High GC wait 11.2.0.3
High GC wait 11.2.0.3+++ After went through related log, we can see the following info:
1. alert and lm* doesn't have much info, however, dia0 shows following:
Chain 1 Signature: 'gc current request'<='gc buffer busy acquire'
..
Chain 40 Signature: 'gc current request'<='gc buffer busy acquire'
Chain 41 Signature: 'gc current request'<='gc buffer busy acquire'
Chain 42 Signature: 'gc current request'<='gc buffer busy acquire'
2. Active session history dump shows :
- start from 2014-01-18 19:07:59, total session number increased from 5 to 110, seems worklad increased.
- Instance 1 many sessions wait for 'gc buffer busy acquire' and holder wait for 'gc current request'.
So we got Chain Signature: 'gc current request'<='gc buffer busy acquire'.
3. alert log shows DRM is disabled:
_gc_undo_affinity = FALSE
_gc_policy_time = 0
_gc_defer_time = 3
+++ Based on the above analysis, A hang occur in a RAC environment with waits of the form 'gc current request'<='gc buffer busy acquire' ,
This symptom is match Bug 13787307https://bug.oraclecorp.com/pls/bug/webbug_edit.edit_info_top?rptno=13787307 : HANG IN RAC WITH 'GC CURRENT REQUEST'<='GC BUFFER BUSY ACQUIRE' SIGNATURE.
We recommend you to implement the following solutions to fix this issue:
- Upgrade to 11.2.0.4 (Server Patch Set)
or
- If apply patch is not available, then Set _gc_bypass_readers=false temporary to workaround this issue, there is no major impact for database.
+++ Now # of LMS processes on all nodes (currently it's 3 on node1 but 2 on node2),
please confirm that you have 32 CPU on one node, If so, you may set # of LMS processes to 3 on all nodes.
4. Oracle strongly recommended to run OSW to monitor OS and NETWORK state
note 301137.1 - OSWatcher Black Box (Includes: )
页:
[1]