Community
Participate
Working Groups
This URL doesn't react -- blocking me since don't get access to the TCF repo: http://git.eclipse.org/c/tcf/org.eclipse.tcf.agent.git/
cGit is killing us again. I'll write up some rules today to prevent that.
*** Bug 501205 has been marked as a duplicate of this bug. ***
I've re-enabled the cGit cache. It has a tendency to break, so I added a cron to clear the cache 2x daily. cGit can't live without it anymore.
Why are we seeing this now? We did not have issues in the past.
To be clear, I think only cgit (git web access) was down. Gerrit was up, though it seemed slightly degraded (out of 10 push/pull, I saw 2 timeouts approximately).
(In reply to Martin Oberhuber from comment #5) > To be clear, I think only cgit (git web access) was down. No, that is not true. Pull via SSH also failed.
Search fuctionality on cgit is not working anymore When I searched for swt it genered the following redirection url https://git.eclipse.org/c/%20SomeCustomInjectedHeader:injected_by_wvs/org.eclipse.cdt.old.git/diff/old2/cdt-home/developer/featurespec/2.0/contentassist.html/bpmn2-modeler/org.eclipse.bpmn2-modeler.git/cbi/org.eclipse.cbi.maven.plugins.git/babel/server.git/actf/org.eclipse.actf.visualization.git/actf/org.eclipse.actf.ai.git/cdo/cdo.incubator.git/cdo/parent.git/b3/b3.git/cbi/org.eclipse.cbi.examples.git/chemclipse/org.eclipse.chemclipse.baselinedetec.git/chemclipse/org.eclipse.chemclipse.filtermfremover.git/?q=SWT
> Why are we seeing this now? We did not have issues in the past. We've seen it before, in the past months. When the Oomph installer cannot reach certain files on www.eclipse.org, it pings those files on cGit. It's just too much. We've re-enabled cGit cache. However, bug 453438 is still problematic.
(In reply to Denis Roy from comment #8) > > Why are we seeing this now? We did not have issues in the past. > > We've seen it before, in the past months. When the Oomph installer cannot > reach certain files on www.eclipse.org, it pings those files on cGit. It's > just too much. Can we stop this?
See bug 463967. It's resolved/fixed but I would suggest it is not.
(In reply to Denis Roy from comment #10) > See bug 463967. It's resolved/fixed but I would suggest it is not. Of course fixes are only fixed in the code with the fix. I would suggest that the problem last time was that www.eclipse.org returned 404 (incorrectly) and that it's currently returning 403 incorrectly/inappropriately. This new problem is not as severe a problem because the code that's out in the wild will use the locally cached version in this case, unlike the 404, which did consult gitc, but will no longer do that ever in new installations and new installers. But for first time users with older installations or installers that they don't update, the 403 will also lead to gitc usage.
Embarassingly, we had a commit in August to block java clients to specific data types, and this affected old oomph clients. Through the magic of an unintentional merge commit, that code made its way in, which triggered the entire thing. We've reverted the commit and we've also kept the cgit cache mechanism. This will make cgit much, much more robust when faced with massive amounts of requests.
Yes, I can confirm that the problem is gone! I found the gitc caching a bit annoying because without the caching the log would immediately show my commit: http://git.eclipse.org/c/oomph/org.eclipse.oomph.git/log Previously, with caching, I'd often have to wait for 10 minutes for a link to the commit...
