Bug 764413 (GLUSTER-2681) - marker returns dummy data for volume-mark attribute if xtime is switched off
Summary: marker returns dummy data for volume-mark attribute if xtime is switched off
Keywords:
Status: CLOSED CURRENTRELEASE
Alias: GLUSTER-2681
Product: GlusterFS
Classification: Community
Component: geo-replication
Version: mainline
Hardware: All
OS: Linux
low
medium
Target Milestone: ---
Assignee: Junaid
QA Contact:
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2011-04-06 17:12 UTC by Csaba Henk
Modified: 2013-08-06 22:37 UTC (History)
3 users (show)

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Last Closed:
Regression: RTNR
Mount Type: ---
Documentation: ---
CRM:
Verified Versions: 3.2.3

Attachments (Terms of Use)
Add an attachment (proposed patch, testcase, etc.)

Description Csaba Henk 2011-04-06 17:12:31 UTC 
In a client mounted with --client-pid=-1, if marker is disabled on the server side, we can observe the following (part of strace output):

getxattr(".", "trusted.glusterfs.volume-mark", "\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00", 27) = 27

Proper behavior would be failing the syscall with ENODATA.

This is pretty bad b/c then gsyncd actually starts to synchronize in the name of volume 00000000-0000-0000-0000-000000000000, instead of waiting for a proper master to come up. (The phenomenon can be clearly blamed on marker and not aggregation, as I was using a simple non-clustered volume).
Comment 1 Lakshmipathi G 2011-09-15 08:48:27 UTC 
tested with 3.2.3 -mounted with client-pid=-1 and marker disabled, it gives

# getfattr -n trusted.glusterfs.volume-mark .
.: trusted.glusterfs.volume-mark: Invalid argument

 
getxattr(".", "trusted.glusterfs.volume-mark", 0x0, 0) = -1 EINVAL (Invalid argument)

Its successful when marker is enabled.
# getfattr -n trusted.glusterfs.volume-mark .
# file: .
trusted.glusterfs.volume-mark=0sAQBKH/x/m7dGp4IQHRSYJpV8AE5xnuAADywv
Note You need to log in before you can comment on or make changes to this bug.