ceph / ceph-medic

find common issues in ceph clusters
MIT License
22 stars 18 forks source link

checks: check for bucket indexes needing index sharding in RGW #66

Open haklein opened 7 years ago

haklein commented 7 years ago

We recommend to shard the RGW bucket index above 100k objects. The command would be "radosgw-admin bucket limit check" on one of the RGW nodes. This will print out a JSON with the field "objects_per_shard" that we're interested in. A warning should be spit out if that exceeds 100k.

example output:

# radosgw-admin bucket limit check
[
    {
        "user_id": "hari",
        "buckets": [
            {
                "bucket": "mybucket",
                "tenant": "",
                "num_objects": 610001,
                "num_shards": 0,
                "objects_per_shard": 610001,
                "fill_status": "OVER 100.000000%"
            },
            {
                "bucket": "mybucket2",
                "tenant": "",
                "num_objects": 2,
                "num_shards": 0,
                "objects_per_shard": 2,
                "fill_status": "OK"
            },
            {
                "bucket": "test",
                "tenant": "",
                "num_objects": 0,
                "num_shards": 0,
                "objects_per_shard": 0,
                "fill_status": "OK"
            }
        ]
    }
]
concubidated commented 7 years ago

Isn't that the information the fill_status is giving you? When you are over the rgw_safe_max_objects_per_shard (which is 100*1024 objects by default)

alfredodeza commented 5 years ago

when I tried running this on the CLI it hang for ever, and there aren't any possible timeout flags we can pass. Do you guys know if there is any other way to capture this information?