Troubleshooting
Problem
mmvdisk is designed for ease of use and allows for streamlined easier management of complex Spectrum Scale RAID storage configurations when compared with the previous administrative methods.
As part of the conversion process, mmvdisk will apply configuration adjustments that are proven to work well in most use cases. With the current behavior, the mmvdisk command will apply these settings by default, so care should be taken if your system has custom tunings for specific workloads.
These configurations belong to the following broader categories:
1.) File system and Spectrum Scale RAID server-side recommended tunings.
2.) Additional mmchconfig parameters whose names starts with “nsdRAID” will get reverted to their default values.
Category (1) covers the following settings that are applied as specified in /usr/lpp/mmfs/data/cst/default.paired.gnrcfg.
Systems with 128G or RAM will have the page pool raised from the previous 60% of physical memory to 70%.
This may cause memory contention with other components on the system, such as the memory used by OFED, and could result in poor performance or out of memory errors.
Category (2) covers all mmchconfig parameters not specified in /usr/lpp/mmfs/data/cst/default.paired.gnrcfg starting with “nsdRAID”.
The default values for all nsdRAID mmchconfig parameters may be examined with the following command:
“grep defaultValue /usr/lpp/mmfs/bin/mmlsconfig | grep nsdRAID”
The following mmchconfig parameters applied by mmvdisk after the conversion process in category (2) are correct and do not need to be reexamined:
nsdRAIDDiskFastFailset to “yes”
Symptom
Performance characteristics and system behavior change noticeably after an mmvdisk system conversion, resulting in poor performance or out of memory errors.
Environment
Systems deployed prior to ESS 5.3.6 will have custom tuning settings altered as a result of the mmvdisk conversion process to match new systems deployed with ESS 5.3.6 and later.
Resolving The Problem
If your ESS systems have been custom tuned for your environment, it may not be desirable to apply the new mmvdisk defaults. Since mmchconfig changes are hard to diagnose, it is recommended to capture the output of “mmlsconfig” on all ESS systems being converted to mmvdisk prior to the conversion process and compare them to the new values that fall into category (1) and category (2).
If the configuration values were not saved explicitly by the user prior to conversion, the following internal mmvdisk files will show the configuration differences for pre and post conversions:
/var/mmfs/tmp/mmvdisk.convert.<rg name>.before.<yymmdd>
/var/mmfs/tmp/mmvdisk.convert.<rg name>.after.<yymmdd>
For any additional questions, please contact IBM support.
Document Location
Worldwide
[{"Line of Business":{"code":"LOB26","label":"Storage"},"Business Unit":{"code":"BU058","label":"IBM Infrastructure w\/TPS"},"Product":{"code":"STHMCM","label":"IBM Elastic Storage Server"},"ARM Category":[{"code":"a8m3p000000GoICAA0","label":"Recovery Group Resign"}],"ARM Case Number":"","Platform":[{"code":"PF016","label":"Linux"}],"Version":"5.3.0"}]
Was this topic helpful?
Document Information
Modified date:
24 March 2021
UID
ibm16435583