We have a customer on CDOT 8.2 that needs to know "top talkers" for CIFS. This was available in 7-mode, and I have heard that it will make it back into a future release of CDOT (can't currently get this from OPM). I have not had a chance to spin up a lab to test it out, but I'm wondering if I can get this data via Harvest/Graphite/Grafana.
In 8.2 you can view ops from the CLI; see KB1014391 for more. My experience is that cifs_ops doesn't increment, but total_ops does and nlm_ops does, so maybe you can use those.
In 8.3 you can view ops from the CLI using the normal statistics interface and the "clients" object. My experience is CIFS stats do increment.
In 8.2 there is no API interface to get the stats, so Harvest can't collect. In 8.3 there is, but so far I haven't had enough demand to add it, so not there today either. Someone can however add it themselves, or if enough people speak up I can add it in a future release.
Cheers, Chris Madden
Storage Architect, NetApp EMEA (and author of Harvest)
I think the top talkers graph would be invaluable. It'd be great if it was included in the next release.
Also another thing that could be included is all of the E/EF series data. Right now i have to use the API and some scripting to send all of the Eseries data to graphite. However if Harvest was confgured to pull that data and if the pre made dashboards were available, it''d reduce all of the script/dashboard maintenance work.
btw the harvest tool is great, far better than any existing solution going around.
A colleague wrote an E-Series collector and we are currently refactoring it into Harvest for release. So I fully expect it to be in a future version of Harvest but unfortunately don't have a date for you right now. We are using the SMI-S Server as our proxy to the E-Series, but the Web Services Server is also possible. So if you have an urgent need you could look at one of those to grab data and post to Graphite in the meantime.
The list of top files and top clients has changing instance names and is poorly suited for use with Graphite [where you would end up with potentially hundreds of thousands of .wsp files with just a few data points in them].
If you need top files or clients I would write a script and put this data into Elasticsearch or similar because these can easily handle indexing with instance names that are frequently changing.
Cheers, Chris Madden
Solution Architect - 3rd Platform - Systems Engineering NetApp EMEA (and author of Harvest)