Hello
I have a cluster, (Nautilus 14.2.4) where one pool I'd like to keep on a
dedicated OSDs. So I setup a rule that covers *3* dedicated OSDs (using
device classes) and assigned it to pool with replication factor *3*. Only
10% PGs were assigned and rebalanced, where rest of them stuck in
*undersized* state.
What mechanism prevents CRUSH algorithm to assign the same set of OSDs to
all PGs in a pool? How can I control it?
Jacek
--
Jacek Suchenia
jacek.suchenia(a)gmail.com
Hi
We have some oldish servers with ssds - all on 25gbit nics. R815 AMD - 2,4ghz+
Is there significant performance benefits in moving to a new NVMe based, new cpus?
+20% IOPs? + 50% IOPs?
Jesper
Sent from myMail for iOS
Hi Ceph Community,
Wondering what experiences good/bad you have with EC pools for iops intensive workloads (IE: 4Kish random IO from things like VMWare ESXi). I realize that EC pools are a tradeoff between more usable capacity, and having larger latency/lower iops, but in my testing the tradeoff for small IO seems to be much worse than I had anticipated.
On an all flash 3x replicated pool we’re seeing 45k random read, and 35k random write iops testing with fio on a client living on an iSCSI LUN presented to an ESXi host. Average latencies for these ops are 4.2ms, and 5.5ms, which is respectable at an io depth of 32.
Take this same setup with an EC pool (k=2, m=1, tested with both ISA and jerasure, ISA does give better performance for our use case) and we see 30k random read, and 16k random write iops. Random reads see 6.5ms average, while random writes suffer with 12ms average.
Are others using EC pools seeing similar hits to random writes with small IOs? Any way to improve this?
Thanks,
Anthony
cephfs is considered to be stable. I am using it with only one mds for
2-3 years in low load environment without any serious issues.
-----Original Message-----
Sent: 17 February 2020 16:07
To: ceph-users
Subject: [ceph-users]
=?eucgb2312_cn?q?=D7=AA=B7=A2=3A_Causual_survey_on_the_successful_usage_
of_CephFS_on_production?=
Dear Folks,
I am planning a files systems project and CephFS is under serious
consideration. But browsering on the web i found some negative comments
on CephFS about stability and losing data. So i would like to learn more
on the latest development about CephFS statbility, either a successful
story or a failure
Could you please speak out, if you have recently successfully deployed
CephFS on production, and which version? any statbility issues with
CephFS?
thanks a lot,
samuel
huxiaoyu(a)horebdata.cn
_______________________________________________
ceph-users mailing list -- ceph-users(a)ceph.io To unsubscribe send an
email to ceph-users-leave(a)ceph.io
Dear Folks,
I am planning a files systems project and CephFS is under serious consideration. But browsering on the web i found some negative comments on CephFS about stability and losing data. So i would like to learn more on the latest development about CephFS statbility, either a successful story or a failure
Could you please speak out, if you have recently successfully deployed CephFS on production, and which version? any statbility issues with CephFS?
thanks a lot,
samuel
huxiaoyu(a)horebdata.cn
Hi all,
Is there a way to estimate how much storage space is required for
CephFS metadata given an expected number of files in the filesystem?
thx
Frank
On Fri, Feb 14, 2020 at 3:19 PM Marc Roos <M.Roos(a)f1-outsourcing.eu> wrote:
>
>
> I have default centos7 setup with nautilus. I have been asked to install
> 5.5 to check a 'bug'. Where should I get this from? I read that the
> elrepo kernel is not compiled like rhel.
Hi Marc,
I'm not sure what you mean by "not compiled like RHEL".
Follow [1] to enable elrepo repository and then:
$ sudo yum --enablerepo=elrepo-kernel install kernel-ml
"ml" stands for "mainline". This is the most recent mainline stable
kernel from kernel.org, it should get you 5.5.3.
[1] http://elrepo.org/tiki/tiki-index.php
Thanks,
Ilya
Hey, I've been running a ceph cluster of arm64 SOCs on Luminous for the
past year or so, with no major problems. I recently upgraded to 14.2.7, and
the stability of the cluster immediately suffered. Seemed like any mon
activity was subject to long pauses, and the cluster would hang frequently.
Looking at ceph -s, it appeared the cluster was electing new masters very
frequently - masters didn't seem to last longer than about 1-2 minutes.
Looking further at the mons, two out of three of which are running on
relatively slow-performing SD card storage on these SoCs, I saw them
absolutely maxing out the root device's IO with writes. Logs show rocksdb
constantly running compactions. I temporarily moved these mons to devices
with better performing IO (but shouldn't be mons as they're also cephfs
clients) and saw a sustained write rate of ~50MB/s. This seems pretty
excessive, and is at least an order of magnitude higher than anything I saw
when running Luminous. Not to mention this isn't really nice for SSD
lifespan.
As downgrading is not an option here, is there anything I can look to to
figure out what exactly the mons are doing and how to prevent such heavy
load? I seem to remember some bug related to telemetry, but can't find it
on this list..
Hello Everyone,
i've run into problems with placement groups.
We have a 12 Host Ceph-Cluster with 408 OSDs (hdd and ssd).
I create a replicated pool with a large pg_num (16384). No problems
everything works.
If I do this with an erasure pool, i get a warning which is fixable by
extending the max_pgs_per_osd and afterwards i get a health_warn in the
ceph status due too few PGs.
Checking the pg_num after pool creation, it is capped at 1024.
I'm stuck at this point. Maybe i did something fundamentally wrong?
To illustrate my steps I tried to summarize everything in a small example:
# ceph -v
ceph version 14.2.7 (fb8e34a687d76cd3bd45c2a0fb445432ab69b4ff) nautilus
(stable)
# ceph osd erasure-code-profile get myerasurehdd
crush-device-class=hdd
crush-failure-domain=host
crush-root=default
jerasure-per-chunk-alignment=false
k=7
m=5
plugin=jerasure
technique=reed_sol_van
w=8
# ceph osd crush rule dump sas_rule
{
"rule_id": 0,
"rule_name": "sas_rule",
"ruleset": 0,
"type": 3,
"min_size": 1,
"max_size": 12,
"steps": [
{
"op": "take",
"item": -2,
"item_name": "default~hdd"
},
{
"op": "chooseleaf_firstn",
"num": 0,
"type": "rack"
},
{
"op": "emit"
}
]
}
# ceph osd pool create sas-pool 16384 16384 erasure myerasurehdd sas_rule
Error ERANGE: pg_num 16384 size 12 would mean 196704 total pgs, which
exceeds max 102000 (mon_max_pg_per_osd 250 * num_in_osds 408)
# ceph tell mon.\* injectargs '--mon-max-pg-per-osd=500'
mon.ceph-fs01: injectargs:mon_max_pg_per_osd = '500' (not observed,
change may require restart)
mon.ceph-fs05: injectargs:mon_max_pg_per_osd = '500' (not observed,
change may require restart)
mon.ceph-fs09: injectargs:mon_max_pg_per_osd = '500' (not observed,
change may require restart)
ceph-fs01:/opt/ceph-setup# ceph osd pool create sas-pool 16384 16384
erasure myerasurehdd sas_rule
pool 'sas-pool' created
# ceph -s
cluster:
id: b9471b57-95a2-4e58-8f69-b5e6048bea7c
health: HEALTH_WARN
Reduced data availability: 1024 pgs incomplete
too few PGs per OSD (7 < min 30)
# ceph osd pool get sas-pool pg_num
pg_num: 1024
Best regards,
Gunnar