Understanding oracle rac internals part 2 - slides

Understanding Oracle RAC Internals – Part 2
for the Oracle RAC SIG
Markus Michalewicz (Markus.Michalewicz@oracle.com)
Senior Principal Product Manager Oracle RAC and Oracle RAC One Node

Safe Harbor Statement
The following is intended to outline our general product direction. It is intended for
information purposes only, and may not be incorporated into any contract. It is not a
commitment to deliver any material, code, or functionality, and should not be relied upon
in making purchasing decisions. The development, release, and timing of any features
or functionality described for Oracle’s products remains at the sole discretion of Oracle.

Agenda
• Client Connectivity
• Node Membership
• The Interconnect

Client Connectivity
Direct or indirect connect
Production
Email
BATCH
• Connect Time Load Balancing (CTLB)
• Connect Time Connection Failover (CTCF)
• Runtime Connection Load Balancing (RTLB)
• Runtime Connection Failover (RTCF)
Connection
Pool
SCAN

Client Connectivity
Connect Time Connection Failover
jdbc:oracle:thin:@MySCAN:1521/Email
PMRAC =
(DESCRIPTION =
(FAILOVER=ON)
(ADDRESS = (PROTOCOL = TCP)(HOST = MySCAN)(PORT = 1521))
(CONNECT_DATA = (SERVER = DEDICATED) (SERVICE_NAME = Email)))
Production
Email
BATCH
Connection
Pool
MySCAN

Client Connectivity
Runtime Time Connection Failover
Production
Email
BATCH
Connection
Pool
MySCAN
PMRAC =
(DESCRIPTION =
(FAILOVER=ON)
(CONNECT_DATA = (SERVER = DEDICATED) (SERVICE_NAME = Email)
...))

Client Connectivity
Runtime Time Connection Failover
Production
Email
BATCH
Connection
Pool
MySCAN
PMRAC =
(DESCRIPTION =
(FAILOVER=ON)
(CONNECT_DATA = (SERVER = DEDICATED) (SERVICE_NAME = Email)
(FAILOVER_MODE= (TYPE=select)(METHOD=basic)(RETRIES=180)(DELAY=5))))
?

Client Connectivity
More information
Production
Email
BATCH
Connection
Pool
MySCAN
?
??
• If problems occur, see:
• Note 975457.1 – How to Troubleshoot Connectivity Issues with 11gR2 SCAN Name
• For more advanced configurations, see:
• Note 1306927.1 – Using the TNS_ADMIN variable and changing the default port
number of all Listeners in an 11.2 RAC for an 11.2, 11.1, and 10.2 Database

Client Connectivity
Two ways to protect the client
Production
Email
BATCH
Connection
Pool
MySCAN
1. Transparent Application Failover (TAF)
• Tries to make the client unaware of a failure
• Provides means of CTCF and RTCF
• Allows for pure ‘selects’ (reads) to continue
• Write transactions need to be re-issued
• The Application needs to be TAF aware
2. Fast Application Notification (FAN)
• FAN wants to inform clients ASAP
• Client can react to failure asap
• Expects clients to re-connect on failure (FCF)
• Sends messages about changes in the cluster
?

Client Connectivity and Service Definition
Define settings on the server
Production
Email
BATCH
MySCAN
[GRID]> srvctl config service
-d ORCL -s MyService
Service name: MyService
...
DTP transaction: false
AQ HA notifications: false
Failover type: NONE
Failover method: NONE
TAF failover retries: 0
TAF failover delay: 0
Connection Load Balancing Goal: LONG
Runtime Load Balancing Goal: NONE
TAF policy specification: BASIC
• HA (and LB) settings
can be defined per service
• Clients connecting to the service will
adhere to the settings considering the
client used.

Client Connectivity
Use a FAN aware connection pool
Production
Email
BATCH
Connection Pool
MySCAN
• If a connection pool is used
• The clients (users) get a physical
connection to the connection pool
• The connection pool creates a physical
connection to the database
• It is a direct client to the database
• Internally the pool maintains logical connections
1

Client Connectivity
Use a FAN aware connection pool
Production
Email
BATCH
Connection Pool
MySCAN
• The connection pool
• Invalidates connections to one instance
• Re-establishes new logical connections
• May create new physical connections
• Prevent new clients to be misrouted
• The application needs to handle the
transaction failure that might have occurred.
2

Client Connectivity
The Load Balancing (LB) cases
• On the Client Side
• On the Server Side
Production
Email
BATCH
Connection
Pool
MySCAN

Client Connectivity
Connect Time Load Balancing (CTLB) – on the client side
Production
Email
BATCH
Connection
Pool
MySCAN
PMRAC =
(DESCRIPTION =
(FAILOVER=ON)(LOAD_BALANCE=ON)
(CONNECT_DATA = (SERVER = DEDICATED) (SERVICE_NAME = Email)))

Client Connectivity
Connect Time Load Balancing (CTLB) – on the server side
Production
Email
BATCH
Connection
Pool
MySCAN?
• Traditionally, PMON dynamically registers the services to the specified listeners with:
• Service names for each running instance of the database and instance names for the DB
• The listener is updated with the load information for every instance and node as follows:
• 1-Minute OS Node Load Average all 30 secs.
• Number of Connections to Each Instance
• Number of Connections to Each Dispatcher

Client Connectivity
Use FAN for the Load Balancing cases
• Connect Time Connection Failover (CTCF)
• Runtime Connection Failover (RTCF)
RAC
Database
Instance1
Instance2
Instance3
I’m busy
I’m very busy
I’m idle
30% connections
10% connections
60% connections

Client Connectivity
Use FAN for the Load Balancing cases
• Also via AQ (Advanced Queuing) based notifications
• Background is always the Load Balancing Advisory
• For more information, see:
• Oracle® Real Application
Clusters Administration and
Deployment Guide 11g Release 2:
5 Introduction to Automatic Workload Management
RAC
Database
Instance1
Instance2
Instance3
I’m busy
I’m very busy
I’m idle
30% connections
10% connections
60% connections
MySCAN

Node Membership

The Oracle RAC Architecture
Oracle Grid Infrastructure 11g Release 2 process overview
OS OS
Oracle Grid Infrastructure
Node
Membership
HA Framework
ASM Instance
OS• My Oracle Support (MOS)
• Note 1053147.1 - 11gR2 Clusterware
and Grid Home - What You Need to Know
• Note 1050908.1 - How to Troubleshoot
Grid Infrastructure Startup Issues

Oracle Clusterware Architecture
Node Membership Processes and Basics
SAN
Network
SAN
Network
Public LanPublic Lan
CSSDCSSDCSSD
Voting
Disk
Private Lan /
Interconnect
Oracle Clusterware
Main processes involved:
• CSSD (ora.cssd)
• CSSDMONITOR
• was: oprocd
• now: ora.cssdmonitor

What does CSSD do?
• Monitors nodes using 2 communication channels:
– Private Interconnect  Network Heartbeat
– Voting Disk based communication  Disk Heartbeat
• Evicts (forcibly removes nodes from a
cluster) nodes dependent on heartbeat
feedback (failures)
CSSDCSSD
“Ping”
“Ping”
Oracle Clusterware

Interconnect basics – network heartbeat
CSSDCSSD
“Ping”
• Each node in the cluster is “pinged” every second
• Nodes must respond in css_misscount time (defaults to 30 secs.)
– Reducing the css_misscount time is generally not supported
• Network heartbeat failures
will lead to node evictions
– CSSD-log:
[date / time] [CSSD][1111902528]
clssnmPollingThread: node mynodename
(5) at 75% heartbeat fatal, removal
in 6.770 seconds

Voting Disk basics – disk heartbeat
CSSDCSSD
“Ping”
1 • Each node in the cluster “pings” (r/w) the Voting Disk(s) every second
• Nodes must receive a response in (long / short) diskTimeout time
– IF I/O errors indicate clear accessibility problems  timeout is irrelevant
• Disk heartbeat failures
will lead to node evictions
– CSSD-log: …
[CSSD] [1115699552] >TRACE:
clssnmReadDskHeartbeat:
node(2) is down. rcfg(1) wrtcnt(1)
LATS(63436584) Disk lastSeqNo(1)

Voting Disk basics – Structure
2 • Voting Disks contain dynamic and static data:
– Dynamic data: disk heartbeat logging
– Static data: information about the nodes in the cluster
• With 11.2.0.1 Voting Disks got an “identity”:
– E.g. Voting Disk serial number: [GRID]> crsctl query css votedisk
1. 2 1212f9d6e85c4ff7bf80cc9e3f533cc1 (/dev/sdd5) [DATA]
• Voting Disks must therefore not be copied using “dd” or “cp” anymore
Node information Disk Heartbeat Logging

Voting Disk basics – Simple Majority rule
CSSDCSSD
“Ping”
3 • Oracle supports redundant Voting Disks for disk failure protection
• “Simple Majority Rule” applies:
– Each node must “see” the simple majority of configured Voting Disks
at all times in order not to be evicted (to remain in the cluster)
 trunc(n/2+1) with n=number of voting disks
configured and n>=1

CSSDCSSD
• Same principles apply
• Voting Disks are just
geographically dispersed
• http://www.oracle.com/goto/rac
– Using standard NFS to support
a third voting file for extended
cluster configurations (PDF)
Simple Majority rule – in extended clusters

• Oracle ASM auto creates 1/3/5 Voting Files
– Voting Disks reside in one diskgroup only
– Based on Ext/Normal/High redundancy
and on Failure Groups in the Disk Group
– Per default there is one failure group per disk
– ASM will enforce the required number of disks
– New failure group type: Quorum Failgroup
[GRID]> crsctl query css votedisk
1. 2 1212f9d6e85c4ff7bf80cc9e3f533cc1 (/dev/sdd5) [DATA]
2. 2 aafab95f9ef84f03bf6e26adc2a3b0e8 (/dev/sde5) [DATA]
3. 2 28dd4128f4a74f73bf8653dabd88c737 (/dev/sdd6) [DATA]
Located 3 voting disk(s).
Voting Disks in Oracle ASM does not change their usage

Oracle Cluster Registry (OCR) placement in Oracle ASM
• The OCR is managed like a datafile in ASM (new type)
• It adheres completely to the redundancy settings for the diskgroup (DG)
• There can be more than one OCR location in more than one DG (DG:OCR  1:1)
• Recommendation is 2 OCR locations, 1 in DATA, 1 in FRA for example

Backup of Clusteware files is fully automatic (11.2+)
• Clusterware Files (managed in ASM) enables fully Automatic Backups:
• The Voting Disks are backed up into the OCR
• Any configuration change in the cluster (e.g. node
addition) triggers a new backup of the Voting Files.
• A single, failed Voting Disks is restored by Clusterware automatically
within a Disk Group, if sufficient disks are used – no action required
• Note: Do not use DD to back up the Voting Disks anymore!
• The OCR is backed up automatically every 4 hours
• Manual Backups can be taken as required
• ONLY IF all Voting Disks are corrupted or failed
AND (all copies of) the OCR are also corrupted or unavailable
THEN manual interference would be required – the rest is automatic.

• Evicting (fencing) nodes is a preventive measure (it’s a good thing)!
• Nodes are evicted to prevent consequences of a split brain:
– Shared data must not be written by independently operating nodes
– The easiest way to prevent this is to forcibly remove a node from the cluster
Fencing Basics
Why are nodes evicted?
CSSDCSSD
1 2

Fencing Basics
How are nodes evicted? – STONITH
• Once it is determined that a node needs to be evicted,
– A “kill request” is sent to the respective node(s)
– Using all (remaining) communication channels
• A node (CSSD) is requested to “kill itself”  “STONITH like”
– “STONITH” foresees that a remote node kills the node to be evicted
CSSDCSSD
1 2

Fencing Basics
EXAMPLE: Network heartbeat failure
CSSDCSSD
1 2
2
• The network heartbeat between nodes has failed
– It is determined which nodes can still talk to each other
– A “kill request” is sent to the node(s) to be evicted
 Using all (remaining) communication channels  Voting Disk(s)
 A node is requested to “kill itself”; executer: typically CSSD

Fencing Basics
What happens, if CSSD is stuck?
CSSDCSSD
1 2
2
CSSD
CSSDmonitor
• A node is requested to “kill itself”
• BUT CSSD is “stuck” or “sick” (does not execute) – e.g.:
– CSSD failed for some reason
– CSSD is not scheduled within a certain margin
 OCSSDMONITOR (was: oprocd) will take over and execute
• See also: MOS note
1050693.1 -
Troubleshooting 11.2
Clusterware Node
Evictions (Reboots)

Fencing Basics
How can nodes be evicted?
• Oracle Clusterware 11.2.0.1 and later supports IPMI (optional)
– Intelligent Platform Management Interface (IPMI) drivers required
• IPMI allows remote-shutdown of nodes using additional hardware
– A Baseboard Management Controller (BMC) per cluster node is required
CSSDCSSD
1

Fencing Basics
EXAMPLE: IPMI based eviction on heartbeat failure
CSSD
1
• The network heartbeat between the nodes has failed
– It is determined which nodes can still talk to each other
– IPMI is used to remotely shutdown the node to be evicted

Fencing Basics
Which node gets evicted?
CSSDCSSD
1 2
• Voting Disks and heartbeat communication is used to determine the node
• In a 2 node cluster, the node with the lowest node number should survive
• In a n-node cluster, the biggest sub-cluster should survive (votes based)

Fencing Basics
Cluster members can escalate a kill request
Oracle RAC
DB Inst. 1
Oracle RAC
DB Inst. 2
Oracle Clusterware
• Cluster members (e.g Oracle RAC instances) can request
Oracle Clusterware to kill a specific member of the cluster
• Oracle Clusterware will then attempt to kill the requested member
Inst. 1:
kill inst. 2

Fencing Basics
Oracle RAC
DB Inst. 1
Oracle RAC
DB Inst. 2
Oracle Clusterware
• If the requested member kill is unsuccessful, a node eviction escalation can be issued,
which leads to the eviction of the node, on which the particular member currently resides
Inst. 1:
kill inst. 2

Fencing Basics
Oracle RAC
DB Inst. 1
Oracle RAC
DB Inst. 2
Oracle Clusterware
Inst. 1:
kill inst. 2

Fencing Basics
Oracle RAC
DB Inst. 1
Oracle Clusterware

Re-Bootless Node Fencing
With 11.2.0.2 onwards, fencing may not mean re-boot
Oracle Clusterware
• Until Oracle Clusterware 11.2.0.2, fencing meant “re-boot”
• With Oracle Clusterware 11.2.0.2, re-boots will be seen less, because:
– Re-boots affect applications that might run an a node, but are not protected
– Customer requirement: prevent a reboot, just stop the cluster – implemented...
CSSDCSSD
App X App Y
RAC DB
Inst. 1
RAC DB
Inst. 2

How it works…
• With Oracle Clusterware 11.2.0.2, re-boots will be seen less:
– Instead of fast re-booting the node, a graceful shutdown of the stack is attempted
• It starts with a failure – e.g. network heartbeat or interconnect failure
Oracle Clusterware CSSDCSSD
App X App Y
RAC DB
Inst. 1
RAC DB
Inst. 2

How it works…
App X App Y
RAC DB
Inst. 1
• Then IO issuing processes are killed; it is made sure that no IO process remains
– For a RAC DB mainly the log writer and the database writer are of concern

How it works…
App X App Y
RAC DB
Inst. 1
• Once all IO issuing processes are killed, remaining processes are stopped
– IF the check for a successful kill of the IO processes, fails → reboot

How it works…
Oracle ClusterwareCSSD
App X App Y
RAC DB
Inst. 1
• Once all remaining processes are stopped, the stack stops itself with a “restart flag”
OHASD

How it works…
Oracle ClusterwareCSSD
App X App Y
RAC DB
Inst. 1
• OHASD will finally attempt to restart the stack after the graceful shutdown
OHASD

EXCEPTIONS
• With Oracle Clusterware 11.2.0.2, re-boots will be seen less, unless…:
– IF the check for a successful kill of the IO processes fails → reboot
– IF CSSD gets killed during the operation → reboot
– IF cssdmonitor (oprocd replacement) is not scheduled → reboot
– IF the stack cannot be shutdown in “short_disk_timeout”-seconds → reboot
App X App Y
RAC DB
Inst. 1
RAC DB
Inst. 2

The Interconnect

The Interconnect
Heartbeat and “memory channel” between instances
Interconnect
with switch
Public Lan
SAN switch
Client
Network
Node 1 Node 2 Node NNode N-1

The Interconnect
Redundant Interconnect Usage
Node 1 Node 2
HAIP1
HAIP2
HAIP3
HAIP4
• Redundant Interconnect Usage can be used as a bonding alternative
– It works for “private networks” only; the nodeVIPs use a different approach
– It enables HA and Load Balancing for up to 4 NICs per server (on Linux / Unix)
– It can be used by Oracle Databases 11.2.0.2 and Oracle Clusterware 11.2.0.2
– It uses so called HAIPs that are assigned to the private networks on the server
– The HAIPs will be used by the database and ASM instances and processes
1

The Interconnect
Redundant Interconnect Usage
Node 1 Node 2
HAIP1
HAIP2
HAIP3
HAIP4
• A multiple listening endpoint approach is used
– The HAIPs are taken from the “link-local” (Linux / Unix) IP range (169.254.0.0)
– To find the communication partners, multicasting on the interconnect is required
– With 11.2.0.3 Broadcast is a fallback alternative (BUG 10411721)
– Multicasting is still required on the public lan for MDNS for example.
– Details in My Oracle Support (MOS) Note with Doc ID 1212703.1:
11.2.0.2 Grid Infrastructure Install or Upgrade may fail due to Multicasting
2

The Interconnect
Redundant Interconnect Usage and the HAIPs
Node 1 Node 2
HAIP1
HAIP2
HAIP3
HAIP4
• If a network interface fails, the assigned HAIP is failed over to a remaining one.
• Redundant Interconnect Usage allows having networks in different subnet
• You can either have one subnet for all networks or a different one for each
• You can also use VLANs with the interconnect. For more information see:
• Note 1210883.1 - 11gR2 Grid Infrastructure Redundant Interconnect and ora.cluster_interconnect.haip
• Note 220970.1 - RAC: Frequently Asked Questions - How to use VLANs in Oracle RAC? AND
Are there any issues for the interconnect when sharing the same switch as the public network by using VLAN to separate the network?
HAIP1 HAIP3

Understanding oracle rac internals part 2 - slides

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to Understanding oracle rac internals part 2 - slides (20)

Recently uploaded (20)

Understanding oracle rac internals part 2 - slides