Unsolved
This post is more than 5 years old
4 Posts
1
769
Resiliency testing the VPLEX
Hi there,
We've a VPLEX Metro cluster here and the project wants to do some resiliency testing of the setup, to prove that if various things on the VPLEX fail the hosts see nothing bad.
So far the simple things like failed networks, failed FC connections have been tested out, but they want to test what happens when a director fails, is there a way to simulate a director failure ?
They want some tests that prove the VPLEX itself handles sudden hardware failures seamlessly to the host.
Thanks!
Mark
hernster
1 Rookie
1 Rookie
•
63 Posts
0
June 1st, 2016 06:00
The best test would be to perform an actual NDU. We have using VPLEX for about 2 years, this was the only way we could test director outage.
During an NDU, Directors A are upgraded first, so they go offline, come back online, and then a few minutes later, Directors B go offline while being upgraded. Hosts lose paths, but of course not all of them. We have also asked if there was any way to pause NDU after Directors A came back - so we could validate hosts re-gained all paths after first half og upgrade - but at this time the NDU cannot be paused, so it is all or nothing.
In our experience, various OS' and configs (with and without PowerPath), seem to handle path loss differently. Some are more sensitive than others. However, if your zoning is correct, and you are up to spec with the VPLEX Simple Support Matrix, you should be ok.
On the flipside, if anything is not properly configured on the host or zoning side of the house, i.e. per EMC VPLEX Best Practice, an NDU will quickly reveal any gaps. Meaning, yes, hosts will crash. However, the NDU pre-check will inform you if there are anomalies prior to proceeding with code upgrade.
Good luck