RAC Wait Event: gcs log flush sync 等待事件 转

RAC Wait Event: gcs log flush sync

https://www.hhutzler.de/blog/rac-wait-event_gcs_log_flush_sync/#overview-gcs-log-flush-sync-wait-event

Table of Contents

• Overview : gcs log flush sync wait event
• Monitor gv$session wait events for LGWR, LMS processes
• Gcs Log Flush Sync Wait histogramms
• References

Overview : gcs log flush sync wait event

• Before sending a reconstructed CR block or CUR block, LMS will verify that corresponding redo vectors are flushed to disk.
• LMS will flush request a log flush when there are uncomitted transactions within the required block -> Increasing commit frequence will increase this wait event too
• Use lfsdiag.sql script to understand the wait time suffered by LGWR process
• If the redo vector are not flushed, LMS on node2 need to wait for ‘gcs log flush sync’ event after requesting LGWR for a log flush, analogous to ‘log file sync’ event.
• in a healthy database 90 % of GCS Log flush Sync should be 2ms ore less
• Always check related instances for LOG FILE SYNC event as this event also reduces the Redo I/O bandwith
• If gcs log flush sync wait event is significant then you need to understand your LGWR performance ( you may use lfsdiag,sql script )

Monitor gv$session wait events for LGWR, LMS processes

SQL> select w.inst_id, w.sid,w.program, w.event , w.STATE , w.SECONDS_IN_WAIT SEC_in_WAIT , w.P1TEXT || ': '  || w.P1 || ' - '    || w.P2TEXT  || ': '  || w.P2 || ' - '
      || w.P3TEXT || ': '  || w.P3 as P1_P2_P3 from gv$session w where
        program like '%LGWR%' or program like '%LMS%' or  program like '%JDBC%' order by program;

INST_ID        SID PROGRAM                          EVENT                         STATE             SEC_IN_WAIT P1_P2_P3
------- ---------- -------------------------------- ---------------------------- ------------------ ----------- ---------------------------------------------     
      2         18 oracle@grac42.example.com (LGWR) log file parallel write      WAITED SHORT TIME  0 files: 1 - blocks: 1 - requests: 1
      2         13 oracle@grac42.example.com (LMS0) gcs log flush sync           WAITING            0 waittime: 30 - poll: 0 - event: 138

Gcs Log Flush Sync Wait histogramms

Script:@event_histogram_from_awr.sql

Acceptable gcs log flush sync histogramm (90 % of wait requests returns in less that 2 ms )
BEGIN_INTERVAL_TIME            Inst    SNAP_ID EVENT_NAME        WAIT_TIME_MILLI WAIT_COUNT
------------------------------ ---- ---------- ----------------  --------------- ----------
28-NOV-13 04.24.27.507 PM         1       876 gcs log flush sync              1      58940
                                  1       876 gcs log flush sync              2      51775
                                  1       876 gcs log flush sync              4       1433
                                  1       876 gcs log flush sync              8        605
                                  1       876 gcs log flush sync             16        271
                                  1       876 gcs log flush sync             32        181
                                  1       876 gcs log flush sync             64        133
                                  1       876 gcs log flush sync            128         78
                                  1       876 gcs log flush sync            256         7
                                  1       876 gcs log flush sync            512         1 

Gcs log flush sync wait histogramm with high wait times
28-NOV-13 04.25.58.021 PM         1       877 gcs log flush sync               1        304
                                  1       877 gcs log flush sync               2        662
                                  1       877 gcs log flush sync               4      11816
                                  1       877 gcs log flush sync               8      46981  
                                  1       877 gcs log flush sync              16      68740 
                                  1       877 gcs log flush sync              32        202
                                  1       877 gcs log flush sync              64        141
                                  1       877 gcs log flush sync             128         83
                                  1       877 gcs log flush sync             256          8
                                  1       877 gcs log flush sync             512          1

References

http://www.orainternals.com/wp-content/uploads/2010/02/riyaj_advanced_rac_troubleshooting_rmoug_2010_doc.pdf

1、event_histogram_from_awr.sql

set linesize 180
set pagesize 200
col BEGIN_INTERVAL_TIME format A30
col instance_number format 99 head Inst
break on  BEGIN_INTERVAL_TIME
/*
define event_name="gcs log flush sync"
define event_name="log file sync"
*/
select snaps.begin_interval_time, snaps.instance_number,snaps.snap_id,hist.event_name , hist.wait_time_milli, hist.wait_count
  from dba_hist_event_histogram hist, dba_hist_snapshot snaps
 where snaps.snap_id = hist.snap_id and snaps.instance_number = hist.instance_number and
   snaps.begin_interval_time > sysdate - 1/12   -- sysdate - 1/12 -> display AWR reports from the last 2 hours
   AND hist.event_name = lower ('&&event_name')
 order by  snaps.snap_id , snaps.instance_number, wait_time_milli;

Contents

1. 1. Overview : gcs log flush sync wait event
   2. Monitor gv$session wait events for LGWR, LMS processes
   3. Gcs Log Flush Sync Wait histogramms
   4. References

Overview : gcs log flush sync wait event

• Before sending a reconstructed CR block or CUR block, LMS will verify that corresponding redo vectors are flushed to disk.
• LMS will flush request a log flush when there are uncomitted transactions within the required block -> Increasing commit frequence will increase this wait event too
• Use lfsdiag.sql script to understand the wait time suffered by LGWR process
• If the redo vector are not flushed, LMS on node2 need to wait for ‘gcs log flush sync’ event after requesting LGWR for a log flush, analogous to ‘log file sync’ event.
• in a healthy database 90 % of GCS Log flush Sync should be 2ms ore less
• Always check related instances for LOG FILE SYNC event as this event also reduces the Redo I/O bandwith
• If gcs log flush sync wait event is significant then you need to understand your LGWR performance ( you may use lfsdiag,sql script )

Monitor gv$session wait events for LGWR, LMS processes

SQL> select w.inst_id, w.sid,w.program, w.event , w.STATE , w.SECONDS_IN_WAIT SEC_in_WAIT , w.P1TEXT || ': '  || w.P1 || ' - '    || w.P2TEXT  || ': '  || w.P2 || ' - '
      || w.P3TEXT || ': '  || w.P3 as P1_P2_P3 from gv$session w where
        program like '%LGWR%' or program like '%LMS%' or  program like '%JDBC%' order by program;

INST_ID        SID PROGRAM                          EVENT                         STATE             SEC_IN_WAIT P1_P2_P3
------- ---------- -------------------------------- ---------------------------- ------------------ ----------- ---------------------------------------------     
      2         18 oracle@grac42.example.com (LGWR) log file parallel write      WAITED SHORT TIME  0 files: 1 - blocks: 1 - requests: 1
      2         13 oracle@grac42.example.com (LMS0) gcs log flush sync           WAITING            0 waittime: 30 - poll: 0 - event: 138

Gcs Log Flush Sync Wait histogramms

Script:@event_histogram_from_awr.sql

Acceptable gcs log flush sync histogramm (90 % of wait requests returns in less that 2 ms )
BEGIN_INTERVAL_TIME            Inst    SNAP_ID EVENT_NAME        WAIT_TIME_MILLI WAIT_COUNT
------------------------------ ---- ---------- ----------------  --------------- ----------
28-NOV-13 04.24.27.507 PM         1       876 gcs log flush sync              1      58940
                                  1       876 gcs log flush sync              2      51775
                                  1       876 gcs log flush sync              4       1433
                                  1       876 gcs log flush sync              8        605
                                  1       876 gcs log flush sync             16        271
                                  1       876 gcs log flush sync             32        181
                                  1       876 gcs log flush sync             64        133
                                  1       876 gcs log flush sync            128         78
                                  1       876 gcs log flush sync            256         7
                                  1       876 gcs log flush sync            512         1 

Gcs log flush sync wait histogramm with high wait times
28-NOV-13 04.25.58.021 PM         1       877 gcs log flush sync               1        304
                                  1       877 gcs log flush sync               2        662
                                  1       877 gcs log flush sync               4      11816
                                  1       877 gcs log flush sync               8      46981  
                                  1       877 gcs log flush sync              16      68740 
                                  1       877 gcs log flush sync              32        202
                                  1       877 gcs log flush sync              64        141
                                  1       877 gcs log flush sync             128         83
                                  1       877 gcs log flush sync             256          8
                                  1       877 gcs log flush sync             512          1

资料二：

How To Solve “gcs log flush sync” Wait Event

Posted on December 30, 2018 by dbtut

In busy databases you may encounter this wait event. The general reason is the low redo log I/O performance or the slow interconnect connection.

As a solution of this wait event;

• You can increase the performance of Interconnect network. For example, instead of 1g cards, you can create an interconnect interface on 10g cards.
• You can move Redo log files onto faster disks.
• Or, you can set the value of the _cr_server_log_flush parameter to false. This is not suggested.

You can change the value of the parameter as follows.

 

1 2 3

SQL> alter system set "_cr_server_log_flush"=false scope=both sid='*';   System altered.