Krishna
Once saw very high ITIL waits on in single table in high frequency DML system.
Changed PK Index to Partition by Hash that reduced ITIL and quick response.
Thanks
Karth
Sent from my IPhone
On Apr 4, 2024, at 11:07 AM, Krishnaprasad Yadav <chrishna0007@xxxxxxxxx>--
wrote:
Dear Gurus ,
i have made observation of ITL waits in DB , and it lasted more than 4+hrs
i see that certain new DML was introduced newly in database , and same time
these ITL wait events triggered .
I validated top object i see 3 tables were seen as DML are using these tables
, i saw concurrency it was around 3k to 10K in 30 mins of snap .
Also table is non partitioned and it was only 3GB in size .
total execution in DAY was around 400K for all 3 DML statements , seeing
these bit surprise that how with such low concurrency its landed in issue
Also able extent management is AUTO .
i know increasing INITTRANS will help them (probably ?) but i am suspecting
its coming something out of DB , i validated OS watcher details , i see some
potential bottleneck but those stats which i suspect as bottleneck are seen
in good time as well , i validate CPU utilization which was high during start
of issue but after 10 mins it got down to 60% but spike remain in DB ,
validate IO busyness trend is similar to CPU .
It will be helpful if you can share any of experience from which i can relate
and take ahead my case .
Regards,
Krishna
