Max Observed Reads per second

[ldclancy]

Our main cluster is backed by a SAN that we estimate to be capable of approximately 30,000 IOps. vRealize Operations Manager reports a Max Observed Reads per second of 2.2 million IOps and Writes per second of 1.2 million IOps.

Consequently the "Capacity Remaining" dashboard shows that we're only using 1.5% of the capacity of our storage. Not true!

Does anyone know why these statistics are so out of whack?

Can anyone advise how to fix or reset these statistics? Or worse case how to manually set them to a more realistic figure?

Thanks, Liam.

[mark_j]

I'll address this question in 3 parts:

1. ..whack?

First step, remove Disk I/O from the capacity remaining calculations. This isn't a good measure of capacity, and we always disable it by default. I'm pretty certain it IS disabled by default on new installations now.

Second, the stats are whacked because they're MAX OBSERVED as long as that vCenter adapter has been collecting stats from vCenter.. so it' essentially an upper watermark of all IO that's been observed by vR Ops. This stat is interesting to look at, but you really can't take it to the bank when it comes to 'absolutely' judging your storage's capabilities - storage MP will do that better, but even then not great (it's hard to calc true "max" IOPs unless you have intimate knowledge of the storage infrastructure).

2.. ..statistics?

You can't. Disable them in your capacity remaining policy settings and call it a day.

3 .. figure?

Can't do it. The closest you can get is building a supermetric based on a static MAX IOPS calculating and doing some type of aggregate/rollup type of deal to determine util%. Even then, your IO retries and whatnot may obscure your true throughput.