Community

Subscribe
Highlighted

Volume IOPs monitoring in NetApp 6210 and 3170

Hi,

I want to monitor my NetApp 6210 and 3170 SAN storage. I am pretty much done with everything. I am using open source third party NMS and doing SNMP based monitoring of my storage infrastructure. i am struck at below points:

1. Number of IOPs per volume.

2. Aggregate number of IOPs on all volumes.

3. Number of IOPs on each physical disk.

Prompt response in this regard will be highly appreciated.

Regards,

Riaz...

rizi.jbutt@gmail.com

Re: Volume IOPs monitoring in NetApp 6210 and 3170

You can use NetApp "stats" command which is available in the Data ONTAP on any filer.


You can find out the available measureable objects on the filer by entering:

step1:
toaster>stats list objects

As you are initerested in 'lun', 'volume' & 'aggr' related stats, you can see them listed among objects listed by the previous command.


To see 'counters' available for the objects listed in step1:

step2:
toaster>stats list counters volume


Counters for object name: volume
        instance_name
        node_name
        instance_uuid
        vserver_name
        vserver_uuid
        avg_latency
        total_ops
        read_data
        read_latency
        read_ops
        write_data
        write_latency
        write_ops
        other_latency
        other_ops


toaster>stats list counters lun
Counters for object name: lun
        instance_name
        node_name
        display_name
        read_ops
        write_ops
        other_ops
        read_data
        write_data
        queue_full
        avg_latency
        total_ops
        avg_read_latency
        avg_write_latency
        avg_other_latency
        queue_depth_lun

toaster> stats list counters aggregate
Counters for object name: aggregate
        instance_name
        node_name
        total_transfers
        user_reads
        user_writes
        cp_reads
        user_read_blocks
        user_write_blocks
        cp_read_blocks
        total_transfers_hdd
        user_reads_hdd
        user_writes_hdd
        cp_reads_hdd
        user_read_blocks_hdd
        user_write_blocks_hdd
        cp_read_blocks_hdd
        total_transfers_ssd
        user_reads_ssd
        user_writes_ssd
        cp_reads_ssd
        user_read_blocks_ssd
        user_write_blocks_ssd
        cp_read_blocks_ssd


You can go one step further and determine what instances are available for the object;  say ' volume'. The instances are basically the volumes available on that filer.


toaster>stats list instances volume
Instances for object name: volume
        vol0
        vol_QT_CIFS
        vol_show
        centos_iscsi
        cl_test_clone_centos_iscsi_20130204124118
        vol_dell
        centos_nfs
        vol_iscsi_win_test


Now, Start system statistics gathering in the background using '-I' as identifier. For example: To observe stats for volume 'vol_iscsi_win_test':

step3:
toaster> stats start -I volstats volume:vol_iscsi_win_test

Note:volstats is just the name I have given, you can give any meaningful name as you like.

To see the results while I/O is in progress:

toaster> stats show -I volstats
StatisticsID: volstats
volume:vol_iscsi_win_test:instance_name:vol_iscsi_win_test
volume:vol_iscsi_win_test:node_name:
volume:vol_iscsi_win_test:instance_uuid:368fa20c-6263-11e2-ad8d-123478563412
volume:vol_iscsi_win_test:vserver_name:
volume:vol_iscsi_win_test:vserver_uuid:
volume:vol_iscsi_win_test:avg_latency:4926032.41us
volume:vol_iscsi_win_test:total_ops:8/s
volume:vol_iscsi_win_test:read_data:0b/s
volume:vol_iscsi_win_test:read_latency:0us
volume:vol_iscsi_win_test:read_ops:0/s
volume:vol_iscsi_win_test:write_data:494556b/s
volume:vol_iscsi_win_test:write_latency:5479065.05us
volume:vol_iscsi_win_test:write_ops:7/s
volume:vol_iscsi_win_testSmiley Surprisedther_latency:8969.92us
volume:vol_iscsi_win_testSmiley Surprisedther_ops:0/s

Note:I am running these commands on my simulator, so dont read too much into the vaules here.


To stop the background stats and print the result on the console, enter:
toaster>stats stop -I volstats


Similarly for LUNS:

toaster> stats list instances lun
Instances for object name: lun
        /vol/vol_iscsi_win_test/iometer-BWKgW]BaIqNQ
        /vol/cl_test_clone_centos_iscsi_20130204124118/lun_centos_0-BWKgW]BaIqOk
        /vol/centos_iscsi/lun_centos_0-BWKgW]BaIqOA
        /vol/centos_iscsi/lun_centos_1-BWKgW]BaIqOC
        /vol/cl_test_clone_centos_iscsi_20130204124118/lun_centos_1-BWKgW]BaIqOl
        /vol/centos_iscsi/lun_centos-BWKgW]BaIqNE
        /vol/cl_test_clone_centos_iscsi_20130204124118/lun_centos-BWKgW]BaIqOj

toaster>stats start -I lun_stats lun:/vol/vol_iscsi_win_test/iometer-BWKgW]BaIqNQ
toaster>stats show -I lun_stats
toaster>stats stop -I lun_stats


There is alreay a wonerful blog post on this matter:

Performance "stats" without PerfStat or Ops Mgr:
https://communities.netapp.com/groups/chris-kranz-hardware-pro/blog/2009/04/01/performance-stats-without-perfstat-or-ops-mgr

Re: Volume IOPs monitoring in NetApp 6210 and 3170

Hi,

Thanks for your response.

What i want to do is. i want to have IOPs for each volume as separate and aggregate by using a third party NMS (Zabbix) via SNMP. I need to know the OID's and MIBs which i can download and can do the needful.

Regards,

Riaz..

Re: Volume IOPs monitoring in NetApp 6210 and 3170

IOPS, and latency, etc are not available via snmp, only via the Ontap API.

Roll your own, or go with something that has this built in and automated. (e.g. http://www.logicmonitor.com/monitoring/storage/netapp-filers/)

Re: Volume IOPs monitoring in NetApp 6210 and 3170

Hi Steve,

I can have IOPs per protocol level i.e. Number of IOPs on NFS, iSCSI and FC but i wanna have IOPs per volume level. I think if i can have at protocol level then there must be some way out for volume level as well.

Re: Volume IOPs monitoring in NetApp 6210 and 3170

You would think. But unfortunately it's not true. :-)

Having been through the NetApp MIBs comprehensivley several times, I can state categorically there is no way to get per volume IOps or latency.

Stuff like this:

can only be gotten via the API.

(The above is from logicmonitor).