Maintenance mode migrations - configurable?

[lldmka]

Was wondering if there are any configurable options for host maintenance mode - for instance, is it possible to control how many VMs migrate at once?

[kjb007]

I haven't seen a setting such as this per se. Anytime I've used maintenance mode, I haven't seen more than about 5 concurrent migrations at once. There were several queued, but only about 5 or so active. Are you having an issue with a large number?

-KjB

[lldmka]

I have some suspicions that SCSI timeouts seen in the past may have been caused by too many migrations occurring at the same time - with hosts 'fighting' for control of LUNs. I would like to be able to limit the number of simultaneous migrations to 2.

[kjb007]

What kind of shared storage are you using? Is it FC or iSCSI? How many hosts and vm's are you running, and how many vm's do you have per LUN.

I have ~8-10 hosts per cluster, and limit my vm's to about 15-20 vm's per LUN. I have yet to run into SCSI timeout issues, even in extreme cases where I've put a host with about 35 vm's into maintenance mode. As posted earlier, I had about 5 active migration tasks, and the remaining would queue until they were able to migrate. The entire process took about 10-20 minutes to complete and the host to be in maintenance mode, but all completed successfully. I've done this during patch cycles when I had less hosts in my cluster (5), and have had to do this several times to have all hosts patched.

I use FC, with 2 HBA's, and several array ports, for 8 total paths.

-KjB

[lldmka]

We use FC storage, 2 HBA's and 8 total paths. I limit our clusters to 8 hosts, while a typical LUN has 15 VMs.

We used to see excessive SCSI reservations and timeouts in the logs at about the same time I put hosts into maintenance mode. We are also running SAN replication to our DR site, which may be causing some of the errors, but for peace of mind I was hoping to limit active migrations.

As part of a 3.5 upgrade I have been reconfiguring HBA settings, as per advice from IBM - we use an IBM storage virtualisation device. This might optimise our storage response and resolve the errors.

[josha01]

I also have the same question and have not yet been able to find an answer so giving your thread a bump. I would like to reduce the amount of VMs migrated at once to ensure that there is no packet loss or IO contention during migration. I have not been able to find any advanced configuration setting that appears to correspond with this behavior.

vSphere 5.5 Enterprise Plus environment