CPU on AIX running constantly at 80%

Posted by andrew.thornton@redprairie.com on 04-Aug-2010 06:42

Hi, I was hoping someone could give me a bit of confirmation on (or entirely dispute!) our view of something.

We have an OpenEdge 10.2A application running using appserver on an IBM AIX server (with Window's clients). AppServer and database are running on the same machine. We are running the application on AIX because it is our customers operating system of choice. They have recently setup nmon on the box and have started asking questions about cpu usage. Basically it runs at pretty much constantly 80% except for a peak at midnight every day when we run a backup. They actually only use the system between 6am and 11pm, so have convinced themselves that the system running at 80% when they're not actually in the system implies there is some configuration problem. The reason we have given them for this behaviour is that we run a 'daemon' process constantly on the server both overnight and during the day whilst they are working. In our view the Progress application server will essentially run at 80% - or whatever the default kernel settings are - whilst it is in use regardless of the 'amount' of use. So during the day 40+ users are in the system plus the daemon is running, cpu runs at 80% and system performance is fine. Overnight only the daemon is running, but effectively this process is then able to take up the full 80% cpu allocated to the Progress app-service - which it doesn't need because most of the time it is essentially just iteratively looping through a block and not doing anything - but it is allowed to use because nothing else is.

Is our view correct - or is there actually an underlying issue here that we need to look into?

If anything we think that on an AIX box its maybe more of a concern that our application is only ever alowed 80% cpu when its the only thing running on there, and maybe it should be at 90% rather than 80%?

Thanks for any help in advance,

Andrew.

All Replies

Posted by Tim Kuehn on 04-Aug-2010 08:00

If the application / server isn't doing anything, the box shouldn't be at 80% CPU.

Have you used any performance measuring tools to find out which process(es) are using up all the CPU?

Posted by andrew.thornton@redprairie.com on 04-Aug-2010 09:50

Hi, The application server is doing something - its essentially iterating through a loop and polling for interface files on the server. So it is continually doing some fairly basic processing - which shouldn't be processor intensive in terms of my understanding of how it would work on a Windows system.

We have three days output from NMON. Our customer are querying the headline statistics on this which are showing cpu use at 80% pretty much constantly (strictly it is hovering around 80% - its not at exactly 80%). Drilling down into some of the output they have sent  though I'm beginning to wonder if this is actually anything to do with us or not. There is one process ID that is taking up 90+ percent of the cpu usage constantly all day. This process is './slp_srvreg -D'. Anyone know if this is anything at all to do with Progress?

Thanks, Andrew.

Posted by Tim Kuehn on 04-Aug-2010 14:31

From http://publib.boulder.ibm.com/infocenter/dirinfo/toolkit/index.jsp?topic=/com.ibm.director.sdk.t1.doc/slp_srvreg.html

slp_srvreg transmits an SLP Service Registration and prints the results. Registration persists through reboots provided the lifetime of the registration has been exceeded.

based on the "-D" being there, I'd guess someone's got it in debug mode.

It's not associated with a PSC product though.

Posted by ChUIMonster on 11-Aug-2010 12:59

No, this is not normal.  Your customer is right to be concerned.

Unless your background process is exceptionally badly written it should not take up anything like that amount of CPU just to see if there is work for it to do.  It should be sleeping most of the time.  It should also be very easy to see if your processes are responsible -- just pop into NMON and check to see what process is using up all of the CPU.  If it seems to be one of yours then start digging into what it is really doing by poking about with PROMON or better yet, ProTop -- http://dbappraise.coom/protop.html

I think you're on to something with the "slp" process -- that's not Progress and it sounds like a much more likely source of the issue.

This thread is closed