Bug 763303 (GLUSTER-1571) - After self healing client see´s wrong file information
Summary: After self healing client see´s wrong file information
Keywords:
Status: CLOSED WONTFIX
Alias: GLUSTER-1571
Product: GlusterFS
Classification: Community
Component: replicate
Version: 3.0.4
Hardware: x86_64
OS: Linux
low
medium
Target Milestone: ---
Assignee: Pranith Kumar K
QA Contact:
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2010-09-08 14:04 UTC by juergen winkler
Modified: 2011-03-22 07:25 UTC (History)
3 users (show)

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Last Closed:
Regression: ---
Mount Type: fuse
Documentation: ---
CRM:
Verified Versions:


Attachments (Terms of Use)

Description juergen winkler 2010-09-08 14:04:13 UTC
There are 2 glusterfs exports witch have the same file and file attributes after self healing .
(Export with Glusterfs Server 3.0.5 on SLES 11)
But the Client who has both of them mounted via client side replication can´t open the file.
The file has a different md5sum on the client as on the Servers. On the Servers they are the same.
You can do an ls -l on the file but nothing happens in fact that they are the same on the Servers there is no further healing done.
When you do an ls -lR on the whole directory the file will be healed. 
No automated healing is done on file access.
(The client is 3.0.4 on SLES11)

Comment 1 shishir gowda 2010-09-10 04:03:53 UTC
Can you please confirm is performance/stat-prefetch translator is used in the client? There is a known issue if stat-prefetch translator is used.

If it is, can you please disable the stat-prefetch translator and report if the issue still exists?

Comment 2 juergen winkler 2010-09-13 08:19:41 UTC
no the stat-prefetch translator is not in use in our client config.
so i think the issue still exists.

Comment 3 shishir gowda 2010-09-15 05:40:55 UTC
Hi,

We are not able to recreate the issue inhouse.

Can you please upgrade 3.0.5 release, and see if the issue is resolved?

With regards,
Shishir

Comment 4 juergen winkler 2010-09-15 11:16:54 UTC
Hi,

i would love to upgrade on 3.0.5 but wit this Version as Client we had the Probelm that when something tried to write or read on the Mount, the Mount becomes unaccessable.
This happened on 4 of our 20 Servers whitch are using Glusterfs. So this is not an Option for us i think.

If i can provide u more information to reproduce this let me know what u need.

Comment 5 shishir gowda 2010-09-15 11:28:26 UTC
When the issue re-occurs, please get a process state dump and attach it to the bug, along the log files. This might aid us to triage issue better.

To take a process state dump run the following command (This does not affect the glusterfs running)

kill -s SIGUSR1 <PID of glusterfs>  (please provide server and client dumps).

This will create files under /tmp/glusterdump.<pid>

With regards,
Shishir

Comment 6 Amar Tumballi 2010-10-05 07:09:33 UTC
Hi Juergen,
 We had some issues with 3.0.x replicate which were hard to address due to some internal design limitations. Can you try with 3.1.x (currently in beta) releases and see if its fixed?

Comment 7 Pranith Kumar K 2011-03-22 04:25:13 UTC
This bug can not be fixed with the design of 3.0.x. So closing this bug.


Note You need to log in before you can comment on or make changes to this bug.