General

Profile

Jeff Layton's activity

From 02/13/2018 to 03/14/2018

03/14/2018

07:11 PM CephFS Bug #23291: client: add way to sync setattr operations to MDS
Ok, I have a draft patch for this now. What I don't have is a great way to test it.
Hmm...now that I look, we have...
Jeff Layton
07:07 PM Ceph Bug #23358: vstart.sh gives obscure error of dashboard dependencies missing
Ahh, there are some new package dependencies (at least in Fedora 27). Try running install-deps.sh again. I needed:
...
Jeff Layton
05:39 PM Ceph Bug #23358: vstart.sh gives obscure error of dashboard dependencies missing
Passing in MGR=0 seems to be a workaround for now. Jeff Layton

03/12/2018

08:20 PM CephFS Bug #23291: client: add way to sync setattr operations to MDS
ceph_ll_fsync would give us almost the semantics we need. The main problem is that we may not have the file open and ... Jeff Layton
02:34 PM CephFS Bug #23291: client: add way to sync setattr operations to MDS
The big question here is how to achieve this:
The kernel has an AT_STATX_FORCE_SYNC flag, and we could repurpose i...
Jeff Layton
02:11 PM CephFS Bug #23291: client: add way to sync setattr operations to MDS
No. I don't think that bug is related. That one is dealing with setxattr (setting extended attributes) and yes, the c... Jeff Layton
06:22 PM RADOS Bug #23324: delete type mismatch in CephContext teardown
It looks more to me like we're allocating an object of one type (Entry) and then casting it to another (Log)? Is ther... Jeff Layton
04:56 PM RADOS Bug #23324: delete type mismatch in CephContext teardown
Package in this case is:
librados2-13.0.1-2356.gf2b88f364515.fc27.x86_64
Jeff Layton
04:51 PM RADOS Bug #23324 (Resolved): delete type mismatch in CephContext teardown
I've been hunting some memory corruption in ganesha and ran across this. Seems unlikely to be the cause of the crashe... Jeff Layton
02:35 PM CephFS Bug #16739: Client::setxattr always sends setxattr request to MDS
...I think you mean CEPH_CAP_XATTR_EXCL (Xx) here? Though you might need Ax too if someone is updating the ACLs using... Jeff Layton

03/09/2018

04:19 PM CephFS Bug #23291 (Resolved): client: add way to sync setattr operations to MDS
Zheng pointed out that we could end up with a setattr request being cached in the libcephfs client after it has repli... Jeff Layton

03/05/2018

02:29 PM CephFS Bug #22802: libcephfs: allow setting default perms
PR is up:
https://github.com/ceph/ceph/pull/20719
Jeff Layton
02:15 PM Ceph Revision 80e2f32b (ceph): client: add new ceph_mount_perms_set function
Allow programs to craft a UserPerm structure to use as the default
perms on the mount. The UserPerm is copied, so the...
Jeff Layton

03/03/2018

01:32 PM Ceph Revision 91e5bfa2 (ceph): tests: add a test for doing a getattr while caps are recalled but befor...
Add a testcase to reproduce the deadlock I was seeing with delegations
in ganesha.
Tracker: http://tracker.ceph.com/...
Jeff Layton
01:32 PM Ceph Revision f3840ca3 (ceph): test: cap waits in delegation tests at 60s
This adds assertions to all of the looping wait functions that we will
not do more than 60000 loop iterations (~60s)....
Jeff Layton
01:29 PM Ceph Revision dacc7b60 (ceph): client: allow getattr, lookup, fstatx, and readdir to use implemented caps
caps_issued_mask is a function that we call to see if we have all of
the caps provided in a mask. Usually this ignore...
Jeff Layton

03/02/2018

12:06 PM RADOS Bug #23194 (Rejected): librados client is sending bad omap value just before program exits
Thanks Jason. You were absolutely right -- the omap get/put at exit is being driven by ganesha. I had missed that bef... Jeff Layton

03/01/2018

10:38 PM RADOS Bug #23194: librados client is sending bad omap value just before program exits
rados_kv_get does look hinky, but I don't think we're calling into it here. We're basically doing a rados_kv_put into... Jeff Layton
09:16 PM RADOS Bug #23194: librados client is sending bad omap value just before program exits
I do have the ability to collect client logs within the container, and can turn up debugging in there if it'll help. Jeff Layton
08:56 PM RADOS Bug #23194: librados client is sending bad omap value just before program exits
Ahh, the object name is 29 bytes in this case, so maybe there is some confusion about lengths down in the code that i... Jeff Layton
08:49 PM RADOS Bug #23194 (Rejected): librados client is sending bad omap value just before program exits
I've been tracking down a problem in nfs-ganesha where an omap value in an object ends up truncated. It doesn't alway... Jeff Layton
03:23 PM Ceph Bug #23187 (Rejected): librados: segfault when fetching omap vals
Ahh this was my bug. I was passing &io_ctx to rados_read_op_operate instead of io_ctx. Since rados_ioctx_t is an alia... Jeff Layton
03:06 PM Ceph Bug #23187: librados: segfault when fetching omap vals
Sightly updated reproducer. Doug F. suggested that it might be the clnt going out of scope in cluster_connect, but ch... Jeff Layton
11:59 AM Ceph Bug #23187 (Rejected): librados: segfault when fetching omap vals
Build the attached program with something like:... Jeff Layton

02/21/2018

01:24 PM CephFS Bug #23028: client: allow client to use caps that are revoked but not yet returned
PR is up for this:
https://github.com/ceph/ceph/pull/20502
...one patch needs to be squashed before merge, ...
Jeff Layton

02/20/2018

08:57 PM CephFS Bug #23028: client: allow client to use caps that are revoked but not yet returned
Ok, I think I have something that fixes it. Basically, I think we can allow getattr* and ll_getattr* to use "implemen... Jeff Layton
06:43 PM CephFS Bug #23028: client: allow client to use caps that are revoked but not yet returned
This testcase seems to reproduce it reliably. The trick was that I needed the client holding the delegation to open t... Jeff Layton
04:00 PM CephFS Bug #23028: client: allow client to use caps that are revoked but not yet returned
I spent most of yesterday trying to write a reproducer but so far I haven't been successful. I did add some debugging... Jeff Layton

02/19/2018

02:53 PM CephFS Bug #23028: client: allow client to use caps that are revoked but not yet returned
Yes. We don't want to livelock here, so we may want to restrict this access to those functions that complete the use ... Jeff Layton

02/16/2018

09:42 PM CephFS Bug #23028: client: allow client to use caps that are revoked but not yet returned
Patrick Donnelly wrote:
> Sounds reasonable Jeff. I think only a change to the MDS is required?
No, actually this...
Jeff Layton
07:36 PM CephFS Bug #23028 (Resolved): client: allow client to use caps that are revoked but not yet returned
I found a problem with ceph and delegations.
NFS client holds a read delegation on a file. That gets recalled by g...
Jeff Layton

02/15/2018

11:41 AM CephFS Bug #22802: libcephfs: allow setting default perms
Yes. I had started some patches to do that a while back. I'll plan to pick them back up sometime soon. Jeff Layton

02/14/2018

01:45 PM Ceph Bug #22995: OSD_DOWN during ffsb on knfs
Yuri Weinstein wrote:
> note that jobs passed on ovh ?!
> http://pulpito.ceph.com/yuriw-2018-02-13_18:11:12-knfs-lu...
Jeff Layton
01:21 PM Ceph Bug #22995: OSD_DOWN during ffsb on knfs
Around the time of the other failed heartbeat check, I see this:... Jeff Layton
01:09 PM Ceph Bug #22995: OSD_DOWN during ffsb on knfs
So these tests just download ffsb and iozone and run them -- I guess to just generate some I/O with particular patter... Jeff Layton
 

Also available in: Atom