TRU64 allows programs to consume all cpu

[neilrotherham]

Gurus,
I am curious with regard to the following output from top:

load averages: 2.33, 2.43, 2.46 14:49:44
257 processes: 5 running, 34 waiting, 95 sleeping, 122 idle, 1 zombie
CPU states: 23.7% user, 0.0% nice, 75.8% system, 0.4% idle
Memory: Real: 1290M/4002M act/tot Virtual: 1138M/18625M use/tot Free: 530M

PID USERNAME PRI NICE SIZE RES STATE TIME CPU COMMAND
589801 onight 42 0 14M 491K run 64.0H 97.60% daemon
589800 onight 42 0 14M 491K run 63.9H 97.00% daemon
674190 oracle9 44 0 427M 24M sleep 0:26 1.00% oracle
674877 onight 44 0 8968K 6103K run 0:00 0.60% top
524288 root 0 0 5219M 184M run 102:59 0.30% kernel idle
673609 root 42 0 0K 0K WAIT 0:01 0.10% icssvr_daemon_
668441 oracle9 44 0 2296K 417K run 0:00 0.00% ksh
524463 root 42 0 1784K 106K sleep 4:08 0.00% update
524903 root 44 0 18M 1679K sleep 1:57 0.00% caad
525614 oracle9 44 0 1179M 4866K sleep 1:46 0.00% oracle
525325 root 44 0 3080K 548K sleep 1:40 0.00% clu_mibs
671576 oracle9 44 0 1222M 18M sleep 1:34 0.00% oracle
525429 oracle9 44 0 14M 2088K sleep 1:06 0.00% tnslsnr
667445 oracle9 44 0 444M 43M sleep 0:53 0.00% oracle
525619 oracle9 44 0 1176M 2113K sleep 0:44 0.00% oracle

Notice the two processes called daemon using 97% of the cpu apiece. This is a twin processor machine, and these are ordinary (not system or 3rd party) programs. This happens with another program on our DR machine as well. It would appear that when more than one instance of the same program is started, they start to eat up more and more CPU. If we kill the offenders, things return to normal on this box - which is decidedly not stressed (not yet in production). Here is the system info:
>
> /usr/sbin/psrinfo -v
Status of processor 0 as of: 11/05/02 15:17:19
Processor has been on-line since 10/30/2002 18:04:27
The alpha EV6.7 (21264A) processor operates at 667 MHz,
and has an alpha internal floating point processor.
Status of processor 1 as of: 11/05/02 15:17:19
Processor has been on-line since 10/30/2002 18:04:27
The alpha EV6.7 (21264A) processor operates at 667 MHz,
and has an alpha internal floating point processor.
>
> /usr/sbin/sizer -v
Compaq Tru64 UNIX V5.1 (Rev. 732); Wed Oct 2 12:03:37 BST 2002
>
Can anyone explain this to me? Is the problem with the O/S or with the programs? Is there any measure I can take to cure the problem? Currently, we run a cron job to clear up this mess, but in my opinion, we shouldn't have to do that!

 

[JezHodgkinson]

Hi, we've also experienced the same problem - the Daemon consumes virtually all of the procecessor. Any ideas?

Thanks

 

[dblair]

They are using a huge amount of CPU time - Can you give more details of these processes ?

 

[kjetiljensen]

Unfortnatly I don't know what the daemon process is and can not tell you why there is a race condition. From the little information you have given it seems like the daemon process could be io-bound and that the race condition is a race for some io resource. (system load should not be higher than user load (except for eg. backup)).

There is a lot of race condition fixes in the last patchkit if you have not already tried that.

Can you describe the daemon process more, or maby even a pointer to it.