power on option available while system is deploying and already powered on

[Larry Michel]

I was looking at a host which apparently had failed to local boot after being deployed. As it was getting ready to timeout, I was looking at whether it was possible to power off the node and power it back to see whether it could retry the local boot.

That's when I noticed that only power on option is available in the UI (dropdown Take Action). The node was actually still powered on yet the UI allowed me to executed the power on anyway.. I am guessing that it didn't do anything and maas just ignored the error:

Aug 25 21:59:45 maas-trusty-aug-2015 maas.node: [INFO] sirrush: Status transition from READY to ALLOCATED
Aug 25 21:59:45 maas-trusty-aug-2015 maas.node: [INFO] sirrush allocated to user oil-slave-9
Aug 25 21:59:47 maas-trusty-aug-2015 maas.node: [INFO] sirrush: Status transition from ALLOCATED to DEPLOYING
Aug 25 21:59:48 maas-trusty-aug-2015 maas.power: [INFO] Changing power state (on) of node: sirrush (node-02f629dc-4b5b-11e5-91ed-00163e362f6f)
Aug 25 21:59:49 maas-trusty-aug-2015 maas.power: [INFO] Changed power state (on) of node: sirrush (node-02f629dc-4b5b-11e5-91ed-00163e362f6f)
Aug 25 22:37:43 maas-trusty-aug-2015 maas.power: [INFO] Changing power state (on) of node: sirrush (node-02f629dc-4b5b-11e5-91ed-00163e362f6f)
Aug 25 22:37:45 maas-trusty-aug-2015 maas.power: [INFO] Changed power state (on) of node: sirrush (node-02f629dc-4b5b-11e5-91ed-00163e362f6f)
Aug 25 22:39:48 maas-trusty-aug-2015 maas.node: [INFO] sirrush: Status transition from DEPLOYING to FAILED_DEPLOYMENT
Aug 25 22:39:48 maas-trusty-aug-2015 maas.node: [ERROR] sirrush: Marking node failed: Node operation 'Deploying' timed out after 0:40:00.
Aug 25 22:40:08 maas-trusty-aug-2015 maas.node_query: [INFO] sirrush: Power is on.

In any case, I don't think power on is right. Perhaps, power off (then when the system is powered off --> power on) or no power on/off until system is actually deployed?

This is with 1.8.2+bzr4041-0ubuntu1~trusty1.

[Gavin Panella (allenap)]

There is an inherent race in trying to prevent someone from powering-on an already-on node, or powering-off an already-off node. I think MAAS should not try to be clever here, and simply provide both choices all the time.

Having said that, it may be worth preventing or putting serious looking warnings around all power control when the node is in transitional states (commissioning, deploying) so that we lead people to use "abort $operation" instead.

[Newell Jensen (newell-jensen)]

Did we want to still and try and fix this or are we going to leave both power options all the time as Gavin mentions?