【RAC】DRM hang causes frequent RAC Instances Reconfiguration (Doc ID 1528362.1)

In this Document

Symptoms

Changes

Cause

Solution

References

---

APPLIES TO:

Oracle Database - Enterprise Edition - Version 11.1.0.7 to 11.2.0.3 [Release 11.1 to 11.2]
Oracle Database - Enterprise Edition - Version 11.2.0.4 to 11.2.0.4 [Release 11.2]
Information in this document applies to any platform.

SYMPTOMS

- RAC Instances freezes during DRM for 100 secs or more.

- DB Alert log shows that all RAC instances undergo reconfiguration at the same time, but there are no instance crashes

Node 1 DB Alert Log

Node 2 DB Alert Log

Sat Jul 14 14:17:04 2012 Reconfiguration started (old inc 70, new inc 72) List of instances: 1 2 (myinst: 1)  Global Resource Directory frozen Communication channels reestablished Master broadcasted resource hash value bitmaps Non-local Process blocks cleaned out Sat Jul 14 14:17:04 2012 LMS 1: 0 GCS shadows cancelled, 0 closed, 0 Xw survived Sat Jul 14 14:17:04 2012 LMS 0: 0 GCS shadows cancelled, 0 closed, 0 Xw survived Set master node info  Submitted all remote-enqueue requests Dwn-cvts replayed, VALBLKs dubious All grantable enqueues granted Sat Jul 14 14:17:13 2012 minact-scn: Master returning as live inst:2 has inc# mismatch instinc:70 cur:72 errcnt:0

Sat Jul 14 14:17:04 2012 Reconfiguration started (old inc 70, new inc 72) List of instances: 1 2 (myinst: 2)  Global Resource Directory frozen Communication channels reestablished Sat Jul 14 14:17:04 2012 * domain 0 valid = 1 according to instance 1  Master broadcasted resource hash value bitmaps Non-local Process blocks cleaned out Sat Jul 14 14:17:04 2012 LMS 0: 0 GCS shadows cancelled, 0 closed, 0 Xw survived Sat Jul 14 14:17:04 2012 LMS 1: 0 GCS shadows cancelled, 0 closed, 0 Xw survived Set master node info  Submitted all remote-enqueue requests Dwn-cvts replayed, VALBLKs dubious All grantable enqueues granted Sat Jul 14 14:18:03 2012 Submitted all GCS remote-cache requests

- Lmon trace shows that DRM quiesce step hangs:

*** 2012-07-14 14:14:51.187
 CGS recovery timeout = 85 sec
Begin DRM(231) (swin 1)
* drm quiesce

*** 2012-07-14 14:17:03.752
* Request pseudo reconfig due to drm quiesce hang 
2012-07-14 14:17:03.752735 : kjfspseudorcfg: requested with reason 5(DRM Quiesce step stall)

*** 2012-07-14 14:17:03.766
kjxgmrcfg: Reconfiguration started, type 6
CGS/IMR TIMEOUTS:
 CSS recovery timeout = 31 sec (Total CSS waittime = 65)
 IMR Reconfig timeout = 75 sec
 CGS rcfg timeout = 85 sec
kjxgmcs: Setting state to 70 0.

- AWR Top waits are "gcs resource directory to be unfrozen" & "gc remaster"

CHANGES

Large Buffer Cache

CAUSE

This is caused by bug:
Bug 12879027 - Lmon trace file shows that Pseudo reconfigurations triggered by the DRM are hanging. DRM quiesce is timing out..

DRM has a number of steps. During the DRM quiesce step all ongoing block transfers for remastering are completed.
In this case, during the DRM quiesce step a hang occured due to an internal function hitting a timeout.
This is a bug condition that happens when the buffer cache is very large.

This hang then triggers a pseudoreconfiguration to prevent the instance from being killed by another instance.
This is the reason for the instance undergoing a reconfiguration without restarting.

SOLUTION

The issue is fixed in 11.2.0.2 GI PSU7, 11.2.0.3 GI PSU 3, 11.2.0.4 onwards, it's recommended to apply latest patchset and PSU

REFERENCES

NOTE:756671.1  - Oracle Recommended Patches -- Oracle Database
NOTE:390483.1  - DRM - Dynamic Resource management
BUG:12879027  - LMON PROCESS CAN GET STUCK IN DRM QUIESCE STEP TRIGGERING PSEUDO RECONFIGURATION

来自 “ ITPUB博客 ” ，链接：http://blog.itpub.net/29487349/viewspace-2124738/，如需转载，请注明出处，否则将追究法律责任。

转载于:http://blog.itpub.net/29487349/viewspace-2124738/