Thanks for the help. I can trigger the OSD to use 1 replica (which I assume is just the primary) using './ceph osd pool set metadata size 1'. Unfortunately, this seems to prevent the client from properly mounting.
Here is the output I get. It seems once the replication level is changed it has trouble opening a session with the MDS. googoo-10 59> ./ceph osd pool set metadata size 1 09.09.02 14:02:17.313654 mon1 <- [osd,pool,set,metadata,size,1] 09.09.02 14:02:18.345679 mon0 -> 'set pool 1 size to 1' (0) googoo-10 60> ./ceph osd dump -o - 09.09.02 14:02:32.293234 mon1 <- [osd,dump] 09.09.02 14:02:32.294275 mon1 -> 'dumped osdmap epoch 3' (0) epoch 3 fsid 9a900213-f3e3-9cf2-b2ca-e209b866c57a created 09.09.02 14:02:02.880448 modifed 09.09.02 14:02:18.316561 pg_pool 0 'data' pg_pool(rep size 2 ruleset 0 pg_num 16 pgp_num 16 lpg_num 1 lpgp_num 1 last_change 1) pg_pool 1 'metadata' pg_pool(rep size 1 ruleset 1 pg_num 16 pgp_num 16 lpg_num 1 lpgp_num 1 last_change 3) pg_pool 2 'casdata' pg_pool(rep size 2 ruleset 2 pg_num 16 pgp_num 16 lpg_num 1 lpgp_num 1 last_change 1) max_osd 4 osd0 in weight 1 up (up_from 2 up_thru 0 down_at 0 last_clean 0-0) 127.0.0.1:6800/15206/0 osd1 out down (up_from 0 up_thru 0 down_at 0 last_clean 0-0) osd2 out down (up_from 0 up_thru 0 down_at 0 last_clean 0-0) osd3 out down (up_from 0 up_thru 0 down_at 0 last_clean 0-0) 09.09.02 14:02:32.294348 wrote 721 byte payload to - googoo-10 61> ./csyn --syn trace sometracefile 1 --debug_client 15 starting csyn at 0.0.0.0:6802/15311/0 mounting and starting 1 syn client(s) 09.09.02 14:02:41.495397 client-1 waiting for client(s) to finish 09.09.02 14:02:41.495701 client-1 initing 09.09.02 14:02:41.495883 client-1 mounting 09.09.02 14:02:41.519017 3065953168 client0 mounted: have osdmap 0 and mdsmap 0 09.09.02 14:02:41.519146 3065953168 client0 random mds-1 09.09.02 14:02:41.519171 3065953168 client0 chose target mds0 based on hierarchy 09.09.02 14:02:41.519192 3065953168 client0 no address for mds0, requesting new mdsmap 09.09.02 14:02:41.524293 3055463312 client0 handle_mds_map epoch 5 09.09.02 14:02:41.524492 3065953168 client0 opening session to mds0 09.09.02 14:02:41.524761 3065953168 client0 waiting for session to mds0 to open 09.09.02 14:02:42.496894 3044973456 client0 renew_caps() Thanks again. Andrew On Aug 28, 2009, at 3:12 PM, Yehuda Sadeh Weinraub wrote: > On Fri, Aug 28, 2009 at 1:09 PM, Andrew Leung <ale...@soe.ucsc.edu> > wrote: >> >> I am doing some basic Ceph testing using to most recent release >> (v0.13). I'm currently using the simple vstart.sh script to get a >> client/mds/mon/osd up and running on a single machine. >> >> I'm wondering how I can toggle the replication level (e.g., turn on >> 2- >> way replication, turn off replication). I'm particularly interested >> in >> doing so for metadata replication. Is there a flag I can input or >> modify in the vstart script or ceph.conf file? >> >> Thanks. >> Andrew >> > > According to Sage it goes like this: > > $ ./ceph osd pool set metadata size 1 > $ ./ceph pg dump -o - > > You can figure out current replication size from this: > $ ./ceph osd dump -o - > > Then you can run: > $ ./ceph osd pool set metadata size 2 > > And then you can watch the replication happening by running'ceph -w'. ------------------------------------------------------------------------------ Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day trial. Simplify your report design, integration and deployment - and focus on what you do best, core application coding. Discover what's new with Crystal Reports now. http://p.sf.net/sfu/bobj-july _______________________________________________ Ceph-devel mailing list Ceph-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/ceph-devel
