Well, if that issue occurs it will be at the beginning of the
recovery, so you may not notice it until you get inactive PGs. We hit
that limit when we rebuilt all OSDs on one server with many EC chunks.
Setting osd_max_pg_per_osd_hard_ratio to 5 (default 3) helped avoid
inactive PGs for all other nodes, so we leave it like that in case a
server goes down unintentionally.
Zitat von "Szabo, Istvan (Agoda)" <Istvan.Szabo@xxxxxxxxx>:
Good to know thank you, so in that case during recovery it worth to
increase those values right?
Istvan Szabo
Senior Infrastructure Engineer
---------------------------------------------------
Agoda Services Co., Ltd.
e: istvan.szabo@xxxxxxxxx
---------------------------------------------------
-----Original Message-----
From: Eugen Block <eblock@xxxxxx>
Sent: Friday, September 23, 2022 1:19 PM
To: ceph-users@xxxxxxx
Subject: Re: Any disadvantage to go above the 100pg/osd
or 4osd/disk?
Email received from the internet. If in doubt, don't click any link
nor open any attachment !
________________________________
Hi,
I can't speak from the developers perspective, but we discussed this
just recently intenally and with a customer. We doubled the number
of PGs on one of our customer's data pools from around 100 to 200
PGs/OSD (HDDs with rocksDB on SSDs). We're still waiting for the
final conclusion if the performance has increased or not, but it
seems to work as expected. We probably would double it again if the
PG size/objects per PG would affect the performance again. You just
need to be aware of the mon_max_pg_per_osd and
osd_max_pg_per_osd_hard_ratio configs in case of recovery. Otherwise
we don't see any real issue with 200 or 400 PGs/OSD if the nodes can
handle it.
Regards,
Eugen
Zitat von "Szabo, Istvan (Agoda)" <Istvan.Szabo@xxxxxxxxx>:
Hi,
My question is, is there any technical limit to have 8osd/ssd and on
each of them 100pg if the memory and cpu resource available (8gb
memory/osd and 96vcore)?
The iops and bandwidth on the disks are very low so I don’t see any
issue to go with this.
In my cluster I’m using 15.3TB ssds. We have more than 2 billions of
objects in each of the 3 clusters.
The bottleneck is the pg/osd so last time when my serious issue solved
the solution was to bump the pg-s of the data pool the allowed maximum
with 4:2 ec.
I’m curious of the developers opinion also.
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
________________________________
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
