DB performance

[Roshan]

Oracle DB 19c
KVM

Hello Erman,

users were experiencing performance issue from their applications. The only change is that new VMs were deployed on the DBs host. Does deploying new vms affect DB performance?

I checked AWR report but found no issue.

Regards,
Roshan

[Roshan]

awrrpt_1_3829_3830.html

[Roshan]

When I moved the DB VM to host 2 on KVM, performance was better. How can I investigate if there is any issue on a host on KVM?

[ErmanArslansOracleBlog]

you have a load caused by log file sync, library cache mutex X and DB CPU.
you have also a significant I/O waits that contributes to that load.
Your OS load average is high.

Please review -> https://ermanarslan.blogspot.com/2014/09/log-writer-waits-storage-exadatas.html


Did you check your DB while it is running on the second KVM host? Do you see the same workload there? Does your DB really perform better with the same workload in the KVM host 2?
Is KVM Host 1 more crowded than KVM Host 1?

Also do a OS performance check for the KVM.. KVM is on Linux, so Linux performance diagnostics will give you a clue, if you have a performance issue on KVM hypervisor level.

Also check Oracle Linux Virtualization Manager.

[Roshan]

Thanks for the update. The load is almost same on both hosts. DB was working faster on host1 1 week ago. I saw the tcp,nerwork usage and CPU throughput  was higher How can I check if the redo log files are located on SSD based on the ASM diskgroup

[ErmanArslansOracleBlog]

You will check it from OS side.
You will find the device names that are used by the ASM, and then check from OS and see if they are SSD or not.. Get support from your OS admin. That way, it will be quicker for you.

[ErmanArslansOracleBlog]

Now I checked your AWR quickly once again.
You don't have high log fileparallel write.. So probably your issue is not related with I/O..

Go through the following MOS note and check your high log file sync issue from every aspect ->

Troubleshooting: 'Log file sync' Waits (Doc ID 1376916.1)

[Roshan]

Hi,

I checked with support because disk was already on SSD. I was told to set _use_single_log_writer to true. I saw an improvement in logminer wait on DB.

Please restrict the log writer to single worker process:
ALTER SYSTEM SET "_use_single_log_writer"=TRUE SCOPE=SPFILE;

[ErmanArslansOracleBlog]

So now you are ok then? or almost ok..

[Roshan]

I am OK for the logminer wait issue. Now I need to optimise cpu at KVM level. I have opened another thread for this.