Originally posted by: SystemAdmin

Hello,
Does anyone know if there are recommended AIX MPIO PCM settings for the following three parameters for use in relation to EMC Symmetrix:
dist_tw_width
dist_err_percent
timeout_policy

I need to know what the recommended values are for these three parameters so we can properly configure MPIO and I understand that these values should be specific to the storage device that is being communicated with, in this case EMC Symmetrix.

Any help would be much appreciated.

Regards,
Chris Quirke

Originally posted by: AtifSami

emc uses its own multipathing software i.e powerpath which is recomended by emc

Originally posted by: ogniemi

but powerpath doesn't work pretty well on clients using NPIV... especially with rootvg on hdiskpower devices where you can;t do mksysb or bosboot without "hacking" powerpath (some pprootdev combinations) etc. although EMC says that multipathing will still work even "lspv" doesn't show exactly what is after reboot - I don;t like it. it should not work in that way. EMC has money to deliver software working without such combinations (like Hitachi for exampel with their HDS).

it is not 21'st century software from enterprise storage vendor...

Originally posted by: mmveiga

You may not feel confortable using PowerPath on rootvg disks and don't want to change your opinion, but to share my experience I use it on NPIV clients, we always run pprootdev before the bosboot command and never had an issues, even when restoring mksysb images. The version you need to run is 5.5 and also make sure you have the latest service pack and fixes for it.

Regards,
Marcelo.

Originally posted by: tech100

Marcelo,

is pprootdev command safe for running on HACMP cluster systems where diskhb communication lines are configured with /dev/hdiskpowerX devices? is it safe to run pprootdev on running cluster without negative impact to cluster status (communication lines seen on cldump command)?

what is the status of hdiskpowerX devices after running pprootdev when PVIDs "migrates" from hdiskpower devices to plane EMC hdisk devices?

Originally posted by: mmveiga

Hi tech100,

I cannot think of any problem that would be caused by pprootdev on cluster heartbeat disks, those are not part of rootvg.

But please contact EMC support and verify with them too.

Regards,
Marcelo.

Originally posted by: adrianfrederic

nice post thanks for the information

Originally posted by: jvk

First two are path control module attributes and Link: AIX doc. has some info on this.
The timeout_policy is described in the Link: APAR text.

All are disabled by default, means usually not in use and therefore there is not much recommendations. You'll need to go through your own testing to see if it make sense to use them. Also I'm not sure if EMC host attachment kit for aix default PCM (mpio) provides all above hdisk attributes. Do you see the timeout_policy parameter and if yes, what does it show the lsattr -El hdiskN command for your Symmetrix disk?

Originally posted by: SystemAdmin

In ODM version 5.3.0.6 ftp://ftp.emc.com/pub/elab/aix/ODM_DEFINITIONS there were several enhancements for AIX MPIO support for Symmetrix which added dist_tw_width and dist_err_pcnt attributes. These were previously being used for CLARiiON MPIO devices.

dist_tw_width default deflt=50 defines the duration of a time window. This is the time frame which the MPIO error detection will cumulate I/Os returning with an error.

dist_err_percent default deflt=0 defines the percentage of "time windows" having an error allowed on a path before the path is failed due to poor performance.

There was a new AIX MPIO attribute implemented in AIX that both Symmetrix and CLARiiON take advantage of called timeout_policy. The timeout_policy set to "fail_path" will resolve many continuous performance degradation issues on MPIO devices connected via a path (or switch) going up and down. The timeout_policy values are defined as follows:
retry_path = Similar to existing PCM behavior, command timeout on path, will not cause immediate path failure. If a path that failed due to transport issues is recovered by a health check, then that path may be used immediately. This is the default setting.

fail_path = The Path will be failed on command timeout assuming it is not the last path in the path group. If a path that failed due to command timeouts recovers, the path will not be used for read/write I/O until a period of time has expired with no failures on that path. Enabling this feature may add a delay for routing I/O to paths that have just recovered.

There is also a third attribute called disable_path. This value was not implemented because it disables paths and they remain disabled and not used until user specifically runs chpath command. It was chosen not to implement this at this time.

The minimum requirements for timeout_policy support are:
5300-10 - use AIX APAR IZ98520
5300-11 - use AIX APAR IZ98444
5300-12 - use AIX APAR IZ98131
6100-03 - use AIX APAR IZ97848
6100-05 - use AIX APAR IZ97464

Originally posted by: MehtaVikash

In case of fail_path, can someone  help me to understand what is the value of period of time ("period of time has expired with no failures on that path") mentioned with "fail_path" attribute.

 

fail_path = The Path will be failed on command timeout assuming it is not the last path in the path group. If a path that failed due to command timeouts recovers, the path will not be used for read/write I/O until a period of time has expired with no failures on that path. Enabling this feature may add a delay for routing I/O to paths that have just recovered.