I have two questions.
1. According to the manual ( http://www.ibm.com/developerworks/wikis/display/WikiPtype/nmon+Manual )
Disk Statistics - double counting
It has been reported that multiple path I/O like EMC Powerpath and DAC drivers for FAST results in the disk statistics being double what they should be. This is a feature of the libperfstat library - nmon just reports what it finds. If you supply my sample captures I may be able to work out heuristics to avoid this for the next version. This may be improving in future AIX versions.
I am interested in using this tool to support oracle database cluster on Oracle Enterprise Linux 5 and with Powerpath to an EMC Clarrion providing Luns. However I am concerned about the above statement if DBA's cry wolf based on above incorrect information. Does this issue still occur in the current linux version ?
From looking at the linux readme, is the workaround to define disks as user defined disk groups ? ( http://www.ibm.com/developerworks/wikis/display/WikiPtype/README+nmon+for+Linux+version+11d )
Finally, left to its own devices, linux powerpath will enumerate the paths as /dev/emcpowerxx
Is this different from how nmon expects the powerpath devices to be enumerated and will it make the powerpath section in nmon not see the powerpath devices ?
Ahh, I was trying to place where I say a reference to the last question I had. Its from the excel analyser of the nmon output. From the manual ( ref http://www.ibm.com/developerworks/wikis/display/Wikiptype/nmonanalyser ( Word doc in v3.3 download ) )
NMON_Analyser detects the presence of an EMC/PowerPath subsystem by scanning the input file for the string “hdiskpower” before starting the analysis. EMC PowerPath creates devices called “hdiskpowern” which each map to multiple hdisks. They are therefore comparable to the vpaths generated by ESS/SDD. However, unlike vpaths, hdiskpower devices appear to AIX (and, therefore NMON) as real disks and NMON therefore records their activity on the DISK sheets.
Note the string it uses to determine if the device is a powerpath device.
Is this relevent posted here or is it outside nmon discussion here ?
Some of the comments here are really to do with nmon for AIX. The libperfstat is an AIX feature. for Linux the data is extracted from /proc/diskstats (from memory). and yes Linux can and does often double up the stats in this file.
It is a policy in nmon to use what we find in this and other files and not to make judgements of if two items are duplicates or related. This stops nmon from inaccurately reporting the data. It is better to have duplicates that are obvious than accidentally hid good information as you can never spot that.
From looks at lots on captured data for different machines, platforms and Linux distro and can safely say there is no consistence in the names used for disks and paths (in fact barking mad comes in to mind) a year ago we had to allow for punctuation symbols within disk names !
This means we are sticking to the "use what we find" and don't try being clever because its error prone.
Smart guys that notice the duplication can use the User Defined Disk Groups to eliminate them.
nmon for Linux has no PowerPath handling code.
I hope this helps, thanks Nigel Griffiths
I run into this same issue and there is a small patch I used to "tailor" nmon's reporting.
Please see another topic "Patch to exclude duplicated disk/net devices", above.
It is helping me alot right now ;-).