QUESTIONS ON EXPERTS EXCHANGE

QUESTIONS--- Modules in ECC?

Answer (1) ----

The 3 topics ECC server + Store collect + Collect Repository does make sense with EMC
ControlCenter.

May be this "EMC ControlCenter infrastructure" page can help.

EMC ControlCenter PLANNING AND INSTALLATION GUIDE may be used for more detailed
information’s.

QUESTIONS--- SAN storage performance tuning in DMX 3 and DMX4

Answer (1) ----

some general tips can be found here

http://searchstorage.techtarget.com/tip/0,289483,sid5_gci1195077,00.html
and you should check this one
http://sqlblogcasts.com/blogs/christian/archive/2009/01/12/tuning-your-san-too-much-hba-
queue-depth.aspx

Answer (2) ----

Remember that the Storage is only one part of the overall SAN, you need to check, that your
SAN infrastructure is performing correctly and not restricting the performance of the whole
solution.

What storage management tools do you have?

If you are analyzing your logs from server/switches/Storage are all their TOD clocks
synchronized to the same source, so that their event logs make sense?

QUESTIONS--- What are CRC errors .on brocade switch..? What is command to
check CRC errors WE USE CLI? Do CRC errors affect application I/O which run
through ports? What are best practices to resolve CRC errors?

Answer (1) ----

First all, i assume you can see CRC errors logged against some or all of the ports on your
switch?

Are they static or are they increasing?

When was the last time you checked them?
Have you been plugging/unplugging the fibres since you last checked them?

If they are static or only increasing slowly, i suggest you record them on a weekly/daily/hourly
basis so that you get an idea of the rate these errors are happening and are they related to
some kind of activity in your datacentre?

Answer (2) ----

I see CRC errors logged against two ports which is connected to hosts and against one of
the ISLs
I did unplug and plug in again and still it is increasing ....when compared with the last record
which i had taken 3 hrs back
Will the CRC effect the application i/o
What is the measures by which i can get rid if these errors
Yes they will affect i/o

Have the errors increased? are you collected the port stats every hour?

Its not unusual to get a few CRC errors if you unplug/plug a cable or sfp, but if you are getting
them all the time then there is a problem with port/sfp/cable/sfp/port.

Is this a server-2-switch port or a switch-2-switch port?

1) On the switch end move the cable and its sfp into a port on the switch, does the increasing
errors follow the cable? if not its a prob with the port!
2) move the cable into a spare sfp, do the errors (increasing errors) follow the cable, if not
its the sfp

If the errors follow all of this then it could be the other end or the cable. Change the cable.
still no go, if its a server, try replacing the sfp (if it has one) on the HBA.

Tell us more about your config.

Answer (3) ----

FC witches or ethernet witches?
CRC errors are cabling faults. First try is re-plugging the cable.

Answer (4) ----

CRC errors slow down IO
If replugging did not help you need new cables

QUESTIONS--- I need to some commands or links by which i can performs SAN

storage performance tuning in DMX 3 and DM4

Answer (1)
my server team in complaining of some performance issues ...most probably it is the storage
side
I am providing storage from EMC DMX3 array via FC protocol .
Everything seems to be OK from switch side so in need to check what exactly is hampering the
performance.
Is there any way to check the Rpm of the disk .If yes what is the command and is it possible
to change the Rpm of the disk
So what kind of performance are they seeing and what are they expecting?

The only way to change the RPM of the disks is to buy new faster (RPM) disks!

Try using something like IOmeter to get a benchmark of the disk performance to take the
application issue out of the equation. Get the server team to specify what they need before
running this!

For advanced performance testing try IOzone.

www.iometer.org
www.iozone.org

they complain oracle database I/O is happening very slow .....do i need to check anything
from host side or is there anything that needs to be checked from the storage side
What are parameters in need to check either from storage or from host

You may well have an Oracle issue, not a SAN issue. I'd investigate that first.

To get useful information, you're going to have to be very specific about your Oracle
configuration, what application they're using to access the database, etc. The solution that's
best for your users may be to bring in an Oracle professional to evaluate what's going on...
because there's going to be a huge amount of back-and-forth if you try to answer the question
here, and not (IMHO) the kind of thing Experts Exchange was designed to help with.

Thats why i suggested finding out what performance they required and then running a
performance test to show if that performance was possible.
This will put the problem firmly back in their court, otherwise they will try and put the blame
on you!

QUESTIONS--- What are different modules in ECC or emc control center ?

What is WLA agent how does it work?

Answer (1)

EMC Control center

- http://www.lulu.com/items/volume_45/685000/685422/1/print/685422.pdf

WLA - http://www.bakercomputeranddata.com/doc/vendors/emc/emc-
docs/ECC5.2/help/ecc/agents/enw/enw_rf_wlaarchiver_overview.htm
QUESTIONS--- The mirror positions in symmetrix are M1, M2, M3 and M4. what
are these call
As far as i know M4 is the BCV and M3 is the srdf however i am not sure what are M1
and M2
Please advice

Answer (1)---

M1, M2, M3, M4 are mirror devices, or local BCV devices. You can have up to 4 mirrors of a
volume.

R1 and R2 describe SRDF devices (R=Remote). R1 is local, R2 is the remote target

QUESTIONS -- Allocate a LUN in cluster environment. (with 2 or 3 node cluster)

Answer (1)---

This is a useful resource: http://www.bakercomputeranddata.com/doc/vendors/emc/emc-

docs/ECC5.2/help/ecc/symm_mgr/cfgmn_pd_setting_port_flags.htm.

You need the following director bits set for Windows/VMware. Nothing extra required for
Windows or VMware clusters.
Enable Auto Negotiation (EAN)
Enable Point-to-point (PP)
Unique World Wide Name (UWN)
Common Serial Number (C)
SCSI 3 (SC3)
Common Serial Number (C) 4
Disable Queue Reset on Unit Attention (D)
Enable Volume Set Addressing (V)
SCSI-2 Compliance (iSCSI) (SPC-2)

There are some additonal settings for Sun etc clusters - see the link above.

Answer (2)---

The port flag settings are as above. You've got a bit of work to do with VCS
though: http://sfdoccentral.symantec.com/sf/5.1/windows/pdf/VCS_SRDF_Agent.pdf
QUESTIONS -- A lun- LUN012 is configured to a hostA(with WWN - XXXX) but
the server person informed the LUN is not visible form host side.
What is the trouble shoot part required . ( Port Zoning, Mapping and Masking
perfectly done but where is the problem how to trouble shoot it)

Answer (1)---

Many HBAs (qlogic, emulex, etc..) have a Java-based utility and SAN mapping that works at
the HBA/port level that specifies which LUNs the host computer can 'see'.

Check into that.

Answer (2)---

in windows you have to "scan for hardware changes"

Answer (3)---

You need to check:

FC switch zoning - that is; the host is zoned to FA ports.

You need to check what ports on the FA the LUN is mapped to
LUN masking - you need to check that the LUNs are masked to the WWNs of the hosts in the
VMCDB on the Symm.

Answer (4)---
FC switch zoning - Mapping masking and cables connection in perfect .
Still server team says they are not able to see the Lun what are the other troubleshooting i
need to do .
Also how to check that which LUNs the host computer can 'see'. form the host side if try it
from the host

Did you run the java-based HBA config program that you download ? If you had, it will show
you what is on the SAN, or at least what it can see.

Also, you can just telnet into the switch and query the switch port to see if it sees the HBA
port on your PC. If so, then at least you know problem MUST be the HBA config, or perhaps
your host O/S & driver issue. If you are using windows, then boot an ubuntu liveCD disk
(don't install the o/s), and then see if same HW running different OS can see everything, that
will eliminate other possibilities

we use Solaris hosts ...what exactly means HW

Recently we did the patching of host , Is there any possibility that the patch might has
causes some problem.
Is there possibility that the application which runs on the host might cause some problem

Well there is always that possibility. what is make and model of the HBA? Is it SUNs version
of an emulex or qlogic?

Then go to emulex site, download the latest drivers & get their java-based management
software. (Can't remember what it is called) .They also have firmware / bios updates. But
important, if this is the sun-branded emulex card, then you have to find the code on sun's
site, as the card has sun firmware on it.

In any event, the utility software has diags and will let you configure the card, update
firmware, AND see what is on the SAN, and configure the card.

You never responded earlier ... does your switch see the HBA? If not, then your problem is
elsewhere.

yes i can see the HBA wwn reflecting in swich

Good, at least we know it is not hardware. Probably just issue of making sure the drivers are
properly installed

Best way to do it is read the release notes that come with the HBA drivers. They walk you
through proper configuration of /kernel/drv/sd.conf , other settings, and how to set
persistence. It varies slightly depending on versions of solaris, and device driver, so you
have to do it the way release notes say.

QUESTIONS -- How do we configure Raid5 ? What is the main difference

between Raid 5 and Raid6 apart from having performance and fault tolerance and no
of parities? what are the two parities in Raid 6

Answer (1)---

http://technet.microsoft.com/en-us/library/cc938484.aspx
or
First install WIN2003 without the hardware controller installed. Once the OS is up and the
domain is setup, install the hardware controller. Go to disk management and config the
remaining HD's as dynamic. Use your manual for configing the hardware raid controller on the
3 remaining disks. I recommend getting another HD and creating a mirror of your OS, as it
can't reside on the raid 5 volume and you should have a backup of the OS to keep the server
up. I would run backups of the OS onto the raid 5 block just in case you get hit with a virus.
Hope that helps.
or
http://www.techotopia.com/index.php/Configuring_and_Managing_RAID_5_on_Windows_Serv
er_2008
or
As a rule, you’ll need one extra drive (for parity): if you have 146GB SCSI drives and need
about 500GB of drive space, you’ll need 5 drives (4×146= 584 + one parity drive = 5 drives)
Make sure the drives are identical (not an absolute requirement, but it sure makes life
easier…)
Make sure the the drive FIRMWARE IS AT THE SAME REV. LEVEL. this point IS important. You
stand a much higher probability of seeing good drives drop off the array (NOT a good
situation) with mismatched firmware.
have at least one drive available as a spare.
If the server or array enclosure is in a populated area (i.e.: not in a server room), turn off the
alarm (a suggestion only… those of you who have had the pleasure of experiencing the scream
of a RAID contoller with a failed drive will understand).
Decide whether you need more spindles or larger drives, and what your default block size will
be. This is an area which many admins don’t consider but can seriously impact the
performance of an array in any given role if done incorrectly. As a rule (general…), you want
more spindles / smaller capacity drives for a database environment as opposed to fewer drives
of a larger capacity for file & print or e-mail…
or
http://www.tek-tips.com/viewthread.cfm?qid=665716&page=20
and
http://www.hpl.hp.com/hpjournal/95jun/jun95a11.pdf
or
http://digitalcommons.unl.edu/cgi/viewcontent.cgi?article=1068&context=csetechreports
or
http://www.abe.pl/html/samples/b/1590590295.pdf
or
http://technicaldiscussionforum.blogspot.com/2007_07_01_archive.html
or
http://www.sqlservercentral.com/blogs/sqlmanofmystery/default.aspx?PageIndex=2
or
http://dtrace.org/blogs/ahl/page/2/
or
http://storagenerve.com/tag/sun/
or
http://www.axstor.com/PDFS/technology_whitepapers/Understanding%20the%20Financial
%20Benefits%20of%20iSCSI%20and%20SANs.pdf
and
http://www.zetta.net/_wp/?cat=5&paged=2

QUESTIONS -- How much is the raw capacity of symmetrix DMX3 and DMX4
arrays in TB

Answer (1)---

Note that there is no such thing as a 500GB FC drive (146GB, 300GB, 400GB, 450GB and
600GB only). There is also no such thing as a DMX that supports 5,200 drives. The biggest
VMAX array supports 2,400 drives. EMC say a maximum of 2.36PB
raw: http://australia.emc.com/collateral/hardware/comparison/emc-symmetrix-dmx.htm

Ummm. I don't remember what the DMX3 specs were (they're kind of old now), but it would
be similar to the DMX4 at about 1PB

QUESTIONS -- How much is the bandwidth (Mbps or kbps)required between

local and remote DMX3 or DMX4 symmetrix arrays in SRDF /Asynchronous
Replication ?

Answer (1)---

You have to determine the rate of change of the data. How much data do you have and what
is the rate of change? The size of incremental backups will give you an indication.
Posted via EE Mobile

This was asked to me in an interview

can you give any one scenario by taking any example date and rate of change .and how
incremental backups will give an indication.
we use symcli commands

OK.

You're weekly full backup is 100GB. You're daily incremental backup is 3GB, so file rate of
change is 3%. You can now estimate your block rate as it will be less than the file rate of
change at one third of the file rate of change, so block rate of change is 1%. !5 of 100GB is
1GB. Most of that change occurs over an 8 hour window during production hours, so:

1000MB / 8 hours = 125MB/hr

125MB / 3600 = .03472 MB/second, or 34KB/sec
34KB/sec x 8 = 277Kb/sec (note change from Bytes to bits). If you want to be conservative,
change the factor from 8 to 10 to allow for Ethernet and IP overheads.
So, if you have to replicate 1GB per working day, you need a serial link that can handle
277Kb/sec. I would typically increase this number to allow for the peaky nature of replication
workloads, and the fact that you can't really purchase a 277Kb serial link. You'd want a
1Mb/sec link at a minimum. This will allow you to maintain the RPO.

The formula scales to any data size.

QUESTIONS -- Can we make any port an ISL/E port in brocade witch ?If yes
then what are the commands required? How can we figure out in a switch that a
particular port is a ISL/E port when we do not have switches connected by ISLs
(standalone switch)? What is a command to see whether ISL license is there or not?

Answer (1)---
>Can we make any port an ISL/E port in brocade witch ?If yes then what are the commands
required
No commands are needed. The switch auto-detects the topography. It'll detect that another
switch has been connected on that port and will become an ISL once you tell it that it's OK to
merge the fabrics and form an ISL.

If you are using the Brocade web GUI, you can see what the ports are from the landing page
by clicking on them. You can also see the installed licenses by looking at the Properties of the
switch

QUESTIONS --
1) What are the commands to setup ISL between two Brocade switches
Suppose i hve two switches 7500 and 3900 and i need to connect what are the
coammands or the
settings i need to change on ISL's.

2) In a fabric which consists of three brocade switches,if one switches goes down and
when we bring this switch up again what ate the setting or commands we need make
or give so that the swith become fully functional with other switches in same fabric

Answer (1)---

You don't have to do anything to the ports themselves, assuming you've got full fabric license
installed they'll automatically negotiate to E ports.

Set the domain IDs so they are all different, e.g. 10,11 and 12. Put your ISLs in place, you
can have multiple ISLs without ISL trunking license, it still uses both but allocates them in a
round robin.

The switch with the lowest domain ID will become the principle switch and will hold the zoning
table, but the others have a copy of it as well, so if one fails you just have to set the domain
ID on the replacement to match and it'll pick up the config. You must make sure that the
replacement has got its old config wiped out. If it just gets powered off it'll power back up
again without any need to change anything since it';; have the domain ID still set and its zone
table will match the others.

If your switches have current configuration on them it will merge the zones when you connect
them together, you therefore have to make sure you don't currently have two zones with the
same name.

So as i understand we just need to connect the two Switches using ISL by simply plugging in
the cable and the switch will be connected to each other.So no changes required.

If in a fabric any switch goes down we just need to power it back and rest all the
configurations will be matched with other switches.So we do not need to remove any config
wiped out

Can we make any port an ISL/E port ?If yes then what are the commands required?
How can we figure out in a switch that a particular port is a ISL/E port ?
What is a command to see whether ISL license is there or not?

The only command needed to make ISLs available is to change the domain ID, it's easiest to
do that using switchexplorer web GUI. Any and all ports will become E ports as soon as you
connect them to another switch. liecenseshow command will tell what licenses are installed.

Referring to the above statement "" need to change the domain Id to make ISL available ""
When a switch gets power down and power up again do we have to again change the domain
Id so that the ISL are available.
In which scenario is above statement "" need to change the domain Id to make ISL available
"" applicable
We use the CLI interface to manage switches

The domain ID is retained after power outage. For CLI:

To modify the domain ID:
Type switchdisable to disable the switch.
Type configure , then type a new value.
Or press Enter to accept each default value.
At the Fabric parameters prompt, type Y and press Enter :
Fabric parameters (yes, y, no, n): [no] y
Enter a unique domain ID. For example:
Domain: (1..239) [1] 3
Complete the remaining prompts or press Ctrl+D to accept the remaining default settings.
Type switchenable to re-enable the switch.
Type fabricshow to confirm any changes made to the domain ID.

If you have multiple ISL Links between a pair of switches, you will need a Trunking licence to
make sure that some links are not heavily oversubscribed and others only lightly.
E.g. If you have two switches (without trunking) connected by two links a-a (@100MB/s) and
b-b (@100MB/s) and three servers trying to connect across these links if server-1 is
generating 70MB/s of throughput, Server-2 @30MB/s and Server-3 @50MB/s. The default
behaviour is to distribute these on a round-robin basis so Server-1 will go on link a-a and
Server-2 will go on link b-b, but Server-3 will go on link a-a. So link a-a will now be
oversubscribed by 130%, wheras link b-b is only 30%.
Trunking turns these two links into one big logical pipe of 200MB/sec and will therefore only
be 75% loaded. YMMV
Scale these numbers up to suit your config!
Some questions.
Are you still trying to keep 2 independent fabrics? Always a good idea to keep 2 independent
fabrics that don't communicate with one another. Don't merge all of your fabrics into one.
Is one of the switches that you want to ISL brand new with no active zoning config?
Is the cable distance between these switches less than 2km?
If one of those switches has no active zoning config.
1. set the switch IP
2. set the domain ID to a # that isn't already in use in the fabric
3. Add the ISL
4. The new switch will download the zoning config automatically and be good to go.
4. Add at least one more ISL (no steps required - just plug them in)
If both of your switches have active zoning configs - Then use Brocade Fabric Manager - They
have a "Fabric Merge" feature.
Trunking - Usually isn't needed. Typically the default (free) is fine. If you enable bottleneck
detection and see lots of confgested ISLs, add another ISL and then possibly look into
trunking.