Your pg_num is fine, there's no reason to change it if you don't
encounter any issues. One could argue that your smaller OSDs have too
few PGs but the larger OSDs have reasonable values. I would probably
leave it as it is.
Regarding the inactive PGs, how are your pools configured? Can you share
ceph osd pool ls detail
It could be an issue with min_size (is it also set to 3?).
Zitat von Lindsay Mathieson <lindsay.mathieson@xxxxxxxxx>:
Nautilus 14.2.9, setup using Proxmox.
* 5 Hosts
* 18 OSDs with a mix of disk sizes (3TB, 1TB, 500GB), all bluestore
* Pool size = 3, pg_num = 512
According to:
https://docs.ceph.com/docs/nautilus/rados/operations/placement-groups/#preselection
With 18 OSD's I should be using pg_num=1024, but I actually have it
set to 512.
However autoscale is recommending pg_num=128
Additionally, I accidentally set autoscale to on, rather than warn,
so it started the process. I rapidly got a "Reduced data
availability: 2 pgs inactive" warning and io on the pool stopped. I
cleared the warning by restarting the effected OSD's for the pg id,
but then more cropped up. I only made it stop and restored access to
the pool by turning off autoscale and setting pg_num back to 512.
Autoscale warning:
POOL SIZE TARGET SIZE RATE RAW CAPACITY RATIO TARGET RATIO BIAS
PG_NUM NEW PG_NUM AUTOSCALE
ceph 3239G 3.0 31205G 0.3114
1.0 512 128 warn
osd tree
ID CLASS WEIGHT REWEIGHT SIZE RAW USE DATA OMAP META
AVAIL %USE VAR PGS STATUS TYPE NAME
-1 30.47374 - 30 TiB 7.4 TiB 7.4 TiB 28 MiB 23
GiB 23 TiB 24.40 1.00 - root default
-5 5.45798 - 5.5 TiB 1.4 TiB 1.3 TiB 4.4 MiB 3.4 GiB
4.1 TiB 24.78 1.02 - host loc
1 hdd 2.72899 0.95001 2.7 TiB 679 GiB 677 GiB 3.1 MiB 1.9 GiB
2.1 TiB 24.31 1.00 136 up osd.1
8 hdd 2.72899 1.00000 2.7 TiB 706 GiB 704 GiB 1.3 MiB 1.5 GiB
2.0 TiB 25.26 1.04 143 up osd.8
-9 1.36449 - 1.4 TiB 431 GiB 429 GiB 1.4 MiB 2.1 GiB
966 GiB 30.84 1.26 - host lod
5 hdd 0.90970 1.00000 932 GiB 291 GiB 290 GiB 1.1 MiB 1.1 GiB
641 GiB 31.20 1.28 59 up osd.5
12 hdd 0.45479 1.00000 466 GiB 140 GiB 139 GiB 293 KiB 1024 MiB
325 GiB 30.12 1.23 28 up osd.12
-11 10.91595 - 11 TiB 2.4 TiB 2.4 TiB 6.9 MiB 5.8 GiB
8.5 TiB 21.94 0.90 - host vnb
6 hdd 2.72899 1.00000 2.7 TiB 613 GiB 612 GiB 2.5 MiB 1.5 GiB
2.1 TiB 21.95 0.90 124 up osd.6
7 hdd 2.72899 1.00000 2.7 TiB 614 GiB 613 GiB 1.7 MiB 1.4 GiB
2.1 TiB 21.98 0.90 124 up osd.7
9 hdd 2.72899 1.00000 2.7 TiB 617 GiB 615 GiB 1.5 MiB 1.4 GiB
2.1 TiB 22.06 0.90 124 up osd.9
17 hdd 2.72899 1.00000 2.7 TiB 608 GiB 607 GiB 1.3 MiB 1.5 GiB
2.1 TiB 21.76 0.89 124 up osd.17
-3 4.54836 - 4.5 TiB 1.3 TiB 1.3 TiB 11 MiB 7.0 GiB
3.3 TiB 27.76 1.14 - host vnh
0 hdd 0.90970 0.95001 932 GiB 220 GiB 219 GiB 3.0 MiB 1021 MiB
711 GiB 23.64 0.97 44 up osd.0
2 hdd 0.90970 0.95001 932 GiB 252 GiB 251 GiB 536 KiB 1023 MiB
679 GiB 27.06 1.11 51 up osd.2
10 hdd 0.54579 1.00000 559 GiB 158 GiB 157 GiB 1.6 MiB 1022 MiB
401 GiB 28.29 1.16 32 up osd.10
11 hdd 0.54579 1.00000 559 GiB 157 GiB 156 GiB 332 KiB 1024 MiB
402 GiB 28.10 1.15 32 up osd.11
14 hdd 0.54579 1.00000 559 GiB 187 GiB 186 GiB 1.1 MiB 1.0 GiB
372 GiB 33.45 1.37 38 up osd.14
15 hdd 0.54579 1.00000 559 GiB 159 GiB 158 GiB 2.0 MiB 1022 MiB
400 GiB 28.51 1.17 32 up osd.15
16 hdd 0.54579 1.00000 559 GiB 159 GiB 158 GiB 2.8 MiB 1021 MiB
400 GiB 28.46 1.17 32 up osd.16
-7 8.18697 - 8.2 TiB 2.0 TiB 2.0 TiB 4.6 MiB 4.5 GiB
6.2 TiB 24.50 1.00 - host vni
3 hdd 2.72899 1.00000 2.7 TiB 670 GiB 669 GiB 1.2 MiB 1.4 GiB
2.1 TiB 23.99 0.98 134 up osd.3
4 hdd 2.72899 1.00000 2.7 TiB 681 GiB 679 GiB 1.6 MiB 1.6 GiB
2.1 TiB 24.36 1.00 136 up osd.4
13 hdd 2.72899 1.00000 2.7 TiB 703 GiB 701 GiB 1.8 MiB 1.5 GiB
2.0 TiB 25.14 1.03 143 up osd.13
TOTAL 30 TiB 7.4 TiB 7.4 TiB 28 MiB 23
GiB 23 TiB 24.40
Should I be reducing the pg_num? is there a way to do it safely?
Thanks.
--
Lindsay
_______________________________________________
ceph-users mailing list -- ceph-users@xxxxxxx
To unsubscribe send an email to ceph-users-leave@xxxxxxx
_______________________________________________
ceph-users mailing list -- ceph-users@xxxxxxx
To unsubscribe send an email to ceph-users-leave@xxxxxxx
