Hello,
I’m using power management to shutdown slaves, and when that happens, they tend to go to stalled state, only one machine closes cleanly with the right offline message. That machine is also a different hardware though. Thoughts?
Thanks!
Hello,
Can you go to one or two of those machines and send us the deadlinelauncher log so we can see if Launcher is getting the command to shutdown? Thanks.
Hello,
to @dwallbridge: here is a log of the deadline launcher
Thanks in advance!
deadlinelauncher-BOTTLE01-2014-02-10-0005.log (5.04 KB)
Hello,
Could you either post some sanitized deadlineslave logs, or email them to support so we can take a look, from the same time period? Thanks.
Hello,
here you can find attached two short logs of deadlineslave for this time-frame. (One before shutdown and another when the machine starts).
deadlineslave_BOTTLE01-BOTTLE01-2014-02-10-0008.log (722 Bytes)
deadlineslave_BOTTLE01-BOTTLE01-2014-02-10-0007.log (1.86 KB)
I’m not sure why, but it almost seems like the slave isn’t reporting it’s status back properly. Have to look into why that would be.
Would be great if you can figure something out. We’re having wakeonlan problems with the same machines too 
So, basically it seems what is happening is that the network connection on the slave is going down before it can let the repository know it’s shutting down. Usually if the slave comes back up, and starts responding, it should be marked as idle pretty quickly. Can you let me know what the average time it is listing between updates is? It might just me a time offset that, at the beginning, is too much, and the time before shut down is just enough to mark it as a slave. This is my best guess.
Interestingly enough, we stopped seeing that. I’ll write up again if it returns Thanks!
Hello,
Well that’s good. Let us know if it pops back up!