[Gluster-infra] Gerrit down, disk full and more complication

[Michael Scherer]

Le lundi 14 mars 2016 à 09:46 +0530, Kaushal M a écrit :
> On Mon, Mar 14, 2016 at 9:35 AM, Niels de Vos <ndevos at redhat.com> wrote:
> > On Sat, Mar 12, 2016 at 12:23:52PM +0100, Michael Scherer wrote:
> >> Le samedi 12 mars 2016 à 12:08 +0100, Michael Scherer a écrit :
> >> > Le samedi 12 mars 2016 à 10:11 +0100, Michael Scherer a écrit :
> >> > > Hi,
> >> > >
> >> > > so ndevos pinged me on irc this morning (around 7) since gerrit was
> >> > > down. I poked around it once I woke up, and found that the disk was
> >> > > full. However, after adding a new disk to the volume group, the web
> >> > > interface also refuse to start.
> >> > >
> >> > > I am currently looking for that, but not ETA until I find what is wrong.
> >> >
> >> > TLDR: So gerrit is back.
> >> >
> >> > Root cause was a wrong interaction between gerrit config and apache.
> >> >
> >> > For a reason I do not quite understand (since we restart gerrit quite
> >> > often), this:
> >> >         canonicalWebUrl = https://review.gluster.org/
> >> > stopped working after the reboot.
> >> >
> >> > instead, you have to have this:
> >> >         canonicalWebUrl = http://review.gluster.org/
> >> >
> >> > for a reason that I do not understand, this resulted in a redirection
> >> > loop with // url at the end.
> >> >
> >> > I tried various variations (such as removing the /, adding /foo,
> >> > or /foo/), but the only working way was to tell to gerrit to advertise
> >> > the http url.
> >> >
> >> > But, apache redirect to https fine, and do not complain at all, so maybe
> >> > that's a issue on the proxyPass side.
> >>
> >> So this is interesting, cause this work:
> >>    canonicalWebUrl = https://review.gluster.org/
> 
> This is my fault. Sorry.
> 
> I changed it sometime late last week (Thursday I think), so that the
> URL returned when pushing a change, and the URLs used in the
> commit-message when a change is merged used `https`.
> I thought that this setting only affected the UI, and didn't expect it
> to cause problems.

So it seems that the change I made also broke some stuff (ie, the github
integration).

Once the releases are out and jenkins moved, I might need to dig a bit
more there, cause I would have also expected to work like you said it
would work too. 

(also, one more reason to start using a cfg mgmt stuff, since I would
have been able to see what was changed and revert)

> >> if you have proxy-https in the ListenUrl setting.
> >>
> >> And it seems it made gerrit a bit faster (potentially due to lack of
> >> redirection), even if maybe that's just the restart.
> >
> > Interesting, thanks for looking into this!
> >
> > Niels
> >
> > _______________________________________________
> > Gluster-infra mailing list
> > Gluster-infra at gluster.org
> > http://www.gluster.org/mailman/listinfo/gluster-infra

-- 
Michael Scherer
Sysadmin, Community Infrastructure and Platform, OSAS


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part
URL: <http://www.gluster.org/pipermail/gluster-infra/attachments/20160314/8b52f805/attachment-0001.sig>