM, and check that it boots properly Note: You may get error loading operating system messages if the vmdk file has become corrupted 4. Connect to the ISO which contains the Knoppix Boot CD - Using Remote console attach to this ISO file making sure you enable X Connected at Power On 5. Power on the VM 6. Change the boot order on the VM to give the CD-ROM the highest priority allow the system to boot to Knoppix CD or use the [ESC] key to select the CD-ROM as a the boot device 7. Click at the Knoppix KDE button

8. Choose System Tools

9. Choose Qparted Note: Qparted will display the partition/disk layout of the virtual machine.

Page 51 RTFM Education

In this case the IDE CD-ROM and single SCSI boot disk (sda) 10. Select /inionfs/dev/sda Note: This then shows the partition table on that virtual disk

In this case a single C drive of 4GB with free space on the hard-drive of 2GB added by using vmkfstools X 6096mg 11. Select Partition No 01 12. Right-click and choose &Resize 13. In the dialog box drag to increase or decrease the size of Partition 01

Using the

icon and Click OK

14. Click File, and choose &Commit Note: Page 52 RTFM Education

Confirm the warnings by click the Yes button and watch the dialog boxes go by and Confirm the Success dialog by clicking OK 15. Shutdown the Virtual Machine 16. Disconnect the CD-Rom Note: This will allow the VM to boot to the VMDK file rather than to the CD-Rom. ` 17. Reboot the Virtual Machine Note: In Windows 2003 you will see check disk events this appears to be normal. Additionally, you will find Windows 2003 will detect the hard-drive as if it was a new device and request a reboot.

Deleting Virtual Disks

Note: Im not really sure why there is a special command for deleting virtual disks when we could easily use rm namofvirtualdisk*.vmdk this would delete the metadata and also the flat.vmdk file but anyway vmkfstools can do it too 1. Type the command vmkfstools -U /vmfs/volumes/local/instructor.vmdk Note: It deletes both the metadata and the flat file and doesnt currently ask Are you sure. Hee-hee

Using VMware-cmd
Note: It works by manipulating the VMX and sending instructions to the VM via the VMX file There are lots you can do with this command. Beyond what is feasible in this document. For further information look at the Scripting Guide for ESX It also returns 1 for positive results, and 0 for failures so can use it for BASH shell scripts and if statements To List Registered VMS 1. Type: vmware-cmd l Note: List paths and names of VMX files. Useful for when you have type long paths to the VMX file. You can highlight an entry in the list and copy it to your command-line. Unfortunately, vmware-cmd no long supports volume names instead you have to use the UUID path to the VMX file. I will give one example of vmware-cmd with the UUID, after that I will replace the string with <UUID> because its so blummin long!

Page 53 RTFM Education

To Register and Un-register a VM 1. Type: vmware-cmd -s unregister /vmfs/volumes/44a38c72-156b2590-be1500065bec0eb7/instructor/instructor.vmx Note: Also includes register as a command. S stands for set. This un-registers a VM from an ESX host (stand-alone) Unfortunately, it does NOT un-register it from VirtualCenter. Instead you are left with an orphaned VM that needs right-clicking and Remove from the Inventory

I think this happens because what vmware-cmd s is manipulating is what used to be called the vm-list. Its not manipulating the VC database. Note: Why is the command still useful? We say you had a server failure which you had rebuild. You then need to register the VMX file with an ESX server to then be able to manage it. If you had 40 VMX you wouldnt want to do that by hand using the GUI. Power Options on a VM Note If you use trysoft, the Guest OS will be shutdown gracefully. The trysoft is a mode option it tries to run the normal scripts but uses a hard shutdown/startup if the VM is not behaving properly. There are two other modes soft which runs scripts but never does a hard start or stop or hard which powers off/on a VM as if you had hit the power switch. Sometimes people use trysoft first, and then if the VM refuses to power of, they follow it with hard power off

1. To

a virtual machine type:

vmware-cmd /vmfs/volumes/<UUID>/instructor1/instructor1.vmx start trysoft

2. To

a virtual machine type:

vmware-cmd /vmfs/volumes/<UUID>/instructor1/instructor1.vmx stop trysoft

Page 54 RTFM Education

3. To

a virtual machine type:

vmware-cmd /vmfs/volumes/<UUID>/instructor1/instructor1.vmx suspend Note: There isnt much in the way of status/progress here. The old-style Remote Console from ESX 2.x days used to give you dialog box with a status bar

4. To

a virtual machine type:

vmware-cmd /vmfs/volumes/<UUID>/instructor1/instructor1.vmx start trysoft Note: There is no resume switch on vmware-cmd just a power-up which retrieves the suspend file, and resumes the machine

5. To

a virtual machine type

vmware-cmd /vmfs/volumes/<UUID>/instructor1/instructor1.vmx reset trysoft Note: This a soft reboot of the virtual machine 6. To find out the Power Status of a VM use: vmware-cmd /vmfs/volumes/<UUID>/instructor1/instructor1.vmx getstate Note: It will return On; Off; Suspend and Stuck if the VM is waiting for interaction Finding the Heartbeat 1. Type: vmware-cmd /vmfs/volumes/<UUID>/instructor1/instructor1.vmx getheartbeat Note: You should see a number. Repeat the command. If the number increments the machine is alive. It stays the same it is dead. It doesnt have a heart beat! Finding the Status Of Devices Note: The Vmware-cmd tools is able to find the configuration of the VMX file, and also change entries in the file. To find out what the status is of the CD-ROM you would use:

Page 55 RTFM Education

1. vmware-cmd /vmfs/volumes/<UUID>/instructor1/instructor1.vmx getconfig ide0:0.deviceType Note: The CD-ROM is emulated (but not Virtualized) IDE 0:0 channel. This is how the CD-ROM is addressed in a Windows Environment. A safe way of learning the VMX variables is just print the VMX file to the console with: cat /vmfs/volumes/<UUID>/instructor1/instructor1.vmx | more Configuring & Connecting a Device This used to work in the past and none of these commands give an error message but I dont get the desired result They must have changed this from ESX 2.x days and dont what the next step is Note: As long as you know the name of the variable in the VMX file you can change anything you like from the command line. For example there is a variable called ide0:0.fileName = which controls the path for the CD-ROM be it physical or an ISO image. There is a variable called ide0:0.deviceType = which controls if it is the physical or ISO image (atapi-cdrom for the physical and cdrom-image if it is an ISO file). In this example we will attach an ISO to the CD-ROM and Connect the CD-ROM to the VM. You may wish to have VC Console open whilst your doing this to see visually the effect of your changes. I assume that initially the CD-ROM is connected to the physical CD 1. First, Disconnect the current device with: vmware-cmd /home/lavericm/vmware/instructor1/instructor1.vmx disconnectdevice ide0:0 2. Change to the system to use a ISO file instead of a physical CD with: vmware-cmd /home/lavericm/vmware/instructor1/instructor1.vmx setconfig ide0:0.deviceType cdrom-image Then set the image file to use with: vmware-cmd /home/lavericm/vmware/instructor1/instructor1.vmx setconfig ide0:0.fileName /vmfs/volumes/isos/w2k3.iso 3. To Refresh the OS - Reconnect the device with: vmware-cmd /home/lavericm/vmware/instructor1/instructor1.vmx connectdevice ide0:0 Note: Now that the CD is set as an ISO file. You merely need to stages 2 and 3 to switch from one ISO file another. The whole process could be put in SH script like so: vmware-cmd /home/lavericm/vmware/instructor1/instructor1.vmx disconnectdevice ide0:0 vmware-cmd /home/lavericm/vmware/instructor1/instructor1.vmx setconfig ide0:0.deviceType cdrom-image Page 56 RTFM Education

vmware-cmd /home/lavericm/vmware/instructor1/instructor1.vmx setconfig ide0:0.fileName /vmimages/w2k3ent.iso vmware-cmd /home/lavericm/vmware/instructor1/instructor1.vmx connectdevice ide0:0 Notes: To Undo and go back to the physical CD-ROM vmware-cmd /home/lavericm/vmware/instructor1/instructor1.vmx disconnectdevice ide0:0 vmware-cmd /home/lavericm/vmware/instructor1/instructor1.vmx setconfig ide0:0.deviceType atapi-cdrom vmware-cmd /home/lavericm/vmware/instructor1/instructor1.vmx setconfig ide0:0.fileName /dev/cdrom vmware-cmd /home/lavericm/vmware/instructor1/instructor1.vmx connectdevice ide0:0

Changing Parameters within a VM with /proc

Note: It seems like the info in proc is quite limited which surprised me so amount of stuff we can change is now limited as well

Page 57 RTFM Education

Module 7: Performance Monitoring

Note: There are number of commands we can use to see how the Service Console is performing and how virtual machine are performing I will start with virtual machines as they are probably more likely to be source of the concern and problem Be very careful if you try to use the guest OS tools to do performance management/analysis these were designed for OS running on physical hardware our VMs are running on virtual hardware Good example of this is when Windows shows 100% CPU but outside of the VM you have capped the maximum CPU to 50%. The 100% shows Windows is using all of its 50% allocation and does NOT reflect true CPU usage

Using ESXTOP to Monitor Virtual Machines

Note: Theres too much stuff here. Sorry! But one thing to look at if CPU is your constraining resource is the %ready values. The higher the %ready value the more frequently the VM was ready to run but the CPU was too busy to service the request In this example of esxtop we going to do something you do very often on courses to demonstrate share values (the proportional system for relative resource management) To create contention I will peg the two VMs to the SAME CPU you should never do this in the real world in the world we strive (and often fail) to remove or reduce the effects of contention Whats contention when two or more VMs compete for the same resource Two VMs with cpubusy script with the SAME share value 1. Power up JUST two VMs 2. Create this VBS script in each them Dim Dim Dim Dim Dim goal before x y i

goal = 2181818 Do While True before = Timer For i = 0 to goal x = 0.000001 y = sin(x) y = y + 0.00001 Next y = y + 0.01 WScript.Echo "I did three million sines in " & Int(Timer - before + 0.5) & " seconds!" Loop

Page 58 RTFM Education

3. Run the script on just one VM with: cscript //nologo cpubusy.vbs Note: This should produce lots of CPU cycles and this message in a cmd prompt:

Run ESXTOP at the Service Console Note: 1. esxtop is dynamic tool so every couple of second it updates but heres a snapshot in time (you might want to zoom in here!)

Interpretation: Server1 is busy with %used at 99.17. The ESX server is doing it best to load the CPU request across my two physical CPUs (PCPU(%) is at 62.40 and 40.50. The %RDY value is quite low for VM called instructor and server1. So although server1 is busy, they are still performing well Run CPUbusy.vbs on the Second VM and Check out ESXtop

Interpretation: With two VMs running the script my PCPUs are beginning to max out at 99.67 and 99.61. Now the VM called instructor is as busy as server1. The %RDY value has increased but not massively so. With just two VMs with a

Page 59 RTFM Education

vCPU each the vmkernel is analyzing the cpu performance to make sure the two VMs dont fight against each other Deliberate Mis-configuration Creating contention Note: By default a single vCPU VM runs on any CPU that has free CPU cycles We can use processor affinity to cause contention by pegging each VM to the same CPU 1. 2. 3. 4. 5. 6. Do this on BOTH VMs Properties of the Virtual Machine Select the Resources Tab Choose Advanced CPU Under Scheduling Affinity, choose Run on processor(s) Select just one CPU for the VM like so:

7. Click OK Run ESXTOP at the Service Console

Page 60 RTFM Education

Interpretation: Here my CPU0 is no longer busy, but CPU1 is maxd out at 100.00% Both VMs are getting an equal share of the CPU at 50.08% and 49.89%. The %RDY value has grown massively to 50%. This is contention and its bad VMware say anything over 5-10% needs dealing with. So what our options? VMotion one of these VMs to less busy ESX host Increase shares to instructor because it is more important Decrease shares to server1 because it is less important Remove the contention the CPU overlap we have just created Increasing Shares on VM Note: Shares are method of favouring on VM over others when contention takes place. No contention. The share values do nothing and VMs scream along as fast as they can. But when contention occurs resources are divvied up according to the share value In my case I am going to give the instructor a 3000 shares, and the other VM left at 1000. Thats 4000 shares altogether so the instructor VM will receive of CPU time, and server1 will get 1. 2. 3. 4. 5. 6. Right-click one of your VMs Edit Settings Click the Resources Tab Select CPU from the list of settings Next to Shares, choose Custom Increase to 4000

Run ESXtop at the Service Console

Page 61 RTFM Education

Interpretation: My physical CPU is just as busy. The Instructor vm is getting 80% of CPU () and server1 is getting just 20% (). The %RDY value is reflecting an inverse proportion. So as Instructor is getting more CPU time is %RDY value is lower (22%) than server2, which is getting less CPU time (84%) Conclusions High CPU values in esxtop in themselves are not a problem in fact the higher the better. Whats the point of virtualisation if the server only runs at 5% of its capacity. Or put another waywhats the point of building an oil pipeline that only ships 5% of oil. The rest is spare and wasted capacity that could be used to increase output What matters is bottlenecks - %RDY values are good indicator of bottleneck in our pipeline Dont trust what the guest operating system says. Both of my VMs report 100% CPU usage when actually they are using 80% () and 20% () of the processor See screen dumps on the next page

Page 62 RTFM Education

Instructor with 4000 shares

Server1 with 1000 shares mmm Im looking at the time on the screen dumps here. I should call it day

Page 63 RTFM Education

Tools to Monitor the Service Console

Note: Whereas a poor performing VMs will affect your users, a poor performing Service Console will not. It will however, annoy the pants of you Symptoms include: Random disconnects from ESX and VirtualCenter Failure of ESX services to start or stay started Incomplete logging General sluggishness of management tools General annoyance and frustration sometimes so bad you say out loud God I wish I bought Microsoft Virtual Server, then I would have good reason for knowing why things dont work CPU top is a Linux tool which show you processes and process IDs You can use the pidid value or the command name to terminate a process if you think it has gone wrong Killall -9 pidid will kill a running process and it process tree ps ef | grep vmware would give you a full process list, filtering on the string vmware

Memory free m will list how much RAM has been allocated to the Service Console. Interesting on my upgrade ESX 2.x box the amount memory was left at 256MB. On a cleanly install ESX 3.x box its around the 500MB mark vmstat 5 10 will give virtual memory/swap statistics every 5secs in my case with 10 samples. Watch out for s/i and s/o any regular swap activity indicates that Service Console lacks RAM just like in Windows

Disk The likelihood of filling / is now much less but it could still happen Due to excessive logging in /var/log or agent logging in /opt. If these are directories rather partitions you might have problem Also if some miss uses vmkfstools i they could be creating very big files in the wrong location You can use df h or vdf h to see how full your file systems are

Page 64 RTFM Education

Module 8: Upgrading/Patching ESX

Pre-Upgrade Checks
Note: If you are doing an in place upgrade of ESX 2.x check out this script. Of course, you could just re-install the COS, and keep your VMFS2 partition, and just upgrade the file system 1. Insert your ESX 3.x CD into the ESX 2.x server 2. Type: perl /mnt/cdrom/scripts/preupgrade.pl This should give you a print out report of the status of the ESX server

Custom Upgrading with a Tar-Ball

Note: This is a straight cut & paste from my upgrade guide for ESX 3.x and VirtualCenter 2.x You can use a tar ball to either upgrade from ESX 2.x to 3.x OR when VMware release a maintenance release of ESX 3.x it might be a method of patching ESX 3.x Why do your upgrades this way? Well, it allows for unattended upgrades without having to be physically present at the server or having to run around from one server to another with CD! This is a two-stage process involving two scripts and reboot between script1 and 2 I copied the Tar-Ball to area of free-space this happened to be my /vmimages location (I was concerned my /tmp partition might be too small having set the Vmware recommended size of 512MB in ESX 2.x) Upgrading ESX in this way didnt seem to help with my virtual networking I still had some machine which thought they were on the production switch and others thought they were connected to a Legacy Bond0 even though I never had a bond. Run First Script & Set ESX Boot Option 1. Transfer the Tar file to the ESX 2.x Host and using WinSCP 2. Logon to the Service Console with PuTTY, as ROOT 3. Untar the tar.gz file with: tar -xvzf esx-3.?.?-?????-upgrade.tar Note: This will take some time! 4. Change the boot option and Reboot ESX lilo R linux-up reboot Note: When you reconnect to the Service Console, confirm you have not loaded the Vmkernel. An easy check is to ls /vmfs. No Vmkernel, No VMFS

Page 65 RTFM Education

5. cd in to the esx-3.?.?-?????-upgrade directory 6. Run the Perl script called ./upgrade.pl Note: The system will verify files, and then ask you to read & accept an EULA! 7. Choose Q, to quit reading the EULA, and 8. Type Yes [Enter] to accept the EULA Note: You will receive status messages like so: Upgrading packages (this may take a while) ... .... done Removing obsolete packages .... done Upgrading system configuration ... Running vmware-config.pl ... done then lastly *** You must now reboot the system into Service Console mode and run upgrade2.pl *** Reboot the server now?[y/n] 9. At the Reboot the Server now [y/n] choose N and [Enter] 10. Now edit the boot.conf file to make sure the server boots to the Service Console with: nano w /boot/grub/grub.conf change default=0 to default=1 11. Now do the reboot with the command: reboot Note: You might find you loose connectivity to the Service Console. I did on my first attempt. The second time I didnt get a script error and my Service Console networking was intact. The interesting thing is the upgrade was done on an identical ESX 2.x host. I know it was identical as it was the same physical machine cloned with ghost which contain an image of ESX. Perhaps these network errors were related to the boot errors? Run Second Script & Revert to ESX Boot Option 1. Logon to the Service Console with PuTTY, as ROOT Note: You will know you have boot to do the Service Console because of this message.

Page 66 RTFM Education

ALERT [Wed May 17 21:19:45 GMT 2006]: This VMware ESX Server machine is currently booted into troubleshooting mode. 2. cd in to the esx-3.?.?-?????-upgrade directory 3. Run the Perl script called ./upgrade2.pl Note: You will receive this status messages like so: This script will complete upgrading your ESX server to version 3.x.x Verifying Files don Upgrading Packages (this may take a while). INIT: Version X.x.x reloading done *** You must now reboot the system into the ESX Server mode to finish the upgrade *** 4. At the Reboot the Server now [y/n] choose N and [Enter] 5. Now edit the boot.conf file to make sure the server boots to the Service Console with: nano w /boot/grub/grub.conf change default=1 to default=0 6. Now do the reboot with the command: reboot Note: This will boot the ESX 3.x Vmkernel for the first time. The remote upgrade is complete. Note: If you wish to automate the change to grub.conf, this script is from IBM is called bootcontrol.pl you can download the zip file directly here. The very first time you run this command it returns an error which states "Use of uninitialized value in numeric eq (==) at bootcontrol.pl line 99". It does actually shift the boot option - and the error only occurs once. I've emailed the guys who wrote it - but I have a feeling its cause by our partially completed upgrade process. This command re-compiles the grub loader so that the Service Console Only (troubleshooting only) menu is chosen. It does that by changing the value in grub.conf called Default 0 to being Default 1

Upgrading to VMFS-3 from the Service Console

Acknowledgments: Page 67 RTFM Education

I would like to thank Mostafa Khalil, VCP (SE) for his Troubleshooting presentation at VMware TSX in Paris 2006. Gotchas You need free space to upgrade to VMFSv3 Watch out for block sizes. VMFSv2 support a max block size of 16MB. Whereas VMFSv3 supports 8MB. You cannot do an upgrade of VMFS2 to VMFS3 if this is the case. You in backup and restore territory. Hopefully, this will effect a very few number of people Note: From man pages of vmkfstools. VMFS-2 to VMFS-3 file system upgrade is a two step process. Before file system upgrade can begin the vmfs2 and vmfs3 driver must be unloaded and the auxiliary file system driver, fsaux, should be loaded. The first step of upgrade uses the -T option. Once the first step completes, the auxiliary file system driver, fsaux, is unloaded and, vmfs2 and vmfs3 drivers are reloaded. The second step of file system upgrade makes use of the u option. -T, --tovmfs3 converts a VMFS-2 file system on the specified partition to VMFS-3 format, preserving all files on the file system. The conversion is in-place and the auxiliary file system driver (fsaux) module must be loaded. The ESX Server file system locking mechanisms will try to ensure that no local process or remote ESX Server is currently accessing the VMFS file system to be converted. The conversion is a one-way operation and once the VMFS-2 file system is converted to VMFS-3 file system, it cannot be rolled back to VMFS-2. -u, --upgradefinish /vmfs/volumes/<label/UUID>/ once the first step of file system upgrade has completed (using -T), the vmfs2 and vmfs3 modules are reloaded and the -u option is used to complete the upgrade 1. Logon to the Service Console as ROOT 2. Unload the vmfs2 driver with: vmkload_mod -u vmfs2 3. Unload the vmfs3 driver with: vmkload_mod -u vmfs3 4. Load the FS Auxiliary Driver with the upgrade function vmkload_mod fsaux fsauxFunction=upgrade 5. Run the first stage of the upgrade with vmkfstools -T /vmfs/volumes/local -x zeroedthick Note: -x zeroedthick (default) . Retains the properties of VMFS-2 thick files. With Page 68 RTFM Education

the zeroedthick file format, disk space is allocated to the files for future use and the unused data blocks are not zeroed out. -x eagerzeroedthick . Zeroes out unused data blocks in thick files during conversion. If you use this sub-option, the upgrade process might take much longer than with the other options. -x thin . Converts the VMFS-2 thick files into thin-provisioned VMFS-3 files. As opposed to thick file format, the thin-provisioned format doesn't allow files to have extra space allocated for their future use, but instead provides the space on demand. During this conversion, unused blocks of the thick files are discarded. Note: This will give you the following message I got /vmfs/volumes/44a38c72-156b2590-be15-00065bec0eb7 VMware ESX Server Question: Please make sure that the VMFS-2 volume /vmfs/volumes/44a38c72156b2590-be15-00065bec0eb7 is not in use by any local process or any remote ESX server. We do recommend the following: 1. Back up data on your volume as a safety measure. 2. Take precautions to make sure multiple servers aren't accessing this volume. 3. Please make sure you have approximately 1200MB of free space on your volume. Note that the number is an upper bound, as the actual requirements depend on physical layout of files. Continue converting VMFS-2 to VMFS-3? 0) Yes 1) No 6. Type 0, and press [Enter] Note: Currently, you get no status information while this is proceeding. It should complete with this message: Filesystem upgrade step one completed. Step two must be completed after the vmfs2 and vmfs3 modules are reloaded. When ready, run 'vmkfstools -u /vmfs/volumes/local' to complete the upgrade. 7. Once this part has completed confirm that your format is VMFS-3 vmkfstools -P /vmfs/volumes/local Note: This should report something like this: VMFS-3.21 file system spanning 1 partitions. File system label (if any): local Mode: public Capacity 36238786560 (34560 file blocks * 1048576), 5511315456 (5256 Page 69 RTFM Education

blocks) avail UUID: 44a38c72-156b2590-be15-00065bec0eb7 Partitions spanned: vmhba0:1:0:1 8. Confirm you files havent disappeared in the process with: ls l /vmfs/volumes/local 9. Next, unload the Auxiliary Files System Driver and reload you VMFS2 and VMFS3 driver with: vmkload_mod -u fsaux vmkload_mod vmfs2 vmkload_mod vmfs3 Note: You should get Module load of vmfs2succeeded and Module load of vmfs3succeeded 10. Restart the hostd service for these changes to be reflected in the VI client with: service mgmt-vmware restart Note: The guide finishes here if you were doing an upgrade I would progress to Module 4 of my upgrade guide

Page 70 RTFM Education