Skip to main content

Collecting data for Performance issues in AIX

The following are the few commands I used heavily for troubleshooting a performance Issue on the AIX servers:

1. tprof:
tprof -kes -x sleep 60
This will collect tprof data for 60 seconds and will store it in a file called sleep.prof

2. Collecting data for SPLAT:

splat command uses the output create by the trace command to analyze the lock performance on the system where you have multi-threaded applications running.
a. Collect trace data:
generate gensyms data
gensyms > gensyms.out
trace -aC all -o trace -3; sleep 60 ; trcstop
This will collect trace data for 60 seconds and will generate files starting with trace, trace-1 etc depending on the number of virtual cpus.
b. run splat command to analyze the lock performance:
splat -sa -da -S100 -i trace -n gensyms.out -o splat.out
The final analysis of the lock performance can be found in splat.out.

There are various options available for the above commands and also documents in AIX information center on how to interpret the output, which I am not covering here.

Popular posts from this blog

Commands to restart RMC connection (t...

Commands to restart RMC connection (to HMC from LPAR) It has become very common with the IBM HMC to LPAR (logical/micro partition) communication to drop for unknown reasons.  Most of the time this is not a problem unless there is a need to do a dynamic logical partition operation (or DLAPR operation to add/remove resources on the fly).  This will become evident during the DLPAR operation when HMC complains about having no RMC connection to LPAR in operation.  When this happens run the following commands on the LPAR in question before reattemping the operation.  The DLPAR operation will still work with out this connection, but the LPAR needs a restart to see the change in the resources.  Restart the RMC connection on the LPAR: # /usr/sbin/rsct/install/bin/recfgct # /usr/sbin/rsct/bin/rmcctrl -p Verify the connection by running: lsrsrc IBM.ManagementServer This will show the HMC IP/hostname and the LPAR information.

Network throughput test between 2 AIX servers

The easiest way to test throughput between 2 AIX servers is to do a FTP test, generating data transfer with ‘dd’ command.  This will provide the throughput/speed for the specified amount of data transfer: #src-system> ftp dst-system … … ftp>  put "|dd if=/dev/zero bs=32k count=10000" /dev/null 200 PORT command successful. 150 Opening data connection for /dev/null. 10000+0 records in 10000+0 records out 226 Transfer complete. 327680000 bytes sent in 1.406 seconds (2.277e+05 Kbytes/s) local: |dd if=/dev/zero bs=32k count=10000 remote: /dev/null ftp>bye   This is the same test IBM recommends in the redbook and also during a support call

HP-UX tape configuration

Here are few commands to configure tape drives in HP-UX 1.  Scan for new devices: # ioscan –fnC tape 2.  List tape drives currently recognized by the OS (without performing an actual hardware scan) # ioscan –fnkC tape 3.  remove all the special files used by the tape drives # rmsf  /dev/rmt/* 4.  Recreate the specil files for the tape drives # insf –C tape –e 5.  Remove a tape drive (that has a NO_HW state) # rmsf –H 0/7/1/0.1.24.255.5.5.0 Find the WWN for the HBAs: # ioscan –fnkC fc # fcmsutil /dev/fcd0 Find the WWN for the tape drives connected to the HBA: # fcmsutil /dev/fcd0 get remote all   **These are just a handful of commands to deal with the tape drives.  Not to be used as a procedure