Hide Forgot
Steps: 1) Cloned a glusterfs git repository onto the glusterfs mountpoint 2) did ./autogen.sh, configure and make Now doing rm -rf of the repository gives the following error. Its not always reproducible. But most of the times it happens. root@bigbang:/mnt/dir/glusterfs.git# cd .. root@bigbang:/mnt/dir# ls fstest_10292a823178f683f79fe700ef073f2e glusterfs.git root@bigbang:/mnt/dir# rm -rf * rm: cannot remove `glusterfs.git/.git/config': Invalid argument This is the entry in the log file corresponding to the error: [2011-02-25 17:12:11.600672] D [stat-prefetch.c:3587:sp_release] vol-stat-prefetch: cache hits: 0, cache miss: 1 [2011-02-25 17:12:11.603777] D [stat-prefetch.c:3587:sp_release] vol-stat-prefetch: cache hits: 1, cache miss: 0 [2011-02-25 17:12:11.611700] D [client3_1-fops.c:1937:client3_1_lookup_cbk] vol-client-2: gfid changed for /glusterfs.git/.git/config [2011-02-25 17:12:11.611753] W [fuse-bridge.c:907:fuse_unlink_cbk] glusterfs-fuse: 580322: UNLINK() /glusterfs.git/.git/config => -1 (Invalid argument) [2011-02-25 17:12:11.611807] D [dht-common.c:2319:dht_open] vol-dht: no cached subvolume for fd=0x7fbf8231c270 [2011-02-25 17:12:11.611828] W [fuse-bridge.c:584:fuse_fd_cbk] glusterfs-fuse: 580323: OPEN() /glusterfs.git/.git/config => -1 (Invalid argum ent) [2011-02-25 17:12:11.613157] D [stat-prefetch.c:3587:sp_release] vol-stat-prefetch: cache hits: 5, cache miss: 0 [2011-02-25 17:12:11.614779] D [stat-prefetch.c:3587:sp_release] vol-stat-prefetch: cache hits: 0, cache miss: 1 The log file is attached.
*** Bug 2682 has been marked as a duplicate of this bug. ***
In our test setup, disabling stat-pretech xlator seems to be mitigating this issue. Consider this as a work around till a fix is out.
*** Bug 2962 has been marked as a duplicate of this bug. ***
PATCH: http://patches.gluster.com/patch/7267 in master (dht-rename: Unlink older link files before creating new one)
PATCH: http://patches.gluster.com/patch/7323 in release-3.1 (DHT-rename: Unlink older link files before creating new one.)
PATCH: http://patches.gluster.com/patch/7265 in release-3.2 (dht-rename: Unlink older link files before creating new one)
*** Bug 2895 has been marked as a duplicate of this bug. ***
tested with 3.2.1qa2
Issue still exists. Previous fix is incorrect.
The patch you've had me test: http://shell.gluster.com/~avati/0001-cluster-dht-dht_rename-handle-GFID-situations.patch passes all the failure tests I was able to repro with vanilla v3.1.5 on my test VMs.
*** Bug 3188 has been marked as a duplicate of this bug. ***
*** Bug 3216 has been marked as a duplicate of this bug. ***
CHANGE: http://review.gluster.com/78 (Change-Id: Ia024fd136c936c5a73e2cdacdb738e5d2912d619) merged in release-3.1 by Anand Avati (avati)
Incorrectly closed as patches for 3.2 and master still need to be committed.
*** Bug 3220 has been marked as a duplicate of this bug. ***
CHANGE: http://review.gluster.com/90 (In the pre-GFID era, the linkfile of the destination file could be reused) merged in master by Anand Avati (avati)
CHANGE: http://review.gluster.com/88 (In the pre-GFID era, the linkfile of the destination file could be reused) merged in release-3.1 by Anand Avati (avati)
CHANGE: http://review.gluster.com/89 (In the pre-GFID era, the linkfile of the destination file could be reused) merged in release-3.2 by Anand Avati (avati)
*** Bug 3222 has been marked as a duplicate of this bug. ***
After upgrading to 3.2.3 Cypress still sees this issue. I need a response from Engineering on this today please.
The issue seems to exist because few of the clients being used were still running gluster version 3.2.2. On upgradation to 3.2.3, the client has not seen the tests failing. Downgrading the severity to normal, till the client confirms that upgrading to 3.2.3 solves the problem
Closing the bug as the client has not run into the error after upgrading all glusterfs clients to 3.2.3 version