[16:11:02] <elukey>	 good news, it seems that we solved the bugs with s3/apus/ceph, the docker registry's restricted backend seems to work as expected (all the hanging etc.. was related to the radosgw s3 daemon in need of a restart on the apus hosts)
[16:19:27] <brouberol>	 nice!
[16:30:33] <elukey>	 I am a little concerned that all the problems were "solved" with a radosgw roll restart, since from what we know it can happen anytime
[16:30:47] <elukey>	 that is not really great, and there was no indication in the logs about what was happening
[16:31:12] <elukey>	 maybe something worth to discuss in the ceph sig
[16:43:14] <dpogorzelski>	 FYI: the ml-build1001 machine can't login to the registry if we only allow it push to the /v2/ml namespace since. when docker logs in /v2 is called so we might need to add the ml-build docker registry user to the `regular-push.htpasswd` file