harvest/grafana data points

[SamuelTBuckner]

Anybidy know how I can display (or even find) the 'node util %' counter in Harvest/Grafana?

[madden]

Hi,

The node util % shown in OPM is calculated within OPM and is not currently sent to graphite if using the 'external data provider' feature.  Because the calc methodology is within OPM, Harvest can't calculate it either so it can't send it to Graphite.

Cheers,
Chris Madden

Storage Architect, NetApp EMEA (and author of Harvest)

Blog: It all begins with data

[SamuelTBuckner]

First off, thank you so much for harvest and the grafana templates. Incredibly useful and fills a huge hole in presentable preformance data.

Any chance of harvest extracting that out of ocupm in the future? Also I've made a single page system performance dashboard that I really like. Should I post thst somewhere for others to use?

[madden]

Hi,

>>Any chance of harvest extracting that out of ocupm in the future?

Maybe.  If OPM would forward the metrics to graphite that could solve, or Harvest would need to use API Services to fetch it from OPM.  I haven't written integration with API services yet but if it this metric is exposed I could get it.

>>Also I've made a single page system performance dashboard that I really like. Should I post thst somewhere for others to use?

Sure, you can post in communities if you like.  Start a new conversation and title it approriately, include a screenshot, and attach your XML file.  If you use templates in that dashboard you might want to use the netapp-manager -export feature to save the dashboard to an XML file whcih removes the items from the template list..

Cheers,
Chris Madden

Storage Architect, NetApp EMEA (and author of Harvest)

Blog: It all begins with data

[prcpa8w3p]

Hi Chris - Thank you for presenting at the break-out session at Insight in Las Vegas.

Do you know how node utilization is currently calculated? Stetson Webster mentioned node utilization included things such as CPU utilization, disk busy, NVRAM card, and Kahuna Domain.

Can you please elaborate more in detail about this?

[jweihs]

Hello

Thanks for using OPM, and Grafana.

We have our engineering team working on including the node and aggregate utilization metrics in the OPM external provider feed. It will come in a future release. I will update this thread when we have committed date for that release.

As for sharing a dashboard, we will be launching a space for customers to share Grafana Dashboards ; this will be 'curated' - meaning that posting your dashboard to the community would be a first step, and when we see a dashboard that we feel is getting traction, we would then add it to that download area.

Best regards,

Yossi

[waynemcc]

Hi Chris,

Harvest can't calculate it or won't calculate it?  It seems to me that if we can use a plugin to manipulate the Kahuna domain per CPU into a utilization across all CPUs, then we should be able to use a plugin to calculate node utilization.  However, since this calculation seems to occur only within OCI and OPM, I have to assume that maybe NetApp is keeping that information proprietary?

Also, can harvest collect capacity metrics directly from the filers or does that only come via OCUM?

Thanks,

Wayne

[madden]

Hi Wayne,

The utilization numbers displayed by OPM and OCI are calculated in those products and I don't have the implementation details to re-implement them in Harvest.  Also, I expect as time passes they will be further optimized and extended presenting a challenge to keep the different calcs displaying the same numbers.  So rather than re-implement I'd rather consume them (API-Serivces v1.1 can pull from OPM) or get OPM to emit them as part of the "external data provider" feature.  For full control of the metrics API-Services seems to be the way to go but I can't say when or if I will have the time to add it to Harvest.  Maybe someone out there will take this as a challenge, do it, and share to all!

For capacity metrics I decided to get these from OCUM 6.x to avoid multiple polling of the systems.  It seems to work quite well so I have no plans right now to add logic and pull capacity info directly from cDOT.  Maybe you can explain your use case?  A consequence of fetching from OCUM 6.x is that there is no capacity info for 7-mode.  For 7-mode there was a request for IO density so I did collect vol used info (in MB) as a counter like: netapp.perf7.$Group.$Node.vol.$Volume.wv_fsinfo_blks_used

This same counter exists in cDOT as well, so you could add it if that woudl help.  There are also more counters in the volume object so you could see if others are interesting for your use case.

Hope this helps!

Cheers,
Chris Madden

Storage Architect, NetApp EMEA (and author of Harvest)

Blog: It all begins with data

P.S. Please select “Options” and then “Accept as Solution” if this response answered your question so that others will find it easily!

[waynemcc]

Chris,

If I can find some time in my schedule, I may have to take on that challenge.

In the meantime, let's say that I collect node utilization from OPM (as well as the ~115 counters OPM provides).  What if I want to collect others counters as well?  Your blog post OPM Vs. Harvest already shows that Harvest is the superior method for collecting stats (it collects more, is customizable, etc.).  If I collect only from OPM so I can get node utilization, then I lose out on all the other counters.  Or do I hit up the filers with both tools?  Seems "expensive" to do ("avoid multiple polling of the system").

I too want to do IO density reporting so that I can determine if I am meeting my SLA/SLO goals.  That's my reason for wanting to collect capacity stats as well.  I'll do the research to find out how to add the C-Mode counter to the collection stats, then I'll update (if necessary) the dashboards to use those stats.

Thanks!

[madden]

Hi Wayne,

Today node utilization is not sent via the 'external data provider' feature either so the only way to get it is to write your own logic to fetch via API-Services.  You can also get capacity info via API-Services, and it uses a web services restful interface (so typically easier to code vs. using the NetApp SDK), so that would be the path I suggest.

Cheers,
Chris