[ceph-users] Large omap objects - how to fix ?

Ben Morrice ben.morrice at epfl.ch
Fri Oct 26 06:09:09 PDT 2018


Hello all,

After a recent Luminous upgrade (now running 12.2.8 with all OSDs 
migrated to bluestore, upgraded from 11.2.0 and running filestore) I am 
currently experiencing the warning 'large omap objects'.
I know this is related to large buckets in radosgw, and luminous 
supports 'dynamic sharding' - however I feel that something is missing 
from our configuration and i'm a bit confused on what the right approach 
is to fix it.

First a bit of background info:

We previously had a multi site radosgw installation, however recently we 
decommissioned the second site. With the radosgw multi-site 
configuration we had 'bucket_index_max_shards = 0'. Since 
decommissioning the second site, I have removed the secondary zonegroup 
and changed 'bucket_index_max_shards' to be 16 for the single primary zone.
All our buckets do not have a 'num_shards' field when running 
'radosgw-admin bucket stats --bucket <bucketname>'
Is this normal ?

Also - I'm finding it difficult to find out exactly what to do with the 
buckets that are affected with 'large omap' (see commands below).
My interpretation of 'search the cluster log' is also listed below.

What do I need to do to with the below buckets get back to an overall 
ceph HEALTH OK state ? :)


# ceph health detail
HEALTH_WARN 2 large omap objects
2 large objects found in pool '.bbp-gva-master.rgw.buckets.index'
Search the cluster log for 'Large omap object found' for more details.

# ceph osd pool get .bbp-gva-master.rgw.buckets.index pg_num
pg_num: 64

# for i in `ceph pg ls-by-pool .bbp-gva-master.rgw.buckets.index | tail 
-n +2 | awk '{print $1}'`; do echo -n "$i: "; ceph pg $i query |grep 
num_large_omap_objects | head -1 | awk '{print $2}'; done | grep ": 1"
137.1b: 1
137.36: 1

# cat buckets
#!/bin/bash
buckets=`radosgw-admin metadata list bucket |grep \" | cut -d\" -f2`
for i in $buckets
do
   id=`radosgw-admin bucket stats --bucket $i |grep \"id\" | cut -d\" -f4`
   pg=`ceph osd map .bbp-gva-master.rgw.buckets.index ${id} | awk 
'{print $11}' | cut -d\( -f2 | cut -d\) -f1`
   echo "$i:$id:$pg"
done
# ./buckets > pglist
# egrep '137.1b|137.36' pglist |wc -l
192

The following doesn't appear to do change anything

# for bucket in `cut -d: -f1 pglist`; do radosgw-admin reshard add 
--bucket $bucket --num-shards 8; done

# radosgw-admin reshard process



-- 
Kind regards,

Ben Morrice

______________________________________________________________________
Ben Morrice | e: ben.morrice at epfl.ch | t: +41-21-693-9670
EPFL / BBP
Biotech Campus
Chemin des Mines 9
1202 Geneva
Switzerland



More information about the ceph-users mailing list