Simpler Navigation for Servers and Operating Systems
Completed: a much simpler Servers and Operating Systems section of the Community. We combined many of the older boards, so you won't have to click through so many levels to get at the information you need. Check the consolidated boards here as many sub-forums are now single boards.
Operating System - Tru64 Unix
cancel
Showing results for 
Search instead for 
Did you mean: 

Problem with Collect file.

Orrin
Valued Contributor

Problem with Collect file.

Hi all,

Host running Tru64 5.1B patch kit 3.
Collect version
Collector Version = 2.0.6, Collect Datafile Version = 18, Compiled: Sep 28 2003

Command used to launch Collect is as follows.
collect -i360,720 -f /usr/users/perf/collect -H d23:50,12m -W 1h -M 10,15

After which we try to get the information we require using the following command.
/usr/sbin/collect -F -p /usr/users/perf/collect_user@23-Jul-23:50:02.cgz -s ccmd | /usr/sbin/cfilt 'cpu:IDLE' 'mem:Free#' > /usr/users/perf/res/result

We get the follwing error...

"Flushing Error: stream error

termination record missing from data file"

On investigating the problem I found the follwing in the patch kit 3 documentation.

Patch 1404.00
provides a fix for collect
- Corrects a condition in which collect cannot create a new data file from a data file that does not include a termination record.

Checked to see if this patch is applied using dupatch and confirmed that this patch is applied.

The same command works on a Tru64 5.1A system.
with no problems.

Any ideas are welcome. please help.. need these statistics, thank you...

Regards,
Orrin.
5 REPLIES
Hein van den Heuvel
Honored Contributor

Re: Problem with Collect file.

So is that data collection terminated or still ongoing?

Looks like you are analyzing on the same version as used for the collection. So there is no Datafile version mismatch problem

Still. Do you have a chance to try an older collect? On my (unpatched) V5.1B system I seem to have:
Collector Version = 2.00, Collect Datafile Version = 15, Compiled: Oct 16 2002

Does simple stuff like: 'collect -p | more' work?
How about:
collect -p | grep "termination"

I have alerted (Email) an collect devo to this topic. Maybe they'll respond.
If you get no response here, then I guess you could try an Email to: collect_support@compaq.com

Cheers,
Hein.

Orrin
Valued Contributor

Re: Problem with Collect file.

Hi Hein,

Thanks for your quick reply...

I have four data files, three of them are complete, the last one is still collecting data.

collect -p .cgz | more
shows me all the records with all the detail and the file seems ok...

collect -p .cgz | grep "termination"
return null.
The termination record is missing....

Have logged a call with HP , lets see what they come back with.

On running collgui, I can get statistics for the CPU ( summary ) busy idle etc.
But statistics for any other devices, like memory disk etc do not work.

The only older version of collect and collect utilities are on Tru 64 5.1A, and there is a datafile mismatch on that.

Once again, thanks for your prompt reply, will keep you informed.

Regards,
Orrin.
Fabrice Rafart_1
Occasional Visitor

Re: Problem with Collect file.

Hello,

Collect had some problem with compression. Try whitout (-oz) and see.

Regards,
Hein van den Heuvel
Honored Contributor

Re: Problem with Collect file.



just relaying a reply from Sreeni:

-----------
From my initial analysis it looks like some one has issued "kill -9" during data collection. Please confirm the same. Please let me know the exact steps followed to collect/play back the files in order to reproduce the problem.
------------

Hein.

Orrin
Valued Contributor

Re: Problem with Collect file.

Hi,

Thanks for your replies..

Have logged a call with HP, and an HP Engineer is working with me on the problem.

It's not a compression problem as collect -p works, and gives valid records.

The collect process is still running but the last record in the .cgz file is at 4:09 am. On further investigation the advfs cloning before backup, causes a problem with the collect process.

Have changed the time collect starts and rollover occurs, to circumvent the problem.

Altough, collect -p gives the details of alldevices etc in each record, cfilt and collgui can show only CPU Summary statistics..

Once again Thank You for replying.

Regards,
Orrin.