topic DATA COLLECTOR in Operating System - HP-UX

DATA COLLECTOR

Madanagopalan S — Wed, 18 Oct 2000 02:39:49 GMT

I am trying to collect the data using 'extract' command. I can able to extract
cpu, disk, network and memory performance data. How can I extract data
for 'bycpu' metrics from scopeux?

Currently I am using glance adivser mode to collect the data and I found scopeux
is running and its also log the performance data.

How can I collect bycpu and system table metrics from scopeux.

Re: DATA COLLECTOR

Sean Bastille — Wed, 18 Oct 2000 02:47:20 GMT

Sorry, MeasureWare doesn't collect per CPU metrics. The data is obviously there, as you can view it through glance, but MeasureWare does not collect it for history.

Re: DATA COLLECTOR

Madanagopalan S — Wed, 18 Oct 2000 02:51:06 GMT

How about the system tables values. Is it possible to collect the
system tables values from measureware using 'tbl' metrics.

Re: DATA COLLECTOR

Sean Bastille — Wed, 18 Oct 2000 02:53:36 GMT

I suppose I could have been a little more detailed in my response

scopeux is a MeasureWare process. Most likely you installed the GlancePlus Pak, which includes MeasureWare and Glance. MeasureWare is used for collecting historical information, and for generating alarms when certain monitoring thresholds are reached. Glance is a real-time kernel monitoring tool.

scopeux gets it's data from a process called midaemon. Both midaemon and glance get there data from ki (kernel implimentation). Both have access to by cpu metrics, however MeasureWare is not configured to log that information.

extract is also a MeasureWare tool that looks at the data in /var/opt/perf/datafiles, so it is limited by what MeasureWare collects.

Re: DATA COLLECTOR

Sean Bastille — Wed, 18 Oct 2000 02:58:49 GMT

I'm not sure exactly what metrics you're looking for, but a complete list of metrics available on various platforms and their descriptions should be included with your distribution in /opt/perf/paperdocs/mwa/C/metrics.txt.

Re: DATA COLLECTOR

Jim Welch — Wed, 18 Oct 2000 21:18:45 GMT

On bycpu metrics in mwa; they don't exist in the current versions...That has been frequently requested and *may* be added in a future version. You can log that data to disk in an ascii text file using glance in adviser mode if you need the info over a day or so to troubleshoot a problem.

On the tbl_* metrics, it depends. Several metrics were added to MeaureWare in version C.02.40 for HP-UX 10.20, (they were added earlier on 11.0). There should be documentation on the available metrics in /opt/perf/paperdocs/mwa/C/ (either text or pdf files). You can also run extract and select export and create a new report file (point to /var/opt/perf/datafiles/logglob). This will create a report file template with all the available metrics that extract sees and you can uncomment out the ones you want to export. You can also run /opt/perf/bin/repinfo if you have Perfview installed and it will list the available metrics from all running rep_server processes on the system you point it to. (repinfo -s yoursystem.yourdomain)

Re: DATA COLLECTOR

Tk Lau — Thu, 19 Oct 2000 02:20:57 GMT

By looking to the report file of measureware /var/opt/perf/reptall, stats by CPU cannot be found there. So I would suggest you to use "sar" to get the statistic

To collect the current date stats
# sar -M
To collect particular date stats
# sar -M -f /var/adm/sa/sa[$Day}

Before hand, you need to enable the sar collecting.

Re: DATA COLLECTOR

Tk Lau — Thu, 19 Oct 2000 02:20:58 GMT