Grafana volumes by node

[AllenTR]

Hi,

Is it possiable to create a query, in Grafana, that would only display the volumes hosted by a node of the cluster? 

[yannb]

There is currently no way to do that.

You can, for a given node, go to "vol_summary", but that gives you an average for all the volumes in that node, which is probably not what you're looking for.

We started to think about options to give you an idea of what volumes are impacting a given node the most, but it's early talk so far, there are still loose end needed to tie.

another alternative would be to dynamically create sylbolic links into the whisper structure of graphite to establish aggregate/volume relationship, and regularly update it. It's kind of hacky so not really sure that's a good approach but it would probably work. More something you would find in NAbox than in Harvest itself I'm afraid, but it's being considered.

[StevePutre]

Good timing on this thread; I have been looking into the same issue.  

Often, we want to see which volumes are causing a given cluster node to have high workload, so that we can migrate high-workload volumes to another node.  But OnTAP only shows volumes by vserver, making it difficult to perform node-based analysis.

I came up with the workaround of creating the symlinks in the Graphite directory tree. I agree that its a hack, but it proved the concept that you can list volumes by node and build Grafana dashboards on top of them.

In my case, I developed a Powershell script which gets all the volumes from the cluster, then generates a ksh script file to remove old symlinks and create new ones.  I then have to manually copy the file to the Grafana appliance and run it.  Total hack.

I am starting to look into the NMSDK in to hopes of maybe developing a Perl interface which could run natively on the Grafana appliance.  

Wondering if this kind of functionality could be included in a future release of Harvest?

-Steve

[yannb]

That was considered some time ago, the main issue is that Harvest doesn't maintain de databases and path, it only talks to Graphite.

Maybe the best option would be to have a tool alongside Harvest distribution to put in a cron that would recreate aggregate/volume hierarchy. We just need to find someone to do it 🙂

[AllenTR]

Thank you all for you input.  I cannot believe how long ago I was pulled off this.   I guess I will look to this as a possible future enhancement.  In the mean time I like the SDK option.  I have been doing more with that lately as upgrades have taken center stage.