This post is the next in my series examining an “out of the ordinary” or less than well documented command for the excellent Xen Cloud Platform cloud computing system. Today’s command is xe vm-reset-powerstate, a VERY important command to know when things go wrong!
In normal operation, the XCP pool master maintains a database of the every virtual machine running on any slave host in an XCP cloud. The problem comes in when the pool master loses communication with the slave. This could either be from a pool master crash, and network problem, or a slave crash.
When communication is lost, the pool master and the slave operate in emergency mode. In this mode, as long as the orphaned slave stays running, all virtual machines on that slave stay running as well, except that it isn’t possible to make any changes to virtuals running on that slave (if an orphaned slave could be changed, it would get out of sync with the pool master, with bad results). In the same way, the pool master will not allow any changes to be made to the orphaned slave until communication is restored.
In the case of a crashed slave, we don’t want the pool master to expect the slave to still be running virtual machines when it reconnects. vm-reset-powerstate allows us to reset the status of the lost virtual machines on the pool master to reflect the fact that they’re no long running anywhere. vm-reset-powerstate accepts the usual virtual machine selectors (vm=name or uuid, power-state=) and requires the “force” parameter as an indicator that you’re REALLY sure you want to do this. The end result is that the selected virtual machines are now set to halted on the pool master and can now be restarted on the pool master or any connected slave.
Looking for green data center or cloud computing services without the hype? Call or email me or visit the SwiftWater Telecom web site today!