harvest counters

[maskajan09]

Hello

I'm new to the harvest and I'm really lost. I use Nabox with harvest 1.6 right now. I would like to know which counters (mainly IOPS, latency and throughput) are related to the frontend and which are related to the backend (aggregate). Is there anywhere list of collected counters with some explanation? And also the list of commands used by harvest to collect the counters.

Thank you

Jan

[vachagan_gratian]

Hi Jan,

You can take a look at the template file, which lists the objects and counters that are collected by Harvest. E.g. open template/default/cdot-9.6.0.conf (or replace "9.6.0" with the version that most closely matches to your cDot).

Next, you can use the utility util/perf-counters-utility to get info of about the counters. E.g. to see the counters of object volume, run:

/util/perf-counters-utility -host HOST -user USER -pass PASS -f volume -c

(replace HOST, USER, PASS with the hostname, usename and password of your Cdot system).

You can also use this utility to explore objects / counters that are not collected by Harvest, add them to the template to make Harvest collect them.

Let me know if you need more help.

[maskajan09]

Hello Vachagan and thank you for your reply. I have seen this answer already on this forum and it is too general. I'll give you an example. First of all... I use nabox and it contains graphite. I can browse the graphite and I assume that the structure there is the same as I can get from the cli utility, right? So for example

..../node/node01/aggr/data_aggr1/vol_summary/total_ops. I would say that this is IOPS on the backend/aggregate level.

But then we have

 ..../node/node01/vol_summary/total_ops    -here I'm not sure. Is it frontend? Is it the same counter as for the aggr? The graphs put together are almost the same but there are differences. So it looks like to different sets of data.

What iops are combined in total_ops? I can see a lot of ops around - other, read, write, cifs, nfs, cifs_read, cifs_write, nfs3, nvf3_read, ........ endless list of ops scattered around all possible sub categories. Without detailed documentation it is just a mess.

And another confusion

there is a panel in nabox called Node Protocol Backend Drilldown and there is NFS IOPs graph that takes data from

..../node/*/system/nfs_ops .... so the system branch gathers the data from backend??

What I really miss is a document that would describe

..../node/node01/vol_summary/total_ops    is the counter for ?total iops on the frontend bla bla? and the command used by harvest to collect the data from netapp storage is <command>

Is there something like that? Because I think there must be something from the guys that wrote/write the harvest.

Thank you in advance for any help that can shed some light to this.

Jan

[maskajan09]

One another question. I need counter for frontend NFS and CIFS iops. If I assume that the frontend is under 

.../node/node01/vol_summary/  then there is only nfs_other_ops, nfs_read_ops and nfs_write_ops. No total nfs_ops. But this counter can be found under .../node/node01/system/nfs_ops but that should be backend/aggr ops. Soooo where can I find total nfs_ops for frontend??

Thank  you

[maskajan09]

Hello all

I'm back to this post after some time and I'm still in the same spot. Haven't found any documentation that would clarify this issue. Anything new? Anyone?

Help please!