Orphaned iostat run away consuming excessive cpu

saffer

Hi All,

I have noticed the following behaviour on all the versions of lpar2rrd agent on AIX.
Randomly we get a case of orphaned iostat processes that runaway and consume a lot of cpu.
What I believe may be the cause, is if the lpar2rrd server that the agent is sending information to is rebooted, then the lpar2rrd agents get into some weird state and disconnect the iostat and orphan it.
When I get this issue, it is usually a whole bunch of systems simultaneously, but not all systems.
The impact is pretty severe when it is on a virtual IO server.
Any thoughts.

Pavel

Hi,

iostat is not our tool, we use it as per its documentation, nothing special.

iostat -Dsal  60 1

we have not got reported such behaviour.

What I suspect are either hdisk or fcs adapters in strange status, just a guess ...

Involve AIX support, it is AIX issue in fact, we are not able to help you.

let us know if you get anything from support.

let us know if you want to skip iostat in the agent (you miss adapter stats them), we would create special build for you or better implemented some parametr in the agent ...