[Gluster-devel] difficult bug in 2.5 mainline

Mon Jul 2 12:00:33 UTC 2007

While most backups are now failing with loss of connectivity, One test I ran on a mailbox that successful backuped up under patch 245 now fails late in the process after copying over 16000 files from gluster to backup/tmp when failing to create a link. I know Zimbra uses hard links to save space.

Here is the error message from Zimbra:
ERRORS

harrisl at lhjonline.com: link(/opt/zimbra/backup/sessions/full-20070629.175030.298/accounts/69e/308/69e308b0-0923-4aa7-9036-3bbae4806b2c/blobs/3/4/bZyRbDLjWxuGkqeDEktny,oLmXY=20179-19047.msg1, /opt/zimbra/backup/tmp/full-20070702.111546.524/accounts/69e/308/69e308b0-0923-4aa7-903
java.io.IOException: link(/opt/zimbra/backup/sessions/full-20070629.175030.298/accounts/69e/308/69e308b0-0923-4aa7-9036-3bbae4806b2c/blobs/3/4/bZyRbDLjWxuGkqeDEktny,oLmXY=20179-19047.msg1, /opt/zimbra/backup/tmp/full-20070702.111546.524/accounts/69e/308/69e308b0-0923-4aa7-903
at com.zimbra.znative.IO.link0(Native Method)
at com.zimbra.znative.IO.link(IO.java:48)
at com.zimbra.common.io.AsyncFileCopier$WorkerThread.link(AsyncFileCopier.java:167)
at com.zimbra.common.io.AsyncFileCopier$WorkerThread.run(AsyncFileCopier.java:130) 

Harris

----- Original Message -----
From: "Harris Landgarten" <harrisl at lhjonline.com>
To: "Harris Landgarten" <harrisl at lhjonline.com>
Cc: "gluster-devel" <gluster-devel at nongnu.org>, "Amar S. Tumballi" <amar at zresearch.com>
Sent: Monday, July 2, 2007 7:04:40 AM (GMT-0500) America/New_York
Subject: Re: [Gluster-devel] difficult bug in 2.5 mainline

Re ran backup test with bricks running under gdb

Brick1 crash bt:

0xb7e4731d in vasprintf () from /lib/libc.so.6
(gdb) bt
#0  0xb7e4731d in vasprintf () from /lib/libc.so.6
#1  0xb7e2e6be in asprintf () from /lib/libc.so.6
#2  0xb75c11df in server_lookup_cbk () from /usr/lib/glusterfs/1.3.0-pre5/xlator/protocol/server.so
#3  0xb7f5d5be in call_resume (stub=0xb5c03938) at call-stub.c:2697
#4  0xb75d1770 in iot_reply () from /usr/lib/glusterfs/1.3.0-pre5/xlator/performance/io-threads.so
#5  0xb7f2d3db in start_thread () from /lib/libpthread.so.0
#6  0xb7eb726e in clone () from /lib/libc.so.6

brick2 seqfault

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread -1243444336 (LWP 14631)]
dict_serialized_length (dict=0x80cc448) at dict.c:333
333         len += strlen (pair->key) + 1;
(gdb) bt
#0  dict_serialized_length (dict=0x80cc448) at dict.c:333
#1  0xb7fc748b in gf_block_to_iovec (blk=0x80c6a58, iov=0xb5e28290, cnt=3) at protocol.c:410
#2  0xb762e538 in generic_reply () from /usr/lib/glusterfs/1.3.0-pre5/xlator/protocol/server.so
#3  0xb7631229 in server_lookup_cbk () from /usr/lib/glusterfs/1.3.0-pre5/xlator/protocol/server.so
#4  0xb7fcd5be in call_resume (stub=0x8099b38) at call-stub.c:2697
#5  0xb7641770 in iot_reply () from /usr/lib/glusterfs/1.3.0-pre5/xlator/performance/io-threads.so
#6  0xb7f9d3db in start_thread () from /lib/libpthread.so.0
#7  0xb7f2726e in clone () from /lib/libc.so.6

Harris

----- Original Message -----
From: "Harris Landgarten" <harrisl at lhjonline.com>
To: "Harris Landgarten" <harrisl at lhjonline.com>
Cc: "gluster-devel" <gluster-devel at nongnu.org>, "Amar S. Tumballi" <amar at zresearch.com>
Sent: Monday, July 2, 2007 6:54:12 AM (GMT-0500) America/New_York
Subject: Re: [Gluster-devel] difficult bug in 2.5 mainline

Backup test run on patch 252:

Zimbra client crashed with BT:

#0  0xb7f81f4c in raise () from /lib/libpthread.so.0
(gdb) bt
#0  0xb7f81f4c in raise () from /lib/libpthread.so.0
#1  0xb7fac628 in gf_print_trace (signum=6) at common-utils.c:211
#2  <signal handler called>
#3  0xb7e7e986 in raise () from /lib/libc.so.6
#4  0xb7e80043 in abort () from /lib/libc.so.6
#5  0xb7e7812d in __assert_fail () from /lib/libc.so.6
#6  0xb7fafe90 in inode_unref (inode=0x80b5fc8) at inode.c:336
#7  0x0804b7a4 in fuse_loc_wipe (fuse_loc=0x847e678) at fuse-bridge.c:97
#8  0x0804b82d in free_state (state=0x847e670) at fuse-bridge.c:129
#9  0x0804efb4 in fuse_entry_cbk (frame=0x84cb380, cookie=0x84ce360, this=0x8058db0, op_ret=8, op_errno=107, inode=0x80b5fc8, buf=0x84b8d90) at fuse-bridge.c:368
#10 0xb7fa9cac in default_lookup_cbk (frame=0x84ce360, cookie=0x84bb7e8, this=0x80587f0, op_ret=8, op_errno=107, inode=0x80b5fc8, buf=0x84b8d90) at defaults.c:40
#11 0xb7fa9cac in default_lookup_cbk (frame=0x84bb7e8, cookie=0x847dc28, this=0x8058760, op_ret=8, op_errno=107, inode=0x80b5fc8, buf=0x84b8d90) at defaults.c:40
#12 0xb75edbbd in unify_sh_opendir_cbk (frame=0x847dc28, cookie=0x8052500, this=0x80579e8, op_ret=8, op_errno=17, fd=0x843ffa0) at unify-self-heal.c:380
#13 0xb75f5f62 in client_opendir_cbk (frame=0x84b9088, args=0x80929b8) at client-protocol.c:3213
#14 0xb75f9077 in notify (this=0x8052a68, event=2, data=0x80902b8) at client-protocol.c:4191
#15 0xb7fada27 in transport_notify (this=0x6e5d, event=6) at transport.c:152
#16 0xb7fae499 in sys_epoll_iteration (ctx=0xbffcfff8) at epoll.c:54
#17 0xb7fadafd in poll_iteration (ctx=0xbffcfff8) at transport.c:260
#18 0x0804a170 in main (argc=6, argv=0xbffd00d4) at glusterfs.c:341

brick2 with namespace crashed as well.
brick1 stayed up
client2 recovered when brick2 was restarted.

No data was written from gluster to backup tmp.

Harris

----- Original Message -----
From: "Harris Landgarten" <harrisl at lhjonline.com>
To: "Harris Landgarten" <harrisl at lhjonline.com>
Cc: "gluster-devel" <gluster-devel at nongnu.org>, "Amar S. Tumballi" <amar at zresearch.com>
Sent: Sunday, July 1, 2007 10:03:02 PM (GMT-0500) America/New_York
Subject: Re: [Gluster-devel] difficult bug in 2.5 mainline

The backup hung as first described. No data was written from the secondary volume on gluster to the backup tmp dir.

Harris

----- Original Message -----
From: "Harris Landgarten" <harrisl at lhjonline.com>
To: "Amar S. Tumballi" <amar at zresearch.com>
Cc: "gluster-devel" <gluster-devel at nongnu.org>
Sent: Sunday, July 1, 2007 9:46:18 PM (GMT-0500) America/New_York
Subject: Re: [Gluster-devel] difficult bug in 2.5 mainline

Amar,

The rm -rf bug is still there. See the last comment by Daniel to the ml in reply to the problem with rm -rf post to the ml. BTW files are being deleted but at the rate of about 1 every 3 sec with lots of lookups in the logs. I am going to check the other problem now.

Harris

----- Original Message -----
From: "Amar S. Tumballi" <amar at zresearch.com>
To: "Harris Landgarten" <harrisl at lhjonline.com>
Cc: "gluster-devel" <gluster-devel at nongnu.org>
Sent: Sunday, July 1, 2007 7:55:09 PM (GMT-0500) America/New_York
Subject: Re: [Gluster-devel] difficult bug in 2.5 mainline

Hi Harris, 
With the latest patch this bug is fixed. Also, i hope it should fix the problem of 'rm -rf' too.. please confirm. 

i am looking into other strange bug reported by you. 

-bulde 

On 7/2/07 , Harris Landgarten < harrisl at lhjonline.com > wrote: 

Disabling posix-locks changes the problem 

The client crashes along with the lock-server brick 

Here is the bt from the client: 

#0 unify_bg_cbk (frame=0xe080168, cookie=0xe1109c8, this=0x8057730, op_ret=0, op_errno=13) at unify.c:83 
83 callcnt = --local->call_count; 
(gdb) bt 
#0 unify_bg_cbk (frame=0xe080168, cookie=0xe1109c8, this=0x8057730, op_ret=0, op_errno=13) at unify.c:83 
#1 0xb75b96e5 in client_unlink_cbk (frame=0xe1109c8, args=0x8059248) at client-protocol.c:2969 
#2 0xb75beff5 in notify (this=0x8057730, event=2, data=0x8095338) at client-protocol.c:4184 
#3 0xb7f73827 in transport_notify (this=0x0, event= 235405672 ) at transport.c:152 
#4 0xb7f74299 in sys_epoll_iteration (ctx=0xbfb96248) at epoll.c:54 
#5 0xb7f738fd in poll_iteration (ctx=0xbfb96248) at transport.c:260 
#6 0x0804a170 in main (argc=6, argv=0xbfb96324) at glusterfs.c:341 
(gdb) print local 
$1 = (unify_local_t *) 0x0 

Harris 

----- Original Message ----- 
From: "Harris Landgarten" < harrisl at lhjonline.com > 
To: "gluster-devel" < gluster-devel at nongnu.org > 
Sent: Sunday, July 1, 2007 10:56:05 AM (GMT-0500) America/New_York 
Subject: [Gluster-devel] difficult bug in 2.5 mainline 

I am trying to track down a bug that is causing hangs in 2.5-patch-249 and all previous. 

This happens during a full Zimbra backup of certain accounts to /mnt/glusterfs/backups 

The first stage of the backup copies indexes and primary storage to /mnt/glusterfs/backups/tmp 
All of this data resides in local storage and the writing to gluster is successful. 

The next stage copies secondary storage to /mnt/glusterfs/backups/tmp 
This fails in the following way: 

Brick1 hangs with no errors 
Brick2 hangs with no errors 
Zimbra client hangs with no errors 
second client loses connectivity 

The second client bails after 2 min but cannot connect 
The Zimbra client never bails 

I then restart the bricks 

After both bricks are restarted, the second client reconnects and a hung df -h completes 

Zimbra client stays in a hung unconnected start 

ls -l /mnt/glusterfs hangs 

Only way is reset is 

kill -9 pidof glusterfs 
umount /mnt/glusterfs 

glusterfs 

Post mortem examination of /mnt/glusterfs/backups/tmp shows that a few files have the written from the secondary storage volume. I this can over 15,000 files should have been written. 

Note: this only happen with large email boxed with some large >10M files. 

Note: with patch-247 the zimbra client would seqfault. With 249 it just hangs in unrecoverable state. 

Harris 

_______________________________________________ 
Gluster-devel mailing list 
Gluster-devel at nongnu.org 
http://lists.nongnu.org/mailman/listinfo/gluster-devel 

_______________________________________________ 
Gluster-devel mailing list 
Gluster-devel at nongnu.org 
http://lists.nongnu.org/mailman/listinfo/gluster-devel 

-- 
Amar Tumballi 
http://amar.80x25.org 
[bulde on #gluster/irc.gnu.org] 

_______________________________________________
Gluster-devel mailing list
Gluster-devel at nongnu.org
http://lists.nongnu.org/mailman/listinfo/gluster-devel