HPE 9000 and HPE e3000 Servers
1755747 Members
3675 Online
108837 Solutions
New Discussion юеВ

Re: error message from procsize

 
Curtis Mullins
Advisor

error message from procsize

procsize is an unsupported process size monitoring tool from hp. i use this to monitor
growing processes. today i got his message:
"procsize error pid has more than a 1000 pregions. Unable to process." i am not much concerned about procsize as i am the actual process. does this mean that the process is having problems or could have? we are running 11.10 on an n-class box.
3 REPLIES 3
Jeff Schussele
Honored Contributor

Re: error message from procsize

Hi Curtis,

A process can have only one text & one data pregion, but it could have numerous shmem or shlib regions. So it is either attaching to a bunch of shared memory regions - kind of unlikely unless this is a monster program or very sloppily written - OR it can be accessing a boatload of SHLIB files - also unlikely, but still 1000+ is a LOT of pregions.
I suspect that it's creating a shared memory segment, attaching to it & then never cleaning up when it's done with it.

I think I'd attach Tusc to it & see just exactly what it's doing.

So the answer is probably - Yes - you're probably going to have trouble with this process because sooner or later you're going to exhaust a resource - probably memory - because it may not be cleaning up after itself properly.

HTH,
Jeff
PERSEVERANCE -- Remember, whatever does not kill you only makes you stronger!
Curtis Mullins
Advisor

Re: error message from procsize

thanks jeff

yes the process is a memory hog and it will run out of data space at 2gig. this was a fix to a memory leak problem. we monitor the process and restart it gracfully just before it reaches the limit. i just was curious if we were going to have unforseen problems. thanks for your help......
Bill Hassell
Honored Contributor

Re: error message from procsize

If the process is supposed to grab more than 2Gb, then it should be recompiled as a 64bit program. However, the 2Gb limit doesn't sound like local data (with 1000+ pregions). Try ipcs -bmop and see if there are many pages of small shared memory segments. If true, find the programmer and demand an explanation. This is not a design I would want in production. If the program needs a lot of shared memory, then it should ask for it in large chunks. Creating 1000+ pregions is going to create a lot of system overhead. Of course, the program may just be full of bugs and running away so it will ask for more memory forever (find a new programmer).


Bill Hassell, sysadmin