Massive archivelog and extremely high contention

Question

this moring , i discovered that one of my DB is suffering from huge amount of archive log , redo (400M,5 group) , average 50 archive log per hour , comparing to only 3 archive logs duirng normal operation
during that time , i checked session_wait ,
SID EVENT P1 P1TEXTP2 P2TEXT
138 latch free 1.3835E+19 address98 number
163 latch free 1.3835E+19 address98 number
1 pmon timer 300 duration 0
50 enqueue 1414332420 name|mode 32237 id1
55 enqueue 1414332419 name|mode 32237 id1
113 db file sequential read 100 file# 17468 block#
158 db file sequential read 100 file# 1182763 block#
107 db file scattered read 100 file# 1037041 block#
9 smon timer 300 sleep time 0 failed
12 wakeup time manager 0
latch free and enquene are the probablly the most serious problem DB is facing ..
and to dig down undernearth , latch cache buffers chains is the wrost thing from v$latch ,statspack shows enquene takes 60% of the total time
my question is , since i'm kind of new to this troubleshooting things , can any of you share your most favorite SQL to gather info ...
i'd like to start with session_wait , could you share your view of troubleshooting route to check out those common issues
thanks ~

Answer 1

I would start using logminer on those massive archivelogs.
You need to know where those massive inserts/updates/deletes are coming from.
You don't mention a version. This is bad as most answers are inevitably version dependent.
You can run statspack on level 7, this will include segment statistics, so spreport will show which segments are most heavily hit.
Locking info is in v$lock, which links with v$transaction, v$session and v$session with v$sql.
The latch issue points to parsing. Something completely unrelated!
Did your firm recently hire new developers ;)?
Sybrand Bakker
Senior Oracle DBA