Better multipath scan

XenServer wont notice Multipath change when storage system is moving volumes between storage cluster members. In our case we added new storage nodes and clustered them with our current production storage and then moved volumes to new storage nodes and when migrating between old and new members storage system first gives access thru both storage members (old and new), so multipath count does double first, then it removes access from old storage nodes, and multipath count goes back to normal.

That scenario is not supported with XenServers currently.

I did some shell scripting that does allowed us to move SR volumes between different io groups. It does involves running rescan-scsi-bus.sh so running that gives new paths and then run multipath -r and multipathd -k"reconfigure" so it takes new paths to use. Then when storage system old nodes have their access removed from volume it does same scanning again.

Other related problem with max multipath count is that when multipath count in our case goes from 4 -> 8 and then 8 -> 4, XenServer mpathcount never wants to decrease max multipath count down to 4 again when it has seen 8 paths. So i made some shell scripting that does manually drops multipath count back to 4 again, would be nice to reset current max total paths as correct value from XenCenter GUI or from command line with one command.

My suggestion is that XenCenter GUI has this Rescan button under SR's Storage tab, could it force also rescanning scsi bus if multipathing is enabled so it would notice new paths and when move is done in storage system new rescan would remove old paths and old block devices from kernel ?

I attach my script that helped us to migrate SRs without problems. First you start this script with every XenServer hosts in same pool, then start migrating volume in storage system.