1509398: Remove the iSCSI rescan during disconnect
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Mirantis OpenStack |
Invalid
|
High
|
Alexey Stupnikov | ||
7.0.x |
Confirmed
|
High
|
Alexey Stupnikov |
Bug Description
This is a "sub-bug" of https:/
Doing an iSCSI rescan at disconnect time is pretty dangerous to do. iSCSI rescan doesn't make volumes go away, it makes them show back up. So during Nova live migration it will make local scsi volumes show back up.
Nova live migration first loops over volumes and calls disconnect_volume, then after that it loops over the volumes again and call's cinder to terminate_
If disconnect_volume does a rescan, then volumes there were removed during previous calls will show back up, because cinder's driver hasn't had a chance to terminate the connection yet from the array.
Original issue: https:/
Original code-review: https:/
Changed in mos: | |
milestone: | 7.0-mu-2 → 7.0-updates |
Changed in mos: | |
milestone: | 7.0-updates → 8.0 |
status: | Confirmed → New |
Changed in mos: | |
status: | New → Invalid |
Same issue as in bug #1521648. In Kilo Nova uses its own code to provision volumes. It is needed to patch nova/virt/ libvirt/ volume. py file to change the way it operates.