!oPtkUSbKZXLkzMcZBn:matrix.org

#ceph

5 Members
3 Servers

You have reached the beginning of time (for this room).


Timestamp Message
16 Sep 2017
06:58:18@leonardovaz:matrix.org@leonardovaz:matrix.org changed the history visibility to "world_readable" from "shared".
06:58:18@leonardovaz:matrix.org@leonardovaz:matrix.org changed the join rule to "public" from "invite".
06:58:56@leonardovaz:matrix.org@leonardovaz:matrix.org invited @oftc-irc:matrix.org@oftc-irc:matrix.org.
06:58:56@oftc-irc:matrix.org@oftc-irc:matrix.org joined the room.
06:58:57@leonardovaz:matrix.org@leonardovaz:matrix.orgchanged room power levels.
07:08:57@oftc-irc:matrix.org@oftc-irc:matrix.org left the room.
18 Sep 2017
11:56:30@bazzspeed:matrix.orgbazzspeed joined the room.
11:59:15@bazzspeed:matrix.orgbazzspeedHi everyone. Can anyone give me a example of how long a recovery/rebalance would take with Ceph ?
12:02:26@bazzspeed:matrix.orgbazzspeedWe're trialing Luminous 12.2. Bluestore OSDs, sata+bcache + ssd block.db. Performance is great, rebalancing/removing osds is slow as molasses
30 Sep 2017
19:12:38@sakhinov:matrix.org@sakhinov:matrix.org joined the room.
19:18:48@sakhinov:matrix.org@sakhinov:matrix.org left the room.
2 May 2018
11:07:13@dalgaaf:matrix.orgdalgaaf joined the room.
5 Jun 2018
11:25:58@ericbaum:matrix.orgericbaum joined the room.
28 Jul 2018
17:54:48@briantopping:matrix.org@briantopping:matrix.org joined the room.
17:55:36@briantopping:matrix.org@briantopping:matrix.org left the room.
14 Aug 2018
18:41:20@ronny:synapse.dst.fjit.noRonny Aasen joined the room.
18:48:51@ronny:synapse.dst.fjit.noRonny Aasen
In reply to @bazzspeed:matrix.org
Hi everyone. Can anyone give me a example of how long a recovery/rebalance would take with Ceph ?

there are some variables that can control the backfill/recovery speed. eg osd recovery sleep hdd have a delay for each recovery op, to avoid lagging the production traffic with the recovery traffic. ; osd max backfills say something about how many backfills a given osd can deal with at the time.

also, if a disk dies, all the other disks shuffle data around, so more osd/hosts = faster. but when a single osd is restored to the cluster, it is the one disk that needs to write all the data, so in that case it will be the limiting factor.

18:49:45@ronny:synapse.dst.fjit.noRonny Aasenmatrix is perhaps a bit slow, so you need patience :P
23:32:27@leonardovaz:matrix.org@leonardovaz:matrix.orgchanged room power levels.
23:33:12@leonardovaz:matrix.org@leonardovaz:matrix.orgchanged room power levels.
23:33:23@leonardovaz:matrix.org@leonardovaz:matrix.org left the room.
22 Nov 2018
12:01:39@olivier:no.dustinthe.netolivier joined the room.

There are no newer messages yet.


Back to Room List