Subscribe

cache hit reporting

When I use sysstat -x I can see the cache hit % and cache age.   I'd like to get that same information out of Operations manager or Performance Advisor but I don't see those counters anywhere.

Are they available?  And if so, where?

Re: cache hit reporting

Andrew,

You will need to enable 'perfAdvisorShowDiagCounters' option on the Ops Manager server. Then you will be able to see cache related counters.

# dfm perf counter list <filer>:wafl:*:*:*:* | grep cache
wafl      wafl                     inode_cache_hit_tohead             per_sec   diag
wafl      wafl                     name_cache_miss_snapdir            per_sec   diag
wafl      wafl                     inode_cache_hit                    per_sec   diag
wafl      wafl                     inode_cache_miss                   per_sec   diag
wafl      wafl                     name_cache_hit                     per_sec   diag
wafl      wafl                     name_cache_miss                    per_sec   diag
wafl      wafl                     inode_cache_hit_nodes              per_sec   diag
wafl      wafl                     inode_cache_miss_nodes             per_sec   diag
wafl      wafl                     name_cache_miss_long                per_sec   diag

You can then add these counters to custom views in Performance Advisor.

Thanks,

Raj.

Re: cache hit reporting

Thanks,

How do you enable a counter on the Ops Mgr Server?

Andy

Re: cache hit reporting

Hi Andy,

     There is nothing like enabling counter in ops-mgr.

All counters come from ontap through counter manager.

By default PA show all basic counters exposed by counter manger for a ontap version.

Some counters are tagged as advanced/diag. For these counters to be shown

in ops-mgr you need to enable the option using

dfm option set perfAdvisorShowDiagCounters=yes.

Regards

adai

Re: cache hit reporting

Oracle Austin DC is interested in these diag counters. However, they are running some hefty DFM servers, most with a little over 100 filers being monitored. Stability is not too bad right now with the DFM/OpsMgr (3.7 going to 3.7.1). However, they are wondering if enabling these diag counters might drastically increase A) counter collection network load?, B) database storage of long term counters, or C) DFM server CPU load.

Should they be concerned turning these diag counters in terms of A, B  or C? 

For me it it opens up the more general question(s):

  #1) When you have enabled "perfAdvisorEnabled" are you then polling perf counters from every filer and storing them? -or- only when you use PA to create views and look at counters for a particular filer/object?

  #2) Whatever the answer to #1, does enabling "perfAdvisorShowDiagCounters" generate new polling/storage load? -or- are the diag counters being polled and stored all the time anyway and enabling "perfAdvisorDiagCounters" just exposes them in the PerfAdvsr GUI?

As you can see I'm a buit fuzzy on how PerfAdvisor counters flow thru the whole system. Thanks in advance.

Re: cache hit reporting

The performance impact depends on the number of additional counters and the frequency of collection. If the counters you are looking for are the cache counters listed above, then collecting them at a frequency of 1 min or greater should not have a huge impact on the amount of data or the performance.

To the general questions you asked below :

1. Turning on perfAdvisorEnabled does cause DFM to start collecting data for a default set of counters from all filers for which login credentials have been set. This is the data that is used for showing historical data in the default views.

Custom views that are created do not use this default data collection. Any custom view that is created triggers of additional data collection from the storage systems.

2. Turning on perfAdvisorShowDiagCounters does not start any new data collection. It just enables the Performance Advisor GUI to display the diag counters. The diag counters can then be added to custom views which would trigger data collection for for those counters.