Rich,
What's your failure domain (osd? host? chassis? rack?) and how big is each of
them?
For example I have a failure domain of type rack in one of my clusters with
mostly even rack sizes:
# ceph osd crush rule dump | jq -r '.[].steps'
[
{
"op": "take",
"item": -1,
"item_name": "default~hdd"
},
{
"op": "chooseleaf_firstn",
"num": 0,
"type": "rack"
},
{
"op": "emit"
}
]
[
{
"op": "take",
"item": -35,
"item_name": "default~ssd"
},
{
"op": "chooseleaf_firstn",
"num": 0,
"type": "rack"
},
{
"op": "emit"
}
]
# ceph osd tree |grep rack
-74 160.34999 rack rack1
-73 160.68999 rack rack2
-72 160.68999 rack rack3
If I have a single failure my rack sizes become uneven and the balancer can't
achieve a perfect balance. I even find it'll sometimes get into a loop which
causes the mgr's to basically stop working. Ideally I would like to see each
rack become balanced within the rack instead of across the cluster.
Bryan
> On Dec 10, 2019, at 5:30 PM, Rich Bade <[email protected]> wrote:
>
> Notice: This email is from an external sender.
>
>
>
> I'm finding the same thing. The balancer used to work flawlessly, giving me a
> very even distribution with about 1% variance. Some time between 12.2.7
> (maybe) and 12.2.12 it's stopped working.
> Here's a small selection of my osd's showing 47%-62% spread.
>
> 210 hdd 7.27739 1.00000 7.28TiB 3.43TiB 3.84TiB 47.18 0.74 104
> 211 hdd 7.27739 1.00000 7.28TiB 3.96TiB 3.32TiB 54.39 0.85 118
> 212 hdd 7.27739 1.00000 7.28TiB 4.50TiB 2.77TiB 61.88 0.97 136
> 213 hdd 7.27739 1.00000 7.28TiB 4.06TiB 3.21TiB 55.85 0.87 124
> 214 hdd 7.27739 1.00000 7.28TiB 4.30TiB 2.98TiB 59.05 0.92 130
> 215 hdd 7.27739 1.00000 7.28TiB 4.41TiB 2.87TiB 60.54 0.95 134
> TOTAL 1.26PiB 825TiB 463TiB 64.01
> MIN/MAX VAR: 0.74/1.10 STDDEV: 3.22
>
> $ sudo ceph balancer status
> {
> "active": true,
> "plans": [],
> "mode": "upmap"
> }
>
> I'm happy to add data or test things to get this bug fixed.
>
> Rich
> _______________________________________________
> ceph-users mailing list -- [email protected]
> To unsubscribe send an email to [email protected]
_______________________________________________
ceph-users mailing list -- [email protected]
To unsubscribe send an email to [email protected]
