Network and Storage Protocols

SAN switches maintenance


Hi all.


We have a dual SAN fabric for our ESX cluster, backed with a FAS3140 MC.

SAN switches are (NetApp provided) Brocade 300's.

There are 2 completely separate SAN fabrics, each ESX and each FAS being joined to each fabric.


What would be the right and clean way to shutdown a SAN switch for maintenance ?

Should I disable FC links on the ESX side first ?

Can I just halt the switch and let it go and notify nodes, or even just power it down ?

What's the minimal set of steps required for NOT losing any I/O ?




Discussion on the same topic on VMware Communities concludes that FC switches can just be powered down.


Any other opinion ?


You can just power off a switch and the systems will compensate as long as all hosts have active paths through both switches.

If you want to try and avoid the timeouts, re-send of IOs, you can go into the hosts and set the link of the switch that is staying up as the preferred for IO so it isn't using the path to the switch that is going down.

If you have to make changes to both switches, you should wait a few hours / day between reboots to ensure the path to the recently rebooted switch is healthy for all hosts before dropping the other side.