Monitoring performance of an HP EVA SAN

hapklaar · Post by **hapklaar** » Mon Nov 17, 2008 7:27 am

Has anyone been able to graph the performance of an HP EVA SAN, for example disk i/o per disk group, read-write performance, etc?

I believe this could be possible by using evaperf in a script, but I haven't yet been able to get this to work.

hapklaar · Post by **hapklaar** » Wed Nov 19, 2008 3:55 pm

Let me be a little bit more specific. Evaperf can output realtime performance stats of an EVA to a file. I have included some examples. I could run evaperf every 5 minutes and let cacti read from the output it generates. I'm however at a loss how to get this done.

It would be great if someone could help me get a few of the values like latencies and transfer rates in a graph.

koaps · Post by **koaps** » Thu Nov 20, 2008 3:27 pm

I'm currently working on some trending analysis for our IBM SAN's.

I have a perl script that uses dd to transfer 2GB of random data to our SAN and record the MB/s. It verifies the file after the copy with MD5 to look for corruption.

If you wanted to play with it, I can provide the script I'm using, it wouldn't be hard to change it to parse the EVA file instead of running dd.

It creates, updates, and graphs on it's own using the perl module RRDs, but I also have Cacti pointed at the rrd files it makes so I can zoom in on failure areas that behavior detection marks on the script generated graphs.

Only thing is, Cacti doesn't support holt-winters stuff in rrdtool, and while getting the confidence bounds graphed with Cacti is trival, there's no tick support making it so Cacti can't display failures on the graphs.

From that data you provided, it wouldn't be hard to parse it with a script you run every 5 minutes and grab whatever values you want to store into a rrd file. I personally like to use external scripts for collecting data and use Cacti to just display graphs, keeping the poller disabled.

It wouldn't be realtime in any way, you would have 5 minute intervals for updates, and if using trending, 45 minute windows to look for 35 minutes of out of bounds(considered a failure). Of course this can be tuned, but changing the windows for failure detect to something to small will cause a lot more failures to show up and depending on how you plan to react to failures, it might be an issue. I'm working on making detected failures trigger pages in our nagios system, so a lot of failures would be a problem.

My goal was to be able to detect major performance events, for instance a drive failure causing a hot spare to become active and the RAID to rebuild, this should be detectable via trending analysis but still need to test things in our lab to see if it's truly detectable.

There is some info out there on how to roll your own external php scripts for Cacti use, so if you wanted to go that route I might be able to find that stuff, but I know Perl much better and tend to stick with it as much as I can.

jbabcock · Post by **jbabcock** » Thu Mar 05, 2009 8:59 am

Just wondering if you could provide your script to me as I am looking to collect the same kind of data.

Let me know

Thanks

flukester · Post by **flukester** » Fri Mar 13, 2009 2:34 pm

Hello!

I am in the same situation as the Original Poster for monitoring HP EVA SANs. In my case, I was easily able to monitor the Brocade switches but not the EVA itself.

However, it seems that there is a windows service running on our SAN management station that exposes a lot of information through windows Performance Counters.

Using the snmptools2 developed by Erwan.l, I am currently attempting to match this and graph these metrics from Cacti on Linux. It looks like it is going to work.

Antoine

gatorfreak · Post by **gatorfreak** » Fri Mar 27, 2009 11:21 am

Any luck on that?

I've used the Erwan iptools to get the OID of the EVA performance counters but they all just return values of zero, even directly with Erwan's iptools app fetching the values.

All we need are the OIDs to get the graphs going.

boatrke1 · Post by **boatrke1** » Wed Apr 01, 2009 6:37 am

Do you have a EVA3K/5K or 4/6/8?

gatorfreak · Post by **gatorfreak** » Wed Apr 01, 2009 7:26 am

EVA5K here. I did end up getting Erwan's snmptool working so that I can get values via snmpget. For some odd reason I can only get one of the values into cacti so far.
I wrote a post on another thread along these lines here:
http://forums.cacti.net/viewtopic.php?t=25600&start=195
Toward the bottom of that page...but the whole thread is pretty relevant to this topic since it appears to be the way to get this done.

boatrke1 · Post by **boatrke1** » Wed Apr 01, 2009 7:33 am

I was using VB Scripts on windows querying the EVA Perf statistics on the SMA. You need EVAPerf installed in order to do this.

I'm currenly rebuilding to use the wmi client on linux. Hopefully I should be done by the end of the week.

boatrke1 · Post by **boatrke1** » Fri Apr 03, 2009 10:05 am

Here's what I have so far. It's built to monitor two EVAs though the SMA. I'm using the WMI from http://www.parkingdenied.com

The host port graph description will most likely need to be modified to match the host port names.

catchtime · Post by **catchtime** » Sun Apr 19, 2009 8:39 pm

Hi, boatrke1

Wonderful.
That is what i am looking for.
Could you please tell me how to use it?

Thanks

boatrke1 · Post by **boatrke1** » Mon Apr 20, 2009 8:47 am

Download and install the CactiWMI 0.0.5 from http://www.parkingdenied.com/. Sorry, the URL was incorrect in the previous post.

Once you have the WMI configured, import the host templates and point the device to your SMA.

Kevin

palli · Post by **palli** » Thu May 07, 2009 2:20 pm

HI Kevin,

Are you still working on these templates ?

kind regards,
Pall Sigurdsson

boatrke1 · Post by **boatrke1** » Thu May 07, 2009 2:22 pm

That's all I've done with them. Is there something you were looking for?

Kevin

palli · Post by **palli** » Thu May 07, 2009 3:06 pm

I am currently working on a generic data query that can create a graph for every virtualdisk in the eva.

Working on Latency, iops and throughput

Just checking if you were working on similar stuff.

Cacti

Monitoring performance of an HP EVA SAN

Monitoring performance of an HP EVA SAN

Perl Script

Any updates

Who is online