Bug 761955 (GLUSTER-223) - flush not sent
Summary: flush not sent
Keywords:
Status: CLOSED CURRENTRELEASE
Alias: GLUSTER-223
Product: GlusterFS
Classification: Community
Component: fuse
Version: 2.0.5
Hardware: All
OS: Linux
low
medium
Target Milestone: ---
Assignee: Csaba Henk
QA Contact:
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2009-08-20 09:07 UTC by Vikas Gorur
Modified: 2011-02-15 07:39 UTC (History)
4 users (show)

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


Attachments (Terms of Use)

Description Vikas Gorur 2009-08-20 09:07:25 UTC
The following was observed in connection with running Xen.

Both replicate backends have 0x00000001 xattr, strongly indicating that
the decrementing xattrop was never called. The client log shows that after many
writes, release was called, with flush never being called. And replicate does
the decrement only in flush.

[2009-08-20 14:26:02] T [fuse-bridge.c:1567:fuse_write] glusterfs-fuse: 8991: WRITE (0x70c9f
b0, size=2560, offset=107126784)
[2009-08-20 14:26:02] T [fuse-bridge.c:1528:fuse_writev_cbk] glusterfs-fuse: 8991: WRITE =>
2560/2560,107126784/8589934594
[2009-08-20 14:26:06] T [fuse-bridge.c:1610:fuse_release] glusterfs-fuse: 8992: RELEASE 0x70
c9fb0

This is further corroborated by the server logs:

[2009-08-20 14:26:02] T [server-protocol.c:3973:server_writev] brick: 36538: WRITEV 'fd=0 (1
776309); offset=111030272; size=2560
[2009-08-20 14:26:02] T [server-protocol.c:3973:server_writev] brick: 36539: WRITEV 'fd=0 (1
776309); offset=107125248; size=1536
[2009-08-20 14:26:02] T [server-protocol.c:3973:server_writev] brick: 36540: WRITEV 'fd=0 (1
776309); offset=107126784; size=2560
[2009-08-20 14:26:06] T [server-protocol.c:4027:server_release] brick: 36541: RELEASE 'fd=0'

Kernel version: 2.6.18-128.1.10.el5xen
fuse init (API version 7.8)
fuse distribution version: 2.7.4

Comment 1 Anand Avati 2009-09-23 13:19:28 UTC
PATCH: http://patches.gluster.com/patch/1442 in master (fuse: emit a flush from release if we didn't get an adjacent FLUSH message from the kernel)

Comment 2 Anand Avati 2009-09-23 13:19:32 UTC
PATCH: http://patches.gluster.com/patch/1466 in release-2.0 (fuse: emit a flush from release if we didn't get an adjacent FLUSH message from the kernel)

Comment 3 Anand Avati 2010-10-27 03:56:31 UTC
PATCH: http://patches.gluster.com/patch/5362 in master (fuse: retire the faked-FLUSH-upon-RELEASE hack)


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