System Reporter and Alert Rules

[starzero85]

Hey everyone

New 7200 3Par owner here trying to get my reporting and alerting setup correctly.

1.) Does anyone have a document on Alerting best practices? I 've seen a few examples on this board as well as in the help document, but I'm not sure which apply to my configuration. How should I determine what my alert limits are?
2.) A few alerts that I've setup so far are great until our backups kick off and total iops and svctime go through the roof, should I be using AO to migrate my data during backups? Or, is there an exception time frame I can configure to ignore the backup window? Or, is this indicative of an incorrect config that may be impacting performance?

thanks!

[Richard Siemers]

I would recommend that you try to stagger your backup and AO windows so that they do not step on each other, unless you have a heavy surplus of horsepower to slug through that simultaneous workload (sounds like you don't).

You may want to double-check that your AO measurement interval does not include backup/AO windows as well.

For a good start, check out this PDF posted by Afidel:
https://docs.google.com/file/d/0B-0rqB4 ... c1YVE/edit

It's not official customer facing HP documentation, however it appears to be what HP professional services use when they manage a 3PAR for their customers.

Some that I have been using:
15k drivess > 180 iops (count 3)
7.2K drives > 80 iops (count 3)
ports > X kbps where x= 80% throughput
vlun > 40ms service time, count3, minimum_iops>50 (This is a new one I am still tuning)

[starzero85]

Thanks Richard

We will continue to exclude our backup window from AO. I guess my concern is the amount of read iops on our data that AO moved to NL drives is pushing past 80,000 kbps and 1600 iops while backups are happening. Perhaps QOS is the way to go for NL?

[Richard Siemers]

What is your AO policy set to? How many tiers and is it balanced, performance or cost?

[starzero85]

It's setup like this

Tier 0 SSD
Tier1 FC 10k
Tier 2 NL 7k

configured for cost

[Richard Siemers]

Cost is going to try to push as much down to NL as reasonably possible based on performance... you may want to consider balanced if you are concerned with IOPs on the NL.

[hdtvguy]

Thanks Richard

We will continue to exclude our backup window from AO. I guess my concern is the amount of read iops on our data that AO moved to NL drives is pushing past 80,000 kbps and 1600 iops while backups are happening. Perhaps QOS is the way to go for NL?

We had this problem all the time. And it was virtually impossible to have a solution that did not slam the NL drives. Our AO would look at 7am to 7pm then run through the night, but while it was moving data our backups were running. Our daytime IOPS would be 40K, yet during the night we would run 60-80K sustained for hours on end. HP says AO by nature should exclude backup type operations (massive sequential reads), but I do not buy that. Our solution was to get away from traditional backups. We replicate all our data to our DR site so we only used traditional backups for SQL, Oracle and data file shares. Th rest of the systems (over 700 VMs and 40+ TB) we just snap every day and expire the snaps after 14 days. It took a hug load off the array. We also snap the target array as well so we have 14 days of snaps on each side.