This is extremely minor, but just something random I noticed and have been meaning to point out.
If Auto Task Timeout is enabled for a job, the slave prints out messages in its task logs as the task timeout is recalculated. Right now, those messages look like this:
Task timeout is 11065 seconds (Regular Task Timeout)
Task timeout is 11894 seconds (Regular Task Timeout)
Correct my if I’m wrong, but shouldn’t they say “Auto Task Timeout” instead?
Hello Nathan,
I’ll pass this by the devs, and if they think it’s a good fix, I am sure it won’t be too much to fix this. Thanks for pointing it out!
Cheers,
Hello Nathan,
Can you clear up for me which timeout reference you feel should be changed in your examples? Should we add 'Auto ’ to the beginning of the line, or change (Regular Task Timeout) to (Auto Task Timeout)?
Just to inform you, when the timeout is being calculated for the job, Deadline looks at the job’s timeout setting, as well as what the auto timeout setting would be, and prints out the smaller of the two. The regular timeout is the job’s timeout property, and if it uses the auto task timeout, it says so in the log, however if both are enabled, we’re only printing out the smaller of the two and clearly says the one it’s using. Look forward to your response.
Hey Dwight, I’ll try to clarify the source of my confusion.
I’m looking at a job with a “regular” task time limit of 5 hours (18000 seconds) that also has Auto Task Timeout enabled. One of its tasks got stuck and timed out based on the auto timeout calculation, and in the report title, it says as much. However, in the log text, at the end, I can see Deadline recomputing the auto timeout value as more tasks finish, before it finally kills the task:
2015-03-26 20:22:49: 0: Task timeout is 9237 seconds (Regular Task Timeout)
2015-03-26 20:30:49: 0: Task timeout is 10353 seconds (Regular Task Timeout)
2015-03-26 20:38:49: 0: Task timeout is 11677 seconds (Regular Task Timeout)
2015-03-26 20:56:49: 0: Task timeout is 12139 seconds (Regular Task Timeout)
2015-03-26 22:43:11: 0: Task timed out -- canceling current task...
My question is: why do these lines say “Regular Task Timeout,” when this value is clearly the “Auto Task Timeout”?
Hello Nathan,
While we don’t think this is a recalculation happening there, we are going to take a look more deeply at the code to make sure we are representing things right.
OK, but I don’t think there is any question that the values being printed there are NOT the “Regular Taks Timeout” value, so those messages are at the very least misleading.
Hello Nathan,
So I have received confirmation from the devs, after they investigated, that this is actually a bug, just not the one we thought it might be, and which we had ruled out. It wasn’t a bug with the messages displayed in the log as I originally thought, but instead it was actually a bug with how the job’s timeout type is updated while the job is rendering. Hopefully we can see it fixed in 7.1.