Oracle RAC CSS 超时计算 及 参数 misscount， Disktimeout 说明

一.概述
在之前的文章：
RAC的一些概念性和原理性的知识
http://blog.csdn.net/xujinyang/article/details/6837273

提到OCSSD这个进程是Clusterware最关键的进程，如果这个进程出现异常，会导致系统重启，这个进程提供CSS(ClusterSynchronizationService)服务。CSS服务通过多种心跳机制实时监控集群状态，提供脑裂保护等基础集群服务功能。
CSS服务有2种心跳机制：一种是通过私有网络的NetworkHeartbeat，另一种是通过VotingDisk的DiskHeartbeat.
这2种心跳都有最大延时，对于DiskHeartbeat，这个延时叫作IOT(I/OTimeout);对于NetworkHeartbeat,这个延时叫MC(Misscount)。这2个参数都以秒为单位，缺省时IOT大于MC，在默认情况下，这2个参数是Oracle自动判定的，并且不建议调整。

可以通过如下命令来查看参数值：
$crsctlgetcssdisktimeout
$crsctlgetcssmisscount

如：
[oracle@rac1 ~]$ crsctl get css disktimeout
200
[oracle@rac1 ~]$ crsctl get css misscount
60
这是这2个参数的默认值。

二.MOS 上相关的几篇文章
How to start/stop the 10g CRS ClusterWare[ID 309542.1]
10g RAC: Steps To Increase CSS Misscount,Reboottime and Disktimeout [ID 284752.1]
CSS Timeout Computation in OracleClusterware [ID 294430.1]
RAC Assurance Support Team: RAC and OracleClusterware Starter Kit and Best Practices (Generic) [ID 810394.1]

2.1修改CSSMisscount 步骤：
1)Shut down CRS on all but one node. For exact steps use Note 309542.1
2)Execute crsctl as root to modify the misscount:
$ORA_CRS_HOME/bin/crsctl set css misscount <n>
where <n> is the maximum i/o latency to the voting disk +1 second
3)Reboot the node where adjustment was made
4)Start all other nodes shutdown in step 1

With the Patch:4896338 for 10.2.0.1 thereare two additional settings that can be tuned.This change is incorporated into the 10.2.0.2 and 10.1.0.6patchsets.


Thesefollowing are only relevant on10.2.0.1 with Patch:4896338，In addition to MissCount, CSS now has two more parameters:
1)reboottime (default 3 seconds) - the amount of time allowed for a node to complete a reboot after the CSS daemon hasbeen evicted. (I.E. how long does ittake for the machine to completely shutdown when you do a reboot)
2)disktimeout (default 200 seconds) - the maximum amount of time allowed for a voting file I/O to complete; if thistime is exceeded the voting disk will be marked as offline. Note that this is also the amount of timethat will be required for initialcluster formation, i.e. when no nodes havepreviously been up and in a cluster.

$CRS_HOME/bin/crsctl set css reboottime <r> [-force] (<r> is seconds)
$CRS_HOME/bin/crsctl set css disktimeout <d> [-force] (<d>is seconds)

Confirm the new css misscount setting via ocrdump


2.2CSS Timeout Computation in OracleClusterware
2.2.1MISSCOUNTDEFINITION AND DEFAULT VALUES
The CSS misscount parameterrepresents the maximum time, in seconds, that a network heartbeat can be missedbefore entering into a cluster reconfiguration to evict the node. The followingare the default values for the misscount parameter and their respectiveversionswhen using Oracle Clusterware* in seconds:

OS	10g(R1 &R2)	11g
Linux	60	30
Unix	30	30
VMS	30	30
Windows	30	30

*CSS misscount default value when using vendor (non-Oracle)clusterware is 600 seconds. This is to allow the vendor clusterwareample time to resolve any possible split brain scenarios.
On AIX platforms with HACMP starting with 10.2.0.3 BP#1, themisscount is 30. This is documented inNote551658.1

2.2.2CSS HEARTBEATMECHANISMS AND THEIR INTERRELATIONSHIP
The synchronization servicescomponent (CSS) of the Oracle Clusterware maintains two heartbeat mechanisms
1.) the disk heartbeat to the voting deviceand
2.) the network heartbeat across theinterconnect which establish and confirm valid node membership in the cluster.
Bothof these heartbeat mechanisms have an associated timeout value. The diskheartbeat has an internal i/o timeout interval (DTO Disk TimeOut), in seconds,where an i/o to the voting disk must complete. The misscount parameter (MC), asstated above, isthe maximum time, in seconds, that a network heartbeatcan be missed. The disk heartbeat i/o timeout interval is directly related tothe misscount parameter setting. There has been some variation in thisrelationship
between versions as described below:

9.x.x.x	NOTE, MISSCOUNT WAS A DIFFERENT ENTITY IN THIS RELEASE
10.1.0.2	No one should be on this version
10.1.0.3	DTO = MC - 15 seconds
10.1.0.4	DTO = MC - 15 seconds
10.1.0.4+Unpublished Bug 3306964	DTO = MC - 3 seconds
10.1.0.4 with CRS II Merge patch	DTO =Disktimeout (Defaults to 200 seconds) Normally OR Misscount seconds only during initial Cluster formation or Slightly before reconfiguration
10.1.0.5	IOT = MC - 3 seconds
10.2.0.1 +Fix for unpublishedBug 4896338	IOT=Disktimeout(Defaults to 200 seconds) Normally OR Misscount seconds only during initial Cluster formation or Slightly before reconfiguration
10.2.0.2	Same as above (10.2.0.1 with PatchBug:4896338
10.1 - 11.1	During node join and leave (reconfiguration) in a cluster we need to reconfigure, in that particular case we use Short Disk TimeOut (SDTO) which is in all versions SDTO = MC – reboottime (usually 3 seconds)

Misscountdrives cluster membership reconfigurations and directly effects theavailability of the cluster. In most cases, the default settings for MC shouldbe acceptable. Modifying the default value of misscount not onlyinfluences the timeout intervalfor the i/o to the voting disk, but alsoinfluences the tolerance for missed network heartbeats across the interconnect.

2.2.3LONG LATENCIES TOTHE VOTING DISKS
If I/O latencies to the voting diskare greater than the default DTO calculations noted above, the cluster mayexperience CSS node evictions depending on (a)the Oracle Clusterware (CRS)version, (b)whether merge patch has been applied and (c)the state oftheCluster. More details on this are covered in the section "Change inBehavior with CRS Merge PATCH(4896338 on 10.2.0.1)".
Theselatencies can be attributed to any number of problems in the i/o subsystem orproblems with any component in the i/o path. The following is a non exhaustivelist of reported problems which resulted in CSS node eviction due to latenciesto the votingdisk longer than the default Oracle Clusterware i/o timeoutvalue(DTO):
1.QLogic HBA cards with a LinkDown Timeout greater than the default misscount.
2.Bad cables to the SAN/storagearray that effect i/o latencies
3.SAN switch (like Brocade)failover latency greater than the default misscount
4.EMC Clariion Array whentrespassing the SP to the backup SP greater than default misscount
5.EMC PowerPath path errordetection and I/O repost and redirect greater than defaultmisscount
6.NetApp Cluster (CFO) failoverlatency greater than default misscount
7.Sustained high CPU load whicheffects the CSSD disk ping monitoring thread
8.Poor SAN network configurationthat creates latencies in the I/O path.

The mostcommon problems relate to multi-path IO software drivers, and thereconfiguration times resulting from a failure in the IO path. Hardwareand (re)configuration issues that introduce these latencies should becorrected. Incompatible failover timeswith underlying OS, network or storagehardware or software may be addressed given a complete understanding of theconsiderations listed below.

Misscount should NOT be modified to workaround theabove-mentioned issues. Oracle support recommends that you apply thelatest patchset which changes the CSS behaviour.More details covered innext section.

2.2.4Change in BehaviorwithBug:4896338appliedon top of10.2.0.1
Starting with 10.2.0.1+Bug:4896338,CSSwill not evict the node from the cluster due to (DTO) I/O to voting disktaking more than misscount seconds unless it is during the initial clusterformation or slightly before reconfiguration.
So if we have aN number ofnodes in acluster and one of the nodes takes more than misscountsecondsto access the voting disk, the node will not be evicted as long asthe access to the voting disk is completed within disktimeoutseconds.Consequently withthispatch, there is no need to increasethe misscount at all.
Additionallythis merge patch introduces Disktimeout which is the amount of time thata lack of disk ping to voting disk(s) will be tolerated.

Note: applying the patch will notchange your value for Misscount.
The table below explains intheconditions under which the eviction will occur

NetworkPing	DiskPing	Reboot
Completes within misscount seconds	Completes withinMisscount seconds	N
Completes within Misscount seconds	Takes more than misscount seconds but less than Disktimeout seconds	N
Completes within Misscount seconds	Takes more than Disktimeout seconds	Y
Takes more than Misscount Seconds	Completes within Misscount seconds	Y

* By defaultMisscount is lessthan Disktimeout seconds

2.2.5CONSIDERATIONS WHENCHANGING MISSCOUNT FROM THE DEFAULT VALUE
1.Customers drive SLA and clusteravailability. The customer ultimately defines Service Levels and availabilityfor the cluster. Before recommending any change to misscount, the full impactof that change should be described and the impact to cluster availabilitymeasured.
2.Customers may have timeout andretry logic in their applications. The impact of delaying reconfiguration maycause 'artificial' timeouts of the application, reconnect failures andsubsequent logon storms.
3.Misscount timeout values areversion dependent and are subject to change. As we have seen, misscountcalculations are variable between releases and between versions within arelease. Creating a false dependency on misscount calculation in one versionmaynot be appropriate for later versions.
4.Internal I/O timeout interval(DTO) algorithms may change in later releases as stated above, there exists adirect relationship between the internal I/O timeout interval and misscount.This relationship is subject to change in later releases.
5.An increase in misscount tocompensate for i/o latencies directly effects reconfiguration times for networkfailures. The network heartbeat is the primary indicator of connectivity withinthe cluster. Misscount is the tolerance level of missed 'checkins' thattrigger cluster reconfiguration. Increasing misscount will prolong the time totake corrective action in the event of network failure or other anomalieseffecting the availability of a node in the cluster. This directly effectscluster availability.
6.Changing misscount toworkaround voting disk latencies will need to be corrected when the underlyingdisk latency is corrected, misscount needs to be set back to the default Thecustomer needs to document the change and set the parameter back to the defaultwhenthe underlying storage I/O latency is resolved.
7.Do not change default misscountvalues if you are running Vendor Clusterware along with OracleClusterware. The default values for misscount should not be changed when usingvendor clusterware. Modifying misscount in this environment may causeclusterwideoutages and potential corruptions.
8.Changing misscount parameterincurs a clusterwide outage. As note below, the customer will need to schedule
a clusterwide outage to make this change.
9.Changing misscount should notbe used to compensate for poor configurations or faulty hardware
10.Cluster and RDBMS availabilityare directly effected by high misscount settings.
11.In case of stretched clustersand stretched storage systems and a site failure where we loose one storage andN number of nodes we go into a reconfiguration state and then we revert toShortDiskTimeOut value as internal I/O timeout for the votings. Severalcasesare known with stretched clusters where when a site failure happen the storagefailover cannot complete within SDTO. If the I/O to the votings is blocked morethan SDTO the result is node evictions on the surviving side.

ToChange MISSCOUNT back to default PleaserefertoNote:284752.1
THIS IS THE ONLY SUPPORTED METHOD.NOT FOLLOWING THIS METHOD RISKS EVICTIONS AND/OR CORRUPTING THE OCR

10gRelease 2 MIRRORED VOTING DISKS AND VENDORMULTIPATHING SOLUTIONS
Oracle RAC 10g Release 2 allows formultiple voting disks so that the customer does not have to rely on amultipathing solution from a storage vendor. You can have n voting disks (up to31) where n = m*2+1 where m is the number of disk failures you wanttosurvive. Oracle recommends each voting disk to be on a separate physical disk.







-------------------------------------------------------------------------------------------------------