Bug#747649: ITP: ceph-dkms -- Ceph FS and RBD Linux kernel drivers (DKMS version)

2014-05-11 Thread Dmitry Smirnov
On Sat, 10 May 2014 22:46:35 Ben Hutchings wrote:
 I just dislike kernel bugs being addressed by 'use
 this out-of-tree package instead'.  Some projects with both in-tree and
 out-of-tree releases seem to encourage this rather than maintaining
 their in-tree code properly.  (But I'm not saying ceph is among those;
 I'm not familiar enough with it to make a judgement.)

I share your concerns and I think there is no such risk here. Upstream don't 
do standalone module development and I had to extract module(s) sources from 
kernel tree. As far as I'm aware all Ceph modules development is done in-tree.

Just in case I commited README.source file with note not suitable for 
stable.


 If that has been committed to an official repository for
 ceph then I think I can cherry-pick it for sid.  Please report a bug
 against the kernel to request this.

I'll consider this but I hope that won't be necessary if upstream make sure 
that fix is included to point release of Linux_3.14 in a timely manner.

Thank you.

-- 
Cheers,
 Dmitry Smirnov.


signature.asc
Description: This is a digitally signed message part.


Bug#747649: ITP: ceph-dkms -- Ceph FS and RBD Linux kernel drivers (DKMS version)

2014-05-10 Thread Dmitry Smirnov
Package: wnpp
Severity: wishlist
X-Debbugs-CC: debian-de...@lists.debian.org,ceph-maintain...@lists.ceph.com

   Package name: ceph-dkms
Version: 3.14+git20140429
Upstream Author: Sage Weil s...@newdream.net
URL: https://github.com/ceph/ceph-client
License: GPL-2, LGPL-2.1
Description: Ceph FS and RBD Linux kernel drivers (DKMS version)
 DKMS drivers for Ceph file system and RBD devices.
 .
 This package provides DKMS kernel modules for Linux Kernel 3.14+.
 .
 Ceph is a scalable distributed storage system; RBD is a block device
 striped across multiple distributed objects in RADOS, a reliable,
 autonomic distributed object storage cluster developed as part of the Ceph
 distributed storage system.

Packaging is committed to

http://anonscm.debian.org/gitweb/?p=pkg-ceph/ceph-dkms.git

Shipping newer Ceph modules [libceph,ceph,rbd] in DKMS form is helpful to 
backport new features (like support for latest tunables) and fixes as well as 
to use recent ceph-client on stable Linux kernel.


signature.asc
Description: This is a digitally signed message part.


Bug#747649: ITP: ceph-dkms -- Ceph FS and RBD Linux kernel drivers (DKMS version)

2014-05-10 Thread Ben Hutchings
On Sun, 2014-05-11 at 03:59 +1000, Dmitry Smirnov wrote:
 Package: wnpp
 Severity: wishlist
 X-Debbugs-CC: debian-de...@lists.debian.org,ceph-maintain...@lists.ceph.com
 
Package name: ceph-dkms
 Version: 3.14+git20140429
 Upstream Author: Sage Weil s...@newdream.net
 URL: https://github.com/ceph/ceph-client
 License: GPL-2, LGPL-2.1
 Description: Ceph FS and RBD Linux kernel drivers (DKMS version)
  DKMS drivers for Ceph file system and RBD devices.
  .
  This package provides DKMS kernel modules for Linux Kernel 3.14+.
  .
  Ceph is a scalable distributed storage system; RBD is a block device
  striped across multiple distributed objects in RADOS, a reliable,
  autonomic distributed object storage cluster developed as part of the Ceph
  distributed storage system.
 
 Packaging is committed to
 
 http://anonscm.debian.org/gitweb/?p=pkg-ceph/ceph-dkms.git
 
 Shipping newer Ceph modules [libceph,ceph,rbd] in DKMS form is helpful to 
 backport new features (like support for latest tunables) and fixes as well as 
 to use recent ceph-client on stable Linux kernel.

Please don't do this.  If there are specific ceph features and bug fixes
that should be backported, talk to the the kernel team.

Ben.

-- 
Ben Hutchings
Horngren's Observation:
   Among economists, the real world is often a special case.


signature.asc
Description: This is a digitally signed message part


Bug#747649: ITP: ceph-dkms -- Ceph FS and RBD Linux kernel drivers (DKMS version)

2014-05-10 Thread Dmitry Smirnov
On Sat, 10 May 2014 21:02:53 Ben Hutchings wrote:
 Please don't do this.

I had to do it for troubleshooting as well as for delivering bugfix and new 
features support. I agree is should be temporary thing but I see no harm in 
it. I think I'm not the only one who might need it. For example there is 
upstream bug report to package modules for RHEL separately:

http://tracker.ceph.com/issues/6986

In any case I'm planning to target it for experimental only.
I know it is not suitable for release without upstream support.

Why do you think we'd be better without ceph-dkms?

Would it be OK for you if I keep it in experimental (or in unstable with RC 
bug not suitable for release to prevent migration to testing)?


 If there are specific ceph features and bug fixes
 that should be backported, talk to the the kernel team.

Thank you. It might be a good idea to let you know about problems.
For instance I've been hit hard by the following bug (I/O errors on RBD 
device) that was just fixed by upstream (so I'll have to use my DKMS package 
until fix propagate to 3.14):

http://tracker.ceph.com/issues/8226

Patch is included to the above bug report.

Trust me, I'm not doing it from boredom and it will be pity to let the effort 
die in vain...

-- 
All the best,
 Dmitry Smirnov.

---

A man is his own easiest dupe, for what he wishes to be true he
generally believes to be true.
-- Demosthenes, Third Olynthiac, sct. 19 (349 BCE)


signature.asc
Description: This is a digitally signed message part.


Bug#747649: ITP: ceph-dkms -- Ceph FS and RBD Linux kernel drivers (DKMS version)

2014-05-10 Thread Ben Hutchings
On Sun, 2014-05-11 at 07:27 +1000, Dmitry Smirnov wrote:
 On Sat, 10 May 2014 21:02:53 Ben Hutchings wrote:
  Please don't do this.
 
 I had to do it for troubleshooting as well as for delivering bugfix and new 
 features support. I agree is should be temporary thing but I see no harm in 
 it. I think I'm not the only one who might need it. For example there is 
 upstream bug report to package modules for RHEL separately:
 
 http://tracker.ceph.com/issues/6986
 
 In any case I'm planning to target it for experimental only.
 I know it is not suitable for release without upstream support.
 
 Why do you think we'd be better without ceph-dkms?
 
 Would it be OK for you if I keep it in experimental (or in unstable with RC 
 bug not suitable for release to prevent migration to testing)?

I think that's fine.  I just dislike kernel bugs being addressed by 'use
this out-of-tree package instead'.  Some projects with both in-tree and
out-of-tree releases seem to encourage this rather than maintaining
their in-tree code properly.  (But I'm not saying ceph is among those;
I'm not familiar enough with it to make a judgement.)

  If there are specific ceph features and bug fixes
  that should be backported, talk to the the kernel team.
 
 Thank you. It might be a good idea to let you know about problems.
 For instance I've been hit hard by the following bug (I/O errors on RBD 
 device) that was just fixed by upstream (so I'll have to use my DKMS package 
 until fix propagate to 3.14):
 
 http://tracker.ceph.com/issues/8226
 
 Patch is included to the above bug report.
 
 Trust me, I'm not doing it from boredom and it will be pity to let the effort 
 die in vain...

Understood.  If that has been committed to an official repository for
ceph then I think I can cherry-pick it for sid.  Please report a bug
against the kernel to request this.

Ben.

-- 
Ben Hutchings
Horngren's Observation:
   Among economists, the real world is often a special case.


signature.asc
Description: This is a digitally signed message part