Subscribe

Streamlining DFM for separate Protection Manager and OpsMgr/PA servers

So, after doing alot of research on how to best optimize OpsMgr for large object environments, I have decided to go with the recommendation in tr-3440, which has one putting Protection Manager and Provisioning Manager on one server, and having Operations Manager and Performance Advisor on another.  It also goes on to say that we can disable the following options on the Protection Manager server side to increase performance:

agentMonInterval

ccMonInterval

envMonInterval

fcMonInterval

hostRBACMonInterval

opsMonInterval

SANHostMonInterval

srMonInterval

userQuotaMonInterval

Now, as you all know, there are a boat-load more options out there.  Are there not anymore that I can disable than these?  I mean, do I really need the "chargeback" options set, or the "autosupport", or the "cpu" monitoring intervals enabled?  The TR states that we only need that which is related to Protection Manager operations.  I hardly believe that the aforesaid is "related," and thus can be disabled, but then again, who knows?  If there are more, a complete list would be awesome!

Also, I assume that to disable any option you just the following sytax:  dfm options set agentMonInterval=disable

Thanks in advanced!

Re: Streamlining DFM for separate Protection Manager and OpsMgr/PA servers

Currently whatever you see is not relevant to protection manger.

Regards

adai

Re: Streamlining DFM for separate Protection Manager and OpsMgr/PA servers

Just to clarify, Adai, when you say "currently whatever you see," do you mean what I currently just pointed out, namely, the "chargeback," "autosupport," and "cpu," can be turned off? If so, we have a growing list of the following that can be disabled when we separate out DPM from DFM/PA:

agentMonInterval

ccMonInterval

envMonInterval

hostMonInterval

SANHostMonInterval

srMonInterval

userQuotaMonInterval

autosupportMonInterval

cpuBusyThresholdInterval

cpuMonInterval

Here are some more that I think could be turned off:

aggrFullThresholdInterval

qtreeFullThresholdInterval

volFullThresholdInterval

Since these would be monitored on another server, why would we need to have 2 monitoring "copies"?  Does DPM use these in any way?  To my knowledge, these threshold values are for event generation only and have nothing to do with informing DPM of volume, qtree, or aggregate utilization.  If I were to guess, when creating a dataset, DPM makes ZAPI calls to the filer to get its usage numbers before making its conformance calculations.

licenseMonInterval

ndmpMonInterval

Now, I know that the NMC, and thus DPM, uses ZAPI calls to get license and ndmp status information and not from the polling that DFM does, so I don't see that these are necessary as well.

cfMonInterval

clusterMonInterval

These, again, don't seem to have anything to do with DPM.

perfAdvisorPollInterval

perfAdvThreshViolationMonInterval

I would guess that if you disabled Performance Advisor, these would either disappear, or would automatically be disabled, at least in practice and thus would not need to be bothered with, but you never know.

I know you dealt with this sort of thing in the thread: http://communites.netapp.com/messages/55322 .  You gave an excellent list of options to trun off for a dedicated Performance Advisor server.  What I'm looking for is the same thing except for a dedicated DPM server.

Thanks alot for your help Adai!