Top 10 IOPS in Performance Advisor

[JeanFrancois_Marie]

By default the dashboard presents

- top performance events

- top storage systems by network throughput

- top storage by total ops

- top storage by CPU utilization

How do I change that ?

I wish to get:

- top 10 qtree NFS ou CIFS IOPS

- top 10 NFS client IOPS

Thanks for hep

regards

JF Marie

[geddam]

You can do this, follow below steps.

1. Go to custom views, create a custom view with one bar chart for cifs_ops and another bar chart with nfs_ops ( select Global in available objects page. While creating a bar chart you can select the number of objects).

2. Save it

3. Go to view page and in the hierarchy tree click on global. From the views combo box select the new custom view and click on " Set Default view".

4. Now go to Monitors page, you can see the new custom view charts.

[harish]

Please follow these steps:

Create a new view with the following:

1. Bar chart with top Qtrees with NFS ops

2. Bar chart with top Qtrees with CIFS ops

I do not understand what you mean by "top 10 NFS client IOPS "

One you have created the view, open the view by selecting "View" Tab and

then "Global" in the tree.

Now click on "Set Default" button to make this view your dashboard.

Regards

Harish

[JeanFrancois_Marie]

Hello,

This part of your answer answered perfectly my questions.

What I miss is to follow the client part (Solaris or Linux client) of the NFS activity.

I used to monitor this by using the command:" nfsstat - l" or my filer.

May I get this using performance Advisor ?

I saw that monitoring qtree just allow to get ops but not latency for NFS and CIFS activity.

Do we have a mean to monitor QTREE NFS/CIFS latency ?

Thanks for help

Best regards

JF Marie

[harish]

For individual qtrees, we only have cifs_ops, nfs_ops and total_ops counters only. There are no latency counters at the qtree level.

However, with DFM 4.0, you can collect the client statistics at a filer level. This feature will report you the NFS/CIFS operations

performed by each of your linux/solaris/windows hosts. Please let me know if you need more information to collect client statistics.

[tatlee]

Hi Harish,

I also want to know how to collect the client statistics.

In addition, any other way to know the latency at qtree level?

Regards,

Terrence

[harish]

You can collect client statistics either using CLI or NMC. The cli is

"dfm perf clientstat", whose details are as follows:

$ dfm perf clientstat help

NAME

clientstat -- Manage collections of client statistics.

SYNOPSIS

dfm perf clientstat collect [ -p <collection-period> ]

<storage-system-name-or-id>

dfm perf clientstat list [ [ -b <before-time> ] [ -a <after-time> ]

<storage-system-name-or-id> ] | [ -s <stat-id> ]

dfm perf clientstat purge { [ -b <before-time> ] [ -a <after-time> ]

<storage-system-name-or-id> } | { -s <stat-id> }

DESCRIPTION

Manage collections of client statistics.

To collect client stats from NMC, goto "Monitor" -> "client statistics".

This page will list all the hosts.

Select the host for which you want client stats collected and click on

"Collect client statistics" button.

Collection of client statistics can also be done automatically by

setting autoClientStatEnabled to enabled.

Let me know if you are interested in this feature.

Regards

harish

[tatlee]

Harish,

Thanks a lot!

Regards,

Terrence Lee

NetApp

852.3605.7700 Main Line

852.9181.8824 Mobile Phone

terrence.lee@netapp.com

Learn more: http://www.netapp.com/in/future-ready

<http://www.netapp.com/in/future-ready/?REF_SOURCE=ftremsignaturein>

<http://www.netapp.com/in/future-ready/?REF_SOURCE=ftremsignaturein

[anantha_dommeti]

The charts are pretty useful given the fact of more counters in 4.0

Is there any way of getting Top 10 IOPs at a particular time rather than current instant ?...This would help us to co-relate what's running hot at a past incident.

Most of the time, we try to find the cause for a performance degradation for past event and we get to a point of exporting required counters on all volumes and use excel to chart it...I would rather like to see if we can do it from performance advisor.

[harish]

Hi Anantha,

Do you want to know the top 10 volumes (for instance) based on some

counter (average latency for instance) for an particular time period?

There is no way to currently do this. Please submit a RFE.

Regards

Harish