[OpenAFS-devel] meeting minutes 2014-01-29

[Stephan Wiesand]

Log: =
http://conference.openafs.org/release-team@conference.openafs.org/2014-01-=
29.txt

Participants:

* Andrew Deason
* Derrick Brashear
* Marc Dionne
* Mike Meffie
* Stephan Wiesand

The meeting was all about 1.6.7.

=3D=3D Queuing changes in gerrit =3D=3D

Stephan would like to avoid the long gerrit backlog and resulting need =
for many rebases we had for 1.6.6. Before pulling up a new change, =
please check whether it touches the same files as others already pushed =
to the 1_6_x branch. In such cases, it's more helpful to help getting =
the older changes reviewed and thus either merged or fixed or abandoned =
than to pull up more changes of which it's already clear that they'll =
have to be rebased later.

We already have a few such cases.

=3D=3D Proposed items for 1.6.7 =3D=3D

=3D=3D=3D Old items revisited =3D=3D=3D

Gerrit 2984 (not yet pulled up) + 6266 implement a feature allowing to =
interrupt clients accessing volumes the fileserver is trying to take =
offline. This was discussed several times since 1.6.2 and not considered =
quite ready, but now the feature has seen real life use and Andrew is =
more comfortable with it.

Another feature under discussion since 1.6.2 is making the client give =
up callbacks on shutdown. The benefit is obvious, and the Windows client =
has been doing this for years, but it's known to kill servers older than =
1.4.6 and there are objections to adding the feature in the stable =
series without a switch allowing to disable it. Andrew would like the =
switch to be per cell (maybe extending "fs setcell") rather than global, =
and this isn't likely to be implemented anytime soon.

=3D=3D=3D New or not even yet on master =3D=3D=3D

Marc is working on a change which may help the getcwd ENOENT problem =
(see RT #131780 and references therein). Should be in gerrit soon. =
[update: http://gerrit.openafs.org/10774 ]

Andrew implemented gerrit 10773 which should help the "no buffer space =
available" problem (see =
https://lists.openafs.org/pipermail/openafs-info/2013-July/039706.html =
and =
https://lists.openafs.org/pipermail/openafs-info/2013-October/040178.html =
).

Mike has been working on parallel build fixes (see =
http://gerrit.openafs.org/#q,project:openafs+branch:master+topic:parallel-=
make,n,z ).

Andrew proposes these:

10214 namei: Ignore misplaced files (approved but not merged)
10322 salvager: Handle multiple/inconsistent linktables

=3D=3D=3D Accepted on master, to be pulled up =3D=3D=3D

Andrew proposes:

10634 afs: Treat vc_error as a CheckCode-translated code

=3D=3D=3D Pulled up to 1_6_x, to be reviewed =3D=3D

Ben's changes to allow using the OS X buildbot slave for 1_6_x too:
=
http://gerrit.openafs.org/#q,project:openafs+branch:openafs-stable-1_6_x+t=
opic:mavericks,n,z
Many were merged, but some need more discussion or maybe even more work.

Miscellaneous other changes:
10772 vol: Nuke parent vol special inodes
10771 salvager: Ignore linktable-only RW volumes
10770 namei: Set inconsistent linktable linkCount to 0
10769 salvager: Fix in-memory invalid linktable counts
10759 viced: Avoid issuing redundant TMAY requests
10758 viced: Move host quota calculation
10757 viced: Improve client error log messages
10756 viced: Improve CallPreamble error messages
10755 viced: Avoid endless BCB loop
10603 cellconfig: Do not use 'long' for dbserver IPs
10598 LINUX: Use sock_create_kern where available


--=20
Stephan Wiesand
DESY - DV -
Platanenallee 6
15738 Zeuthen, Germany