Cool feature of OnCommand Core - automatic clientstatistics

[hadrian]

Hello Team,

I'm trying to get a better handle on this cool feature that will automatically collect NAS client statistics based on performance events.

This feature is enabled at the CLI to AUTOMATICALLY gather (you can ad-hoc gather whenever you choose).  But what about the other options here.  Do these options decide when the automatic gathering starts, or do something else?

autoclientstatenabled

clientstatcifslatency

clientstatcputhreshold

clientstatmintotalopsrate

clientstatnfslatency

clientstatthresholdperiod

clientstattotalopsrate

[adaikkap]

Hi Hadrian,

Automatic collection of client statistics is based on the values of the following global options:

1. autoClientStatEnabled (default value No😞 specifies if automatic collection of client statistics is enabled.
2. clientStatThresholdPeriod (default value 300😞 specifies the period, in seconds, over which the following thresholds should be evaluated.
3. clientStatCifsLatency (default value 30😞 specifies the percentage increase in CIFS latency over the period.
4. clientStatCpuThreshold (default value 80😞 specifies the absolute, last available, value of the cpu_busy counter.
5. clientStatNfsLatency (default value 30😞 specifies the percentage increase in NFS latency over the period. This applies to both NFSv3 and NFSv4.
6. clientStatTotalOpsRate (default value 30😞 specifies the percentage increase in total operations over the period.

There are two sets of thresholds based on the above values:

1. If both the total_ops threshold and the CPU busy threshold are breached, automatic collection of client statistics will be initiated.
2. If the total_ops threshold and either of the NFS or CIFS latency thresholds are breached, then automatic collection of client statistics will be initiated.

Regards

adai

[MARTINLEGGATT]

Hi Adai,

In a similar way to automatic client statistics has Netapp considered adding a function to Oncommand where when a threshold is breached for one of the counters mentioned in your post it captures a perfstat for a short period?

For example if a Filers breaches the clientStatCifsLatency threshold a perfstat is captured for the Filer for a short time?

With intermittent performance issues Netapp Support often ask for a perfstat after the issue has occurred.

Regards

Martin

[adaikkap]

You can create an alarm when an event is generated, As part of the alarm you can kick in a script. This script can be the one which goes to the filer and collects perfstat. In the below link there is a how to on the same.

https://communities.netapp.com/people/fenton/blog/2011/02/22/diy-guide-to-integrating-perfstat-to-performance-advisor

Incase of client stats the following events are generated.

[root@ ~]# dfm eventtype list | grep -i clie

cifs-per-client-stats-disabled                Normal       cifs.perclientstats

cifs-per-client-stats-enabled                 Warning      cifs.perclientstats

nfs-per-client-stats-disabled                 Normal       nfs.perclientstats

nfs-per-client-stats-enabled                  Warning      nfs.perclientstats

[root@ ~]#

Regards

adai

[MARTINLEGGATT]

Hi Adaikkappan,

Thanks for the reply, that is exactly what I needed.

Regards

Martin