ubuntu-syncdaemon produces "disk wait CPU" death and filesystem crash
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
ubuntuone-client (Ubuntu) |
Expired
|
Undecided
|
Unassigned |
Bug Description
In the last few days, soon after reboot ubuntu-syncdaemon, if not killed immediately, soon keeps the disk so busy that all the CPU goes to "wait CPU" and the system basically dies; after a while, the filesystem is made read-only (due to errors created by the hectic situation, I suppose), which is sometimes fixed by the disk checks on reboot and sometimes not (needing some fsck work).
I'm quite sure that ubuntu-syncdaemon is to blame because 1) if I disable sync the situation never reappears, 2) if I kill it immediately after reboot or when CPU wait begins to peak the system doesn't "crash", 3) if I don't do (1) nor (2) it's consistently reproduced.
It's not possible to run apport when the FS is read-only nor when the CPU is basically unavailable, hence what attached is taken before the actual problem begins.
ProblemType: Bug
DistroRelease: Ubuntu 11.10
Package: ubuntuone-client 2.0.1-0ubuntu1.1
ProcVersionSign
Uname: Linux 3.0.0-22-
ApportVersion: 1.23-0ubuntu4
Architecture: i386
Date: Tue Jul 10 18:58:45 2012
ExecutablePath: /usr/lib/
InstallationMedia: Ubuntu 10.10 "Maverick Meerkat" - Release i386 (20101007)
InterpreterPath: /usr/bin/python2.7
PackageArchitec
SourcePackage: ubuntuone-client
UbuntuOneSyncda
UbuntuOneUserSy
[__main__]
files_sync_enabled = True
log_level = DEBUG
UpgradeStatus: Upgraded to oneiric on 2011-10-27 (257 days ago)
Might be the same as bug 946565.