AWS Thinkbox Discussion Forums

Self-Directed Auto-shutdown?

Does it make sense for power management to be handled through pulse? Obviously you can’t bootleg startup and need a machine for that but I keep running into trouble with shutdown randomly not working.

I propose that slaves do their own auto-shutdown checks. They can update their power management settings when they update their pool/group data. Then they can turn themselves off and there are no random gaps. I haven’t had a problem with auto-start it’s always shutdown. I think it’s somehow related to when security patches are released. I think machines don’t want to auto-shutdown remotely when there is a pending windows update.

That’s an interesting proposal. The nice thing about having Pulse deal with it is that it’s easier to ensure that a certain number of slaves are left running, and it also allows one to override the shutdown order in the Idle Shutdown options. If the slaves did this themselves, there would have to be some coordination between the slaves to achieve this. Also, the Thermal Shutdown, Machine Restart, and Slave Scheduling features of Power Management are all based on the same remote communication system that Idle Shutdown uses, so if you’re having issues with Idle Shutdown, you would probably have the same problem with these other features if you ever wanted to use them.

I think it would be a better idea at this point to try and figure out why it occasionally doesn’t work for you. I know you’ve brought this up before, but unfortunately we’ve had no luck reproducing this problem. The first step would be to figure out where the problem is occurring.

When Pulse determines that a machine should be shutdown, it sends a remote command to the Launcher running on the remote machine, and then the Launcher relays that to the Slave. So the first thing we should check is if the Launcher is receiving this message from Pulse. You can check this in the Launcher’s log on the remote machine (just right-click on the Launcher and select Explore Log Folder). So when this happens again, send us the log and we’ll have a look.

If the Launcher isn’t receiving the message, that would indicate there is some sort of network issue. If it is receiving the message, then that indicates that it isn’t relaying it to the Slave properly.

Cheers,

  • Ryan

I’ve opened a new thread specifically to debug the current difficulty:
viewtopic.php?f=86&t=9070&p=38556#p38556

Privacy | Site terms | Cookie preferences