Thanks in advance !
In case of only a link/path loss (i.e. not a complete device loss), the path(s) should be recover instantly after being available again (it may take a few seconds until the GUI is updated though). That's actually required for e.g. switch updates where reboots are required, or for switch replacements ...
A few weeks ago I migrated a vSphere environment with iSCSI storage to new physical switches, by moving the uplinks to the new switches (one-by-one), always ensuring the "lost" paths were available again before migrating the next uplink.
André
additional question - is there a time period - during which transition from "dead" to "alive" is still possible ?
It should automatically restore, but a rescan of the storage adapter(s) might help.
afalk time period during from dead to alive is not there.
This may depend on whether all paths to the LUN were lost and/or the LUN was in use when the loss occurred. Take a look at http://kb.vmware.com/kb/2014155 to see whether this applies in your case.
André
I should have added more details initially
1) we are considering the case where only "some" of the paths are lost. after a period of time, the "lost" paths are "restored" (reconnected in hardware)
2) iops were still going through "alive" paths during the disconnection of the "problem paths"
So the questions are -
1) should we expect the esxi to change path state from "dead" to "alive" after the missing/disconnected paths become available again ?
2) is there a time period after which such path state transitioin (from "dead" to "alive") is not possible ?
In case of only a link/path loss (i.e. not a complete device loss), the path(s) should be recover instantly after being available again (it may take a few seconds until the GUI is updated though). That's actually required for e.g. switch updates where reboots are required, or for switch replacements ...
A few weeks ago I migrated a vSphere environment with iSCSI storage to new physical switches, by moving the uplinks to the new switches (one-by-one), always ensuring the "lost" paths were available again before migrating the next uplink.
André
