Project

General

Profile

Activity

From 02/03/2011 to 03/04/2011

03/04/2011

01:09 PM Tasks #861: handle_client_rename thinks nonexistent dir is in subtree map
Server.cc:4722 passes ... Greg Farnum
12:51 PM Tasks #861 (Resolved): handle_client_rename thinks nonexistent dir is in subtree map
... Greg Farnum
12:53 PM Tasks #862 (Resolved): cap_refs[CEPH_CAP_FILE_BUFFER] isn't cleared if truncation zaps changes
... Greg Farnum
12:43 PM Bug #858 (Resolved): cfuse craps out with fsstress
I have yet to see any cfuse issues with this. I will continue running and reopen if it does, but in the meantime I wi... Greg Farnum

03/03/2011

09:50 PM Bug #858: cfuse craps out with fsstress
1) Spinning the cpu. I have logs and a core file but haven't looked into them deeply. (one of my spare disks I use fo... Greg Farnum
09:26 PM Bug #858: cfuse craps out with fsstress
Greg Farnum wrote:
> What kind of issues are you expecting to crop up here? I've so far run across:
> 1) an issue w...
Sage Weil
05:28 PM Bug #858: cfuse craps out with fsstress
What kind of issues are you expecting to crop up here? I've so far run across:
1) an issue with cosd spinning that I...
Greg Farnum
09:32 PM Bug #854: unsynchronized clocks between kernel-client/cmds cause PJD fstest failures
The only reasonably sane idea I have here is for the client/mds to compare clocks to estimate skew and have some sort... Sage Weil

03/02/2011

10:01 PM Bug #858 (Resolved): cfuse craps out with fsstress
... Sage Weil
02:44 PM Bug #854: unsynchronized clocks between kernel-client/cmds cause PJD fstest failures
Ah, that makes sense. This is something we're unlikely to fix -- currently a lot of operations occur "on" the MDS (re... Greg Farnum
02:32 PM Bug #854 (Duplicate): unsynchronized clocks between kernel-client/cmds cause PJD fstest failures
I'm seeing a varied number (generally 5-8) of POSIX tests within the PJD fstest suite failing when the tests are bein... Brian Chrisman

02/18/2011

03:40 PM Cleanup #814: hadoop: refactor hadoop shim in terms of java libceph bindings
Sage,
I see the general benefits of a Java-based libceph interface, but what are your long-term plans for Hadoop o...
Noah Watkins
12:54 PM Feature #818 (Resolved): mds: robust lookuphash
Make the lookuphash be thorough about locating the directory inode. Sage Weil

02/17/2011

11:10 AM Cleanup #814: hadoop: refactor hadoop shim in terms of java libceph bindings
I mean that, at the end of the day, we should probably have:
- a java Ceph interface binding that is identical to t...
Sage Weil
09:00 AM Cleanup #814: hadoop: refactor hadoop shim in terms of java libceph bindings
I'm not sure what you're after here -- you mean you want Java bindings for librados, and then the Hadoop patches shou... Greg Farnum
07:14 AM Cleanup #814 (Resolved): hadoop: refactor hadoop shim in terms of java libceph bindings
Refactor the hadoop code in terms of generic Java bindings for libceph, instead of mixing the SWIG crap in with the a... Sage Weil

02/16/2011

08:16 PM Feature #601: mds: order directory commits after rename
Err, I guess actually that should read:
8) a commits
9) hold on b/bar gets removed
10) commit b comes in
11) b co...
Greg Farnum
07:53 PM Feature #601: mds: order directory commits after rename
I haven't gotten too far into it yet, but my current line of attack is to see how feasible it is to place "holds" on ... Greg Farnum
05:07 PM Feature #601: mds: order directory commits after rename
I think this is going to be much ahrder than I originally imagined. For example:
- mv /a/foo /b/foo
- mv /b/bar ...
Sage Weil
03:51 PM Feature #601: mds: order directory commits after rename
I'm going to take a look at this while waiting for data collection to run as I work on #698. Greg Farnum

02/15/2011

09:38 AM Bug #805 (Can't reproduce): mds startup: _replay journaler got error -22, aborting
Well, I can't figure out how the header could have gotten corrupted like that, but I've put in a few more checks for ... Greg Farnum

02/14/2011

04:26 PM Bug #805: mds startup: _replay journaler got error -22, aborting
Okay, somehow expire < trim. That shouldn't happen. Greg Farnum
02:29 PM Bug #805: mds startup: _replay journaler got error -22, aborting
Dumped and bzipped the journal here: http://johnleach.co.uk/downloads/ceph/805-journal.dump.tbz2 John Leach
01:21 PM Bug #805: mds startup: _replay journaler got error -22, aborting
You don't have any other logs from before, do you? At this point there's a problem with the way the MDS journal ended... Greg Farnum
01:02 PM Bug #805 (Resolved): mds startup: _replay journaler got error -22, aborting
As per #803, the assert is now fixed but starting up the cluster I now get:... John Leach

02/11/2011

04:04 PM Tasks #797 (Resolved): Don't _commit_full just because dir is_complete()
This got pushed in commit: fa4a9230c6ba68b9b66d1560abb89114caedf74b Greg Farnum

02/10/2011

11:57 AM Tasks #797 (Resolved): Don't _commit_full just because dir is_complete()
We should use some kind of metric to see if we should commit_partial or commit_full! Greg Farnum

02/08/2011

01:52 PM Bug #791 (Resolved): ls -al waits for writes to complete
On Mon, 6 Dec 2010, Jim Schutt wrote:
> Hi Sage,
>
> On Sat, 2010-12-04 at 21:59 -0700, Sage Weil wrote:
> > >
>...
Greg Farnum

02/03/2011

03:13 PM Feature #764 (Rejected): mds: make anchor table scale
The anchor table is current kept completely in memory. This won't scale forever for large numbers of anchors, especi... Sage Weil
 

Also available in: Atom