2010-10-19 08:23 AM
One of our customers has a fabric metrocluster and has been having some odd issues with it. Most now seem to have been resolved with firmware upgrades but they wanted me to check the configuration. I can't see anything wrong with the way in which it has ben cabled up, but they get a warning message every now and again relating to ports on most of the switches in the config.
config.SameSwitchWarn: Disks on port sw1:9 are also connected to the same switch at port sw1:9. For availability reasons, it should be connected to 2 separate switches.
When I run storage show disk -p it shows that all disks on both controllers have multiple paths, via different switches. Just wondering if anyone else had seen this message and knows whether it indicates something that should be configured differently.
Also, they have 2 shelves in a stack, but each of the shelves is owned by a different controller. I've suggested that one of the shelves should be removed from the stack and placed on its own loop, but this will change the disk name. E.g. from sw1:9.12 to sw1:10.12 when it is attached via a different switch port. I've suggested that the MetroCluster should be taken down in order to make this change due to the disk name changing to avoid the cluster panicing, but is what I've suggested OK? The system uses software disk assignment.
2010-10-20 02:31 AM
Sorry, I can't really help with the error message but removing shelves from a running system is never supported (MC or not, spare disks or not, doesn't matter), so while it *might* work, you're better off shutting down the cluster before moving the shelves.
We had some success with hot-removing shelves (involving pulling single loops, disabling ports and multiple takeover/giveback transistions) but I wouldn't generally recommend it (especially on an FMC)
2011-01-22 03:29 AM
I have seen the message today and the storage show disk -p also gave multipathed output.
We looked at the shelf that was connected to the port in the message. We noticed that one fc port in the esh module was blinking.
In the end we changed the sfp of the switchport that was connected to the blinking port. Now the port is online.