On Tue, Nov 20, 2012 at 21:17:21 +0000, Laszlo Boszormenyi (GCS) wrote:
> On Mon, 2012-11-19 at 11:07 +0100, Julien Cristau wrote:
> > On Mon, Nov 19, 2012 at 01:18:34 +0000, Laszlo Boszormenyi (GCS) wrote:
> > > Agree. That's an other thing upstream should fix. However I don't think
> > > that would happen soon, at least not for Wheezy. I'll ask about it.
> > > Until then this sleep may fixes the majority (maybe all) of the
> > > problems.
> > >
> > Why can't this be fixed in your init script if upstream won't fix it in
> > time?
> Touché! First I thought it's not possible. 'couchdb -d' sends a signal
> to the running process that it should stop. It returns immediately and
> doesn't wait until it completely ends. Then found 'couchdb -s' which
> display the status of the daemon. While it's not my initscript, I've
> changed that to wait until the status is running.
> Changes between the current Wheezy version and the planned 1.2.0-3
> upload is attached.
Thanks, I think that should be acceptable.
> In short, it fixes four RC bugs:
> - now properly create a couchdb owned rundir,
That looks better, I think.
> - wait for complete stop of the daemon, and this allows to:
> - purge the package properly,
> - restart the service without failing,
Doesn't seem crazy.
> - logrotate will properly own the rotated files.
>
OK, I guess. Though why is the dir owned by couchdb in the first place
instead of root?
> +--- couchdb-1.2.0.orig/etc/init/couchdb.tpl.in
> ++++ couchdb-1.2.0/etc/init/couchdb.tpl.in
> +@@ -102,6 +102,8 @@ stop_couchdb () {
> + # Stop the running Apache CouchDB process.
> +
> + run_command "$COUCHDB -d" > /dev/null
> ++ while [ $(couchdb -s 2>/dev/null | grep -c process) -eq 1 ]; \
> ++ do echo -n .; sleep 1; done;
> + }
> +
> + display_status () {
Slightly weird to use $COUCHDB everywhere except in this one place where
you write couchdb.
Cheers,
Julien
Attachment:
signature.asc
Description: Digital signature