After some continued testing, there were some differences when testing the Storage Provider failover timing. Here were the findings through the testing.
|
Test that was Performed |
Storage Provider Failover Timing |
|---|---|
|
The VASA service is stopped on the FlashArray controller for the active storage provider. |
20 to 30 seconds |
|
The FlashArray controller for the active storage provider is rebooted. |
2 to 3 minutes |
|
The management interface on the FlashArray controller for the active storage provider is disabled. |
7 to 8 minutes |
Essentially what we saw with the management interface being disabled was a "transient" transport error. What we saw with the controller reboot was an "unreachable" transport error. There appears to be different timeout values in vCenter and ESXi for transient transport errors vs unreachable transport errors. Which are both different when the service is down/connection is refused with a "Bad Gateway" error and the session is additionally marked as invalid.
Everpure is currently working with VMware to get some clarity on the timeout values for these events and what the expected storage provider failover timing is as well.
VMware has opened up an internal PR for this investigation, PR:2349738
We have also started to track open DCPN cases in JIRA, this one can be followed here: https://jira.purestorage.com/browse/CSG-88
At this time there is no need to share the PR with customers, but if needed, please reach out to Alex Carver.
The goal of this KB is to cover what happens when a Storage Provider (FlashArray VASA Service) is unreachable from vCenter or any ESXi hosts. As well as to explore what is expected in vCenter and what would be expected from the ESXi Host as well.