Hi,
I'm using cephfs for a while, it's great, but I found a serious problem now.
My situation is simple: a lot of servers receive syslog to local disk,
then mount ceph FS via kernel driver and move those received files to
ceph FS.
Some days ago, I found my script missed some file while processing
data, then I noticed those files' size is zero. But it's impossible,
rsyslog won't generate file until data arrives, so I logged the file
size before mv, so here's one example from these exceptions:
In log:
2013-11-07 18:17:29 /scripts/log/collector: file size: 4910689
/gwbn/nat/20131107/dnsquery-pb24-192.168.156.60-by_squid79-20131107_17.log
Zero size file on ceph FS:
-rw-r--r-- 1 root root 0 Jan 1 1970
dnsquery-pb24-192.168.156.60-by_squid79-20131107_17.log
stat dnsquery-pb24-192.168.156.60-by_squid79-20131107_17.log
File: ‘dnsquery-pb24-192.168.156.60-by_squid79-20131107_17.log’
Size: 0 Blocks: 0 IO Block: 4194304 regular empty file
Device: 0h/0d Inode: 1099518968218 Links: 1
Access: (0644/-rw-r--r--) Uid: ( 0/ root) Gid: ( 0/ root)
Access: 2013-11-07 17:00:00.167501860 +0800
Modify: 1970-01-01 08:00:00.000000000 +0800
Change: 2013-11-07 18:17:29.843558000 +0800
Birth: -
I'm running multiple versions of kernel, scales from 3.8.x to 3.11.6.
Most of the zero file event occurs on those servers with kernel
version of 3.10.[45], some on 3.8.6 and 3.11.4, I didn't tested all of
them, these is my running versions.
I think the ceph FS kernel driver should sync before umount, isn't it
? So what can I do to shoot this problem down ? And any advice of
temparory solution to avoid this, manually sync?
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
