GSI Forum - RDF feed
https://forum.gsi.de/index.php
Dashboard not killing timeout tests
https://forum.gsi.de/index.phpindex.php?t=rview&goto=10634&th=2826#msg_10634
I noticed (now for a while) that the nighly dashboard run leaves (sometimes) a root instance running which eats 100% cpu for hours (=until I kill it by hand).
Has this been observed by others, too?
Björn]]>Bjoern Spruck2010-05-06T09:23:18-00:00Re: Dashboard not killing timeout tests
https://forum.gsi.de/index.phpindex.php?t=rview&goto=10635&th=2826#msg_10635
Actually I see the same. Each night one root application poltergeist.
Ralf.]]>Ralf Kliemt2010-05-06T09:34:26-00:00Re: Dashboard not killing timeout tests
https://forum.gsi.de/index.phpindex.php?t=rview&goto=10647&th=2826#msg_10647
I know this problem, which is somehow OS related. I see this problem only on one platform. To get rid of the zombies i put a
killall root.exe
in the last line of Dart.sh. This is no solution but at least a workaround.
Do you have some test which run into a timeout?
Ciao
Florian]]>Florian Uhlig2010-05-06T13:27:59-00:00Re: Dashboard not killing timeout tests
https://forum.gsi.de/index.phpindex.php?t=rview&goto=10648&th=2826#msg_10648
I'd be careful with that! You don't want to kill running jobs
Since you think it being platform related: my machine is leonardo (openSUSE_11.2-GNU_Linux-i686-gcc4.4-fairsoft_Jan10).
Kind regards, Ralf.]]>Ralf Kliemt2010-05-06T14:02:50-00:00Re: Dashboard not killing timeout tests
https://forum.gsi.de/index.phpindex.php?t=rview&goto=10649&th=2826#msg_10649
Since i start no root sessions on my test machines during this time of the night, there is no problem with killing all root jobs. If there is still a root job running this is also a zombie which could be killed.
Somehow on some machines/systems the root job is not terminated correctly. I had the problem only if one of the jobs runs into a timeout.
Ciao
Florian
]]>Florian Uhlig2010-05-06T14:46:04-00:00Re: Dashboard not killing timeout tests
https://forum.gsi.de/index.phpindex.php?t=rview&goto=10651&th=2826#msg_10651
yes, there was a timeout job, i guess its related to that.
I have the problem on different machines OS systems, but not everytime on all machines, but that changes.