We have had success using pgremapper[1] for this sort of thing, in both
index and data augments.
1. Set nobackfill, norebalance
2. Add OSDs
3. pgremapper cancel-backfill
4. Unset flags
5. Slowly loop `pgremapper undo-upmaps` at our desired rate, or allow
the balancer to do this work
There's still going to be the hit from peering, especially when you
change the CRUSH map (before even bringing the OSDs in), but in general
peering has been very quick for us for the last few releases, and s3 has
enough overhead where it's not overly noticed. We haven't done multiple
OSDs per device, however, and have plenty CPU power.
[1] https://github.com/digitalocean/pgremapper
On 2023-08-29 17:51, Szabo, Istvan (Agoda) wrote:
Hello,
Is there a way to somehow fine tune the rebalance even further than
basic tuning steps when adding new osds?
Today I've added some osd to the index pool and it generated many slow
ops due to OSD op latency increase + read operation latency increase =
high put get latency.
https://ibb.co/album/9mN6GQ
osd max backfill, max recovery, recovery ops priority are 1.
1 nvme drive has 4 osd, each osd has around 80pg.
The steps how I add the osds:
1. Set norebalance
2. add the osds
3. wait for peering
4. unset rebalance
It takes like 15-20 mins to became normal without interrupting the
rebalance the user traffic.
Thank you,
Istvan
________________________________
This message is confidential and is for the sole use of the intended
recipient(s). It may also be privileged or otherwise protected by
copyright or other legal rules. If you have received it by mistake
please let us know by reply email and delete it from your system. It is
prohibited to copy this message or disclose its content to anyone. Any
confidentiality or privilege is not waived or lost by any mistaken
delivery or unauthorized disclosure of the message. All messages sent
to and from Agoda may be monitored to ensure compliance with company
policies, to protect the company's interests and to remove potential
malware. Electronic messages may be intercepted, amended, lost or
deleted, or contain viruses.
_______________________________________________
ceph-users mailing list -- ceph-users@xxxxxxx
To unsubscribe send an email to ceph-users-leave@xxxxxxx
_______________________________________________
ceph-users mailing list -- ceph-users@xxxxxxx
To unsubscribe send an email to ceph-users-leave@xxxxxxx
