DM_STAPE process running high.

[Larry Scheetz]

I have a L1000 machine running 11.0. This is my cell server for my OmniBack II software. I am also running some other interface software. When running a top the process DM_STAPE is constantly running at 88%. Any suggestions on why or how to fix?

Thanks
Larry

[Steven E. Protter]

Is this while actually running backups or all the time.

This patch might be relavent.

http://www2.itrc.hp.com/service/patch/patchDetail.do?patchid=PHSS_27809&context=hpux:700:11:00

I'm attaching a performance data collection script to help you gather further data if need be.

Good luck.

http://www.isnamerica.com
Call me anything you want, just don't forget to call me to dinner.
Steven E. Protter

[Steven E. Protter · Accepted solution]

Is this while actually running backups or all the time.

This patch might be relavent.

http://www2.itrc.hp.com/service/patch/patchDetail.do?patchid=PHSS_27809&context=hpux:700:11:00

I'm attaching a performance data collection script to help you gather further data if need be.

Good luck.

http://www.isnamerica.com
Call me anything you want, just don't forget to call me to dinner.
Steven E. Protter

[Larry Scheetz]

This is at all times. I only run backups from 02:00am - 7:00am in that time frame. As we are speaking it is at 88%? Thank you for your help.

[Michael Tully]

If your keen on not patching your system, you could look at not running the dm_stape process. This is quite a common occurance when in conjunction with tape libraries and the use of the omniback product. What it does is prevent SCSI resets to occur when EMS is running. Also you could change the 'st_ats_enabled' kernel parameter to 0 (zero)
Attached is a document that will assist you in turning off this particular EMS monitor.

[Andrew Merritt_2]

Larry,
dm_stape is the EMS Hardware Monitor for tape devices, part of the Online Diagnostics.

One thing to check is the revision of OnlineDiags that you have installed on the system, a new version is released every quarter. To see the version run
swlist -l bundle|grep OnlineDiag

If this is not a recent version, then you should upgrade the OnlineDiags to a more current release (check the patch catalog for any patches for that release, too). There have been changes so that dm_stape does not do any polling of tape devices, which has caused some problems in the past; I would not expect dm_stape to be using such high levels of CPU (unless a lot of errors are being generated when the device is in use, in which case you would expect to see EMS events being logged in /var/opt/resmon/log/event.log).

[Andrew Merritt_2]

Just to explain a bit more, with a recent version of OnlineDiags it should not be necessary to follow the instructions to disable dm_stape. The problem that work-around avoids was caused by dm_stape polling the tape devices on an hourly basis. This is no longer done (check the value for POLL_INTERVAL in /var/stm/config/tools/monitor/dm_stape.cfg, it should be '0' (zero)).

Andrew