Nimble CS215 & Cisco UCS

[erikr]

Hello.  I have a Cisco UCS connected to a Nimble CS215.  I also have a Cisco UCS Mini connected to another Nimble CS215.  Everytime i update the Nimble OS on either system, I receive the following warning on each ESXi host:

"Lost connectivity to the device...backing the boot filesystem...As a result, host configuration changes will not be saved to persistent storage."

I removed the eui number from the above warning.  It's always the boot device datastore that shows up in these warnings for each ESXi host.

I only see the warnings after upgrading the Nimble OS.  I can make the warning go away by putting the host into maintenance mode and bouncing it or restarting the management agents on the host.

I was just curious if others experience this issue.

Thank you.

[SprinkleJames]

Hi Erik,

This is a long-known issue which happens when booting from iSCSI, because iSCSI boot in ESXi does not multipath. I'm not aware of any way to prevent it, and the only fixes I know of are the ones you mention. This is one of the reasons I typically recommend other boot media options for ESXi instead of booting from iSCSI SAN.

[erikr]

Hi James,

Would this still be the case if the ESXi hosts are not rebooting?  They are up and then the Nimble fails over to the standby controller causing this issue.  I can simulate it by manually failing over a controller using the Nimble UI.

I also see a warning on each host that at least two datastores are needed for HA.  This goes away after several minutes.

It would appear that the hosts are losing connectivity to the datastores long enough for the ESXi hosts to generate warnings.

Nimble Connection manager is up to date and the suggested Nimble iSCSI timeouts have been setup on the hosts as well.

I don't really know what to test to try to narrow down why the failvoer is not seemless.

[SprinkleJames]

Erik,

Yes, just having the ESXi OS installed on and booting from an iSCSI LUN leads to this behavior. ESXi is not tolerant (for the boot device) of the brief pause in I/O that happens during a controller failover, and so you see the error referencing the boot filesystem.

As for the warning that at least two datastores are needed for HA, I haven't encountered that before during a controller failover. Are both your heartbeat datastores on the same storage system?

[erikr]

Good morning Jerry,

Yes, the ESXi hosts only have datastores available to them that are on the single Nimble CS215 array.

Erik