Rank: Newbie
Groups: Member
Joined: 11/23/2009
Posts: 5
Points: 21
Location: paris (France)
|
Hello
We've observed the following behaviour on a PBS installation:
A user can run 7 jobs of 32 cpus, but when running 8 , it says "connection refused".
he also can run 1 job of 128 cpus, but a second one fails.
If he uses a new username, no problem to launch a new job....
(using the same queue)
Do you know if there could be a variable limitation for a cpu number per user ?
(As 8*32=256 is ok, not 7*32=224 .... strange)
I couldn't find such a limit in the documentation.
Thx a lot for any help.
|
Rank: Administration
Groups: Altair
Joined: 2/11/2009
Posts: 30
Points: -29
Location: Troy, MI
|
Hello JohnArchive,
I am assuming that qsub is giving the "connection refused" message, can you provide the full error message and which version of PBS Professional you are using?
If qsub is having a connection refused error, that effectively means qsub cannot contact the PBS Server... maybe due to hostname resolution issues, server is down, ...
By default, the user(s) can submit unlimited number of jobs to the system, and PBS will simply queue them up. Unless you have a max_queuable (old versions of PBS) or max_queued attribute set. Then the user would get a different message from qsub.
Scott
|
Rank: Newbie
Groups: Member
Joined: 11/23/2009
Posts: 5
Points: 21
Location: paris (France)
|
Thanks for the answer.
I won't have the answer till next week now.
But you're right: if a variable had been set, I imagine that the message would be more significant than only "connection refused"....
On the other way , it's so strange that it always appears after this 256 cpus max...
Regards.
JohnArchive
|
