When we were adding the fedmsg hooks to pkgs01 and pkgs01.stg, we decided it
would be a nice "extra" to run "git update server-info" on each repo for each
push. This made cloning via http possible.

We ran a script "git check-perms /srv/git/rpms --check=post-update --fix" on
pkgs01 that added this hook for every repo, but I forgot to add it to
"setup_git_package" so that it would be added for every new repo. Now there
are a handful of repos that do not have the hook, while most others do.

Other than the inconsistency, this doesn't matter all too much. It is
annoying, however, due to a cron job that is checking for the new hook,
failing to find it, and bothering sysadmins about it over email.

I'm seeking +1s for a freeze break to make the following two changes:

1) Run "git check-perms /srv/git/rpms --check=post-update --fix" on pkgs01
again to fix the handful of repos that are out of sync.

2) Apply the following patch in puppet which will add the hook for new repos.

diff --git a/modules/gitolite/files/distgit/setup_git_package b/modules/gitolite/files/distgit/setup_git_package
index bd42b95..eeaa16f 100755
--- a/modules/gitolite/files/distgit/setup_git_package
+++ b/modules/gitolite/files/distgit/setup_git_package
@@ -121,6 +121,9 @@ ln -s /usr/share/git-core/post-receive-fedmsg
ln -s /usr/share/git-core/post-receive-chained
$GITROOT/$PACKAGE.git/hooks/post-receive

+# This executes "git update-server-info" on each push for clone via http
+ln -s /usr/share/git-core/templates/hooks/post-update.sample
+ $GITROOT/$PACKAGE.git/hooks/post-update

rm -rf $TMPDIR
echo "Done."
_______________________________________________
infrastructure mailing list
infrastructure@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/infrastructure

>>>>> "RB" == Ralph Bean <rbean@redhat.com> writes:

RB> 1) Run "git check-perms /srv/git/rpms --check=post-update --fix"
RB> on pkgs01 again to fix the handful of repos that are out of sync.

I wouldn't have realized that we were frozen in such a way as to prevent
this kind of thing. I'd +1 myself but I'm not sure if I have standing
to do so.

RB> 2) Apply the following patch in puppet which will add the hook for
RB> new repos.

I'd +1 this as well, but if there's controversy I could also modify the
SCM processing script to do this. It's not, to my knowledge, subject to
any freeze.

- J<
_______________________________________________
infrastructure mailing list
infrastructure@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/infrastructure

On Thu, Aug 16, 2012 at 09:54:46AM -0400, Ralph Bean wrote:
> When we were adding the fedmsg hooks to pkgs01 and pkgs01.stg, we decided it
> would be a nice "extra" to run "git update server-info" on each repo for each
> push. This made cloning via http possible.
>
> We ran a script "git check-perms /srv/git/rpms --check=post-update --fix" on
> pkgs01 that added this hook for every repo, but I forgot to add it to
> "setup_git_package" so that it would be added for every new repo. Now there
> are a handful of repos that do not have the hook, while most others do.
>
> Other than the inconsistency, this doesn't matter all too much. It is
> annoying, however, due to a cron job that is checking for the new hook,
> failing to find it, and bothering sysadmins about it over email.
>
> I'm seeking +1s for a freeze break to make the following two changes:
>
> 1) Run "git check-perms /srv/git/rpms --check=post-update --fix" on pkgs01
> again to fix the handful of repos that are out of sync.
>
Does this cause any end-user visible load on the server?

> 2) Apply the following patch in puppet which will add the hook for new repos.
>
+1

-Toshio
_______________________________________________
infrastructure mailing list
infrastructure@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/infrastructure

On Thu, Aug 16, 2012 at 07:32:39AM -0700, Toshio Kuratomi wrote:
> On Thu, Aug 16, 2012 at 09:54:46AM -0400, Ralph Bean wrote:
> > 1) Run "git check-perms /srv/git/rpms --check=post-update --fix" on pkgs01
> > again to fix the handful of repos that are out of sync.
> >
> Does this cause any end-user visible load on the server?

It takes a significant amount of time to run (a few hours). But it
does not appear to impact user performance:

1) We received no complaints last time we ran it....
2) I just tested cloning a package 3 times from pkgs.stg with and
without "git check-perms" running on the server. The times taken
to complete the operations were indistinguishable.
_______________________________________________
infrastructure mailing list
infrastructure@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/infrastructure

On Thu, Aug 16, 2012 at 10:45:56AM -0400, Ralph Bean wrote:
>
> On Thu, Aug 16, 2012 at 07:32:39AM -0700, Toshio Kuratomi wrote:
> > On Thu, Aug 16, 2012 at 09:54:46AM -0400, Ralph Bean wrote:
> > > 1) Run "git check-perms /srv/git/rpms --check=post-update --fix" on pkgs01
> > > again to fix the handful of repos that are out of sync.
> > >
> > Does this cause any end-user visible load on the server?
>
> It takes a significant amount of time to run (a few hours). But it
> does not appear to impact user performance:
>
> 1) We received no complaints last time we ran it....
> 2) I just tested cloning a package 3 times from pkgs.stg with and
> without "git check-perms" running on the server. The times taken
> to complete the operations were indistinguishable.
>
+1 to the freeze break request.

-Toshio
_______________________________________________
infrastructure mailing list
infrastructure@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/infrastructure

On Thu, 16 Aug 2012 09:54:46 -0400
Ralph Bean <rbean@redhat.com> wrote:
....
>
> I'm seeking +1s for a freeze break to make the following two changes:
>
> 1) Run "git check-perms /srv/git/rpms --check=post-update --fix" on
> pkgs01 again to fix the handful of repos that are out of sync.

+1

(That just needs to run on the list of newer ones? or over the entire
packages?)
> 2) Apply the following patch in puppet which will add the hook for
> new repos.

+1

kevin
_______________________________________________
infrastructure mailing list
infrastructure@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/infrastructure

On Thu, Aug 16, 2012 at 08:55:16AM -0600, Kevin Fenzi wrote:
> (That just needs to run on the list of newer ones? or over the entire
> packages?)

Yeah, I can run it on only the newer ones. That should lighten the
load.
_______________________________________________
infrastructure mailing list
infrastructure@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/infrastructure

On Thu, 16 Aug 2012 09:17:40 -0500
"Jason L Tibbitts III" <tibbs@math.uh.edu> wrote:

> >>>>> "RB" == Ralph Bean <rbean@redhat.com> writes:
>
> RB> 1) Run "git check-perms /srv/git/rpms --check=post-update --fix"
> RB> on pkgs01 again to fix the handful of repos that are out of sync.
>
> I wouldn't have realized that we were frozen in such a way as to
> prevent this kind of thing. I'd +1 myself but I'm not sure if I have
> standing to do so.

Yeah, we are in pre-alpha freeze.
See:
https://fedorahosted.org/fedora-infrastructure/browser/architecture/Environments.png
pkgs01 (where this change would need to happen) is under the freeze.

The rule for +1's for freeze break is sysadmin-main and
sysadmin-releng only. However, feedback is very welcome from anyone.
If you see a problem or have a concern, please do bring it up, or if
you think it's ok, thats good to to know too.

> RB> 2) Apply the following patch in puppet which will add the hook
> RB> for new repos.
>
> I'd +1 this as well, but if there's controversy I could also modify
> the SCM processing script to do this. It's not, to my knowledge,
> subject to any freeze.

Yeah, best to do all the git setup in the same place tho...

kevin
_______________________________________________
infrastructure mailing list
infrastructure@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/infrastructure

Ok, this should be all fixed up now.
_______________________________________________
infrastructure mailing list
infrastructure@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/infrastructure

>>>>> "KF" == Kevin Fenzi <kevin@scrye.com> writes:

KF> Yeah, we are in pre-alpha freeze. See:
KF> https://fedorahosted.org/fedora-infrastructure/browser/architecture/Environments.png
KF> pkgs01 (where this change would need to happen) is under the freeze.

But that's precisely my confusion. If we can't change a repository git
hook without a freeze break, then should we also not be processing SCM
requests? If someone asked me to delete a branch from a repository or
make some other SCM-admin-related change (which admittedly happens
rather rarely), is that also a freeze break?

- J<
_______________________________________________
infrastructure mailing list
infrastructure@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/infrastructure