On Sunday, November 25, 2012 at 9:30 AM, Drunkard Zhang wrote: > 2012/11/26 Sage Weil <s...@inktank.com (mailto:s...@inktank.com)>: > > On Sun, 25 Nov 2012, Drunkard Zhang wrote: > > > I'm using ceph-0.51. I setup 3 monitors. then mount with 3 mon IP at > > > another host with either kernel mode or fuse, neither give me > > > redundancy. > > > > > > Here's commands I used at client (IP: 10.0.0.2) side; > > > mount -t ceph log3,log21,squid86-log12:/ /mnt/bc > > > ceph-fuse -m log3,log21,squid86-log12 /mnt/bc > > > > > > When mounted, read/write is OK, then add a iptables rule at log3 to > > > REJECT/DROP packets from client 10.0.0.2, operation at client stuck > > > due to IO problem. > > > > > > Related stuck processes on client look like this: > > > log1 ~ # ps auwx |g bc > > > root 1325 0.0 0.0 120100 3536 pts/0 Sl 23:10 0:00 > > > ceph-fuse -m log3,log21,squid86-log12 /mnt/bc > > > root 1404 0.0 0.0 16192 700 pts/0 S 23:10 0:00 ls > > > --color=auto /mnt/bc > > > > > > > > Can you repeat teh test with '--debug-monc 20 --debug-ms 1 --log-file > > /tmp/foo' > > on the command line and attach the resulting log? > > > > Thanks! > > sage > > > Thanks for the hint, I think the problem found, I rejected the client > at mon.log3 which also acts as mds, and it's up and active, another > mds is up and standby. So the client cannot connect to mds server? Is > it possible to get more than one mds up and active simultaneously, and > let client know that?
The active MDS servers partition the namespace between them — clients need to be able to communicate with them all. You can configure which MDS is active (at start-up) if you like; unfortunately the best reference is still in the wiki: http://ceph.com/deprecated/Standby-replay_modes -Greg -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
