| 2016-05-09 15:16:32 |
Dmitry Tantsur |
description |
Ironic nodes will be put under maintenance for some reason, for example, temporary BMC offline or unreachable.
Once nodes are put under maintenance, human has to take them out of the state manually, which is unacceptable for large deployment.
Proposal is, when run "_sync_power_states", do not exclude "maintenance" nodes, scan them at longer interval; if "do_sync_power_state" succeeds, flip the "maintenance" status.
This way "maintenance" nodes can get out of "maintenance" without human interference. |
Ironic nodes will be put under maintenance for some reason, for example, temporary BMC offline or unreachable.
Once nodes are put under maintenance, human has to take them out of the state manually, which is unacceptable for large deployment.
Proposal is, when run "_sync_power_states", do not exclude "maintenance" nodes, scan them at longer interval; if "do_sync_power_state" succeeds, flip the "maintenance" status.
This way "maintenance" nodes can get out of "maintenance" without human interference.
UPD from dtantsur: we need a way (probably new flag on a node) to distinguish between nodes that entered maintenance automatically and nodes that were moved into maintenance via the API. |
|
