I launched a slave with the -nogui flag, and then eventually shut it down via deadlineslave -shutdown (from another shell). My running Monitor never updates to reflect this, but continues to list the slave as “Idle”. This occurs whether or not the slave was running on the same machine as the monitor
Running a Pulse instance does not cause any kind of update, even when left running for several minutes. Eventually the monitor list will list the slave as Stalled.
It seems like the Slave process should have a cleanup handler that at least attempts to update the Mongo DB to indicate the slave is offline. I’m assuming the -shutdown flag causes the second slave process to call something like System.Diagnostics.Process.Kill on the headless slave process; perhaps an “external shutdown” event handler could be implemented instead, to allow for a less abrupt shutdown process.
The slave is supposed to be shutdown gracefully by running “deadlineslave -shutdown”, so there must be a bug here. Can you post the log from the slave session that was shutdown? We’ll see if we can reproduce here in the meantime.
The only log I have is the one from the terminal, which is pretty bare:
...
Cleaning up partially (un)archived jobs
Purging old statistics
Purging slave statistics that are older than Jun 17/12 11:31:50
Purging repository statistics that are older than Jun 17/12 11:31:50
Cleaning up orphaned tasks
Purging repository temp files
Purging trash
Purging old job auxiliary files
Purging old job reports
Purging trash
System.Net.Sockets.SocketOptionName 0x1b is not supported at IPv6 level
Killed
Same output as if I had done it myself using SIGKILL.
Thanks! I was just able to reproduce this here. A bug was preventing Deadline from determining the port number that the existing slave was listening on, so it fell back to killing the slave. It appears we have fixed this issue, so this should work properly in beta 3.