Hi,
As far as I know, this is exactly the problem why the new tunables where
introduced, If you use 3 Replicas with only 3 hosts, crush sometimes doesn't
find a solution to place all pgs.
If you are really stuck with bobtail turntables, I can think of 2 possible
workarounds:
1. Add another osd Server.
2. Bad idea, but could work: build your crush rule manually, e.g.: set all
primary pgs to host ceph1, first copy to host ceph2 and second copy to host3.
Micha Krause
Am 08.07.2016 um 05:47 schrieb Nathanial Byrnes:
Hello,
I've got a Jewel Cluster (3 nodes, 15 OSD's) running with bobtail tunables
(my xenserver cluster uses 3.10 as the kernel and there's no upgrading
that....). I started the cluster out on Hammer, upgraded to Jewel, discovered
that optimal tunables would not work, and then set the tunables back to
bobtail. Once the re-balancing completed, I was stuck with 1 pg in
active+remapped. Repair didn't fix the pg. I then upped the number of pgs from
328 to 712 (oddly I asked for 512, but ended p with 712...), now I have 5 pgs
stuck in active+remapped. I also tried re-weighting the pgs a couple times, but
no change.... Here is my osd tree:
ID WEIGHT TYPE NAME UP/DOWN REWEIGHT PRIMARY-AFFINITY
-1 15.00000 root default
-2 5.00000 host ceph1
0 1.00000 osd.0 up 0.95001 1.00000
1 1.00000 osd.1 up 1.00000 1.00000
2 1.00000 osd.2 up 1.00000 1.00000
3 1.00000 osd.3 up 0.90002 1.00000
4 1.00000 osd.4 up 1.00000 1.00000
-3 5.00000 host ceph3
10 1.00000 osd.10 up 1.00000 1.00000
11 1.00000 osd.11 up 1.00000 1.00000
12 1.00000 osd.12 up 1.00000 1.00000
13 1.00000 osd.13 up 1.00000 1.00000
14 1.00000 osd.14 up 1.00000 1.00000
-4 5.00000 host ceph2
5 1.00000 osd.5 up 1.00000 1.00000
6 1.00000 osd.6 up 1.00000 1.00000
7 1.00000 osd.7 up 1.00000 1.00000
8 1.00000 osd.8 up 1.00000 1.00000
9 1.00000 osd.9 up 1.00000 1.00000
Any suggestions on how to troubleshoot or repair this?
Thanks and Regards,
Nate
_______________________________________________
ceph-users mailing list
[email protected]
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
_______________________________________________
ceph-users mailing list
[email protected]
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
