Re: too many PGs per OSD when pg_num = 256??

David Turner <david.turner@xxxxxxxxxxxxxxxx> · Thu, 22 Sep 2016 17:03:55 +0000

So you have 3,520 pgs.  Assuming all of your pools are using 3 replicas, and using the 377 pgs/osd in your health_warn state, that would mean your cluster has 28 osds.

When you calculate how many pgs a pool should have, you need to account for how many osds you have, how much percentage of data each pool will account for out of your entire cluster, and go from there.  The ceph PG Calc tool will be an excellent resource to
 help you figure out how many pgs each pool should have.  It takes all of those factors into account.  http://ceph.com/pgcalc/

David Turner |
Cloud Operations Engineer |
StorageCraft
 Technology Corporation

380 Data Drive Suite 300 |
Draper |
Utah |
84020

Office:
801.871.2760 |
Mobile:
385.224.2943

If you are not the intended recipient of this message or received it erroneously, please notify the sender and delete it, together with any attachments, and be advised that any dissemination or copying of this
 message is prohibited.

From: Andrus, Brian Contractor [bdandrus@xxxxxxx]

Sent: Thursday, September 22, 2016 10:41 AM

To: David Turner; ceph-users@xxxxxxxxxxxxxx

Subject: RE: too many PGs per OSD when pg_num = 256??

David,
I have 15 pools:
# ceph osd lspools|sed 's/,/\n/g'
0 rbd
1 cephfs_data
2 cephfs_metadata
3 vmimages
14 .rgw.root
15 default.rgw.control
16 default.rgw.data.root
17 default.rgw.gc
18 default.rgw.log
19 default.rgw.users.uid
20 default.rgw.users.keys
21 default.rgw.users.email
22 default.rgw.meta
23 default.rgw.buckets.index
24 default.rgw.buckets.data
# ceph -s | grep -Eo '[0-9]+ pgs'
3520 pgs

Brian Andrus
ITACS/Research Computing
Naval Postgraduate School
Monterey, California
voice: 831-656-6238

From: David Turner [mailto:david.turner@xxxxxxxxxxxxxxxx]

Sent: Thursday, September 22, 2016 8:57 AM

To: Andrus, Brian Contractor <bdandrus@xxxxxxx>; ceph-users@xxxxxxxxxxxxxx

Subject: RE: too many PGs per OSD when pg_num = 256??

Forgot the + for the regex.

ceph -s | grep -Eo '[0-9]+ pgs' 

David Turner | Cloud Operations Engineer |
StorageCraft Technology Corporation

380 Data Drive Suite 300 | Draper | Utah | 84020

Office: 801.871.2760 |
Mobile: 385.224.2943

If you are not the intended recipient of this message or received it erroneously, please notify the sender and delete it, together with any attachments, and be advised that any dissemination
 or copying of this message is prohibited.

From: David Turner

Sent: Thursday, September 22, 2016 9:53 AM

To: Andrus, Brian Contractor; 
ceph-users@xxxxxxxxxxxxxx

Subject: RE: too many PGs per OSD when pg_num = 256??

How many pools do you have?  How many pgs does your total cluster have, not just your rbd pool?

ceph osd lspools

ceph -s | grep -Eo '[0-9] pgs'

My guess is that you have other pools with pgs and the cumulative total of pgs per osd is too many.

From: ceph-users [ceph-users-bounces@xxxxxxxxxxxxxx]
 on behalf of Andrus, Brian Contractor [bdandrus@xxxxxxx]

Sent: Thursday, September 22, 2016 9:33 AM

To: ceph-users@xxxxxxxxxxxxxx

Subject: [ceph-users] too many PGs per OSD when pg_num = 256??

All, 

I am getting a warning:

     health HEALTH_WARN
            too many PGs per OSD (377 > max 300)
            pool cephfs_data has many more objects per pg than average (too few pgs?)

yet, when I check the settings:
# ceph osd pool get rbd pg_num
pg_num: 256
# ceph osd pool get rbd pgp_num
pgp_num: 256

How does something like this happen?
I did create a radosgw several weeks ago and have put a single file in it for testing, but that is it. It only started giving the warning a couple days ago.

Brian Andrus
ITACS/Research Computing
Naval Postgraduate School
Monterey, California
voice: 831-656-6238

_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com