On Sep 4, 2007, at 9:22 AM, Jim Kellerman wrote:
On Tue, 2007-09-04 at 08:39 +0300, Enis Soztutar wrote:
AFAIK Hadoop-nightly is killed if it is not terminated in 45 minutes.
Actually, no it isn't anymore (I just updated the erroneous doc that
suggested that it was). The timeout feature of Hudson doesn't
reliably kill the underlying processes (see https://
hudson.dev.java.net/issues/show_bug.cgi?id=789 )
But for the hadoop-patch build, it is. We have seen builds running
for
hours, and worse now for days
(http://lucene.zones.apache.org:8080/hudson/job/Hadoop-Patch/
672/). Can
we also enable termination after some time for the hadoop-patch
build ?
The patch build makes two attempts to build and test the patch, so 45
minutes might not be enough. Maybe 90?
Hudson is supposed to kill long running tests, but often doesn't
manage
to do so. I have entered a Jira for this:
https://issues.apache.org/jira/browse/HADOOP-1831
Timing out a patch build won't have the effect you desire. It will
simply hang the patch queue since the 'current' link on the
filesystem to the patch being tested won't get removed. Probably a
Hudson plugin needs to be written to properly handle timeouts of the
patch build (and the plugin could also make the patch queue visible).
JUnit is *supposed* to timeout a test if it is taking longer than 15
minutes. This doesn't seem to work reliably if a test gets really
'wedged'.
Nige
