The server was overloaded and had to be rebooted. It is up now. It may not
handle all of the slaves doing a svn checkout at the same time, but I'll keep
an eye on it. Making the slaves not throw away their working copies when this
happens would make my job easier.
From a backend hardware standpoint, I think using svn.webkit.org will give you
better results than git.webkit.org on average. Plus I'd have to make sure the
git syncing completed before sending the build event to the master, so it
complicates the svn hooks which potentially slows down or prevents commits
when something is wrong with git.webkit.org.
-Bill
On Dec 21, 2011, at 8:43 AM, Osztrogonac Csaba wrote:
Hi,
it seems something happened with svn.webkit.org, because many bots fail with
svn error again and again. Unfortunately buildbot tries to solve svn errors
with rm -rf and a new checkout. But a new checkout takes min. 4-5 hours. :(
Have you got any idea what happened with svn.webkit.org? If it is stabilized,
I can do a trick on the Qt bots. I'll stopp them, copy an up-to-date svn
working
copy to theirs storage and then restart.
I think we should find an automatic and better way in
the future to avoid similar problems. For example:
- hack buildbot source not to do rm -rf
- migrate buildslaves to git.webkit.org somehow
(It needs many hack to save svn revision number on the bots)
br,
Ossy
___
webkit-dev mailing list
webkit-dev@lists.webkit.org
http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev
