Troubleshooting
Problem
Recent versions of AIX record performance data by default for the last 7 days. This can be useful for a number of purposes:
1. Researching the "general" and very broad performance trends for a server, such as broad trends about CPU use, memory used, and IO values.
2. Providing "limited" analysis when no other, higher-quality performance is available.
3. With other IBM and not IBM tooling, providing some simple and broad trend graphs.
4. It is automatically collected by the system with minimal system impact
5. By default it is continually collected
Please note this topasrec data is NOT collected off the system as part of an AIX snap. It would need to be sent into IBM separately, or as part of a perfpmr.sh collection, which does collect it by default.
Symptom
The performance data is recorded with “topasrec” and you can see the default /etc/inittab entry below:
# cat /etc/inittab | grep topas
xmdaily:2:once:/usr/bin/topasrec -L -s 300 -R 1 -r 6 -o /etc/perf/daily/ -ypersistent=1 2>&1 >/dev/null #Start local binary recording
By default 7 days worth of performance data are recorded, as defined by the -r flag.
By default data is stored in /etc/perf/daily as defined by the -o flag.
Reference:
https://www.ibm.com/support/knowledgecenter/en/ssw_aix_72/t_commands/topasrec.html
Whilst the default is to be collecting this data, this can be verified by the following command from the system in question:
# ps -elf | grep topasrec
The files will have a “.topas” file extension and the file name includes the date:
For example, you might see something like this:
# ls -al /etc/perf/daily total 88 drwxr-xr-x 2 root system 256 Aug 15 19:47 . drwxr-xr-x 5 root system 256 Jul 06 16:22 .. -rw-r--r-- 1 root system 44212 Aug 15 19:53 aix1_191015.topas
topasrec and perfpmr.sh
For any detailed performance analysis neither topas (nor nmon) will not provide enough detail to provide detailed problem analysis - nor likely root cause. However for a sudden AIX issue where perfpmr.sh was not running and no AIX memory dump is available, this has been sometimes known to supply some clues about the major system metrics and either eliminate or provide some possible avenues of further investigation.
One example is when a system fails due to a lack of paging space and it is important to determine if the growth in paging space was gradual over a number of hours or days; or a sudden spike. This single piece of information which can be gathered from topasrec data can be useful for the deployment of other IBM tools such as perfpmr.sh
If you do not have perfpmr data from the time of the problem; the automatically collected topasrec data for the day of the issue may provide the other clues we have.
Contained below is short instructional video by Alan Hunt from IBM UK that summarises and also demonstrates the use of these topasrec files.
.
.
Related Information
Document Location
Worldwide
Was this topic helpful?
Document Information
Modified date:
04 May 2021
UID
ibm11087239