Unsolved
This post is more than 5 years old
21 Posts
0
3842
Grafana Monitoring a Vplex cluster
Hello!
I am hoping to get some ideas as I'm a bit stuck, I'm trying to pull performance characteristics from our Vplex, we run the performance monitor tool but it does leave a lot to be desired. I'd like to create something similar to what we have done with Isilon ( insights connector -> influxdb -> grafana ) and collect stats and deliver them to Grafana.
Digging around I was able to create monitors, and sinks for the monitors. This gets me data for the stats I need (selected from the list I found with "monitor stat-list *") but the output drops it into folder on the management server. It stores these stats in a CSV.
I'm wondering if there is a way to instead grab the stats that go into the sink via the rest API, or if this can't be done maybe a way to retrieve the file via the api- I suppose I'd have to parse externally but csv is pretty friendly.
I suppose I can also scp with a scrip into the mgmt server and just download the file then parse but I'm wondering if there is a cleaner way.
Thank you!
SGieschen
1 Message
3
July 25th, 2017 07:00
Hi!
Have a look here:
https://vzaigrin.wordpress.com/2017/01/28/universal-storage-collector/
eespin
21 Posts
0
August 14th, 2017 17:00
I used his blog heavily in navigating the monitor setup- however due to the state of our environment (government) and restrictions imposed on me I wouldn't be able to use the collector tool. And in all honesty I am much more comfortable implementing something where I understand the build, functionality and can explain each call to security should an audit occur.
I'm actually going to implement this pretty simply, a cronned php script will scp the csv file created, pull the data, write to influx, compare to validate, delete the csv. Vplex proceeds to create a new csv the next period cycle and the process continues. It's not robust but it will work for our needs.
My only issue would be timing, should vplex still be writing to the csv and the script goes to collect and I loose good stat data. Still thinking on that one, maybe I'll move it to another location instead and run a recursive compare against what was stored in influx. Not sure...will update with ultimate solution.
garyo
89 Posts
0
September 14th, 2017 11:00
> I'm wondering if there is a way to instead grab the stats that go into the sink via the rest API, or if this can't be done maybe a way to retrieve the file via the api- I suppose I'd have to parse externally but csv is pretty friendly.
There is a relatively new command "monitor get-stats", which does allow you to get the most recent stats data for a monitor (aka. the last row written.) I think it was introduced in the 5.5.2 code timeframe. It was added as part of the VPLEX Performance Monitor product which polls the perpetual virtual-volume monitors. It can be called via a restful api call.
Additional info here (P. 387).
https://support.emc.com/docu80975_VPLEX_GeoSynchrony_6.0.1_CLI_Reference_Guide.pdf?language=en_US
The obvious downside to this is if you miss the time entry, it cannot be grabbed again - no going back if your poller goes down, or you have network issues, or if VPLEX management station goes down for whatever reason (well, stats are probably not being collected in that situation anyways...)
Pretty sure you can call it on any monitor type - the perpetuals or ones you've created.
Gary
eespin
21 Posts
0
September 26th, 2017 13:00
This is a great recommendation, I've just scraped up a little free team and should be able to rework the code this way. Thank you garyo!