I know a bit the work-arounds for manually editing the crush map. I just think
this is not the best way to get acquainted to with a new ceph cluster. I would
make these hdd,nvme,ssd classes available directly.
> You could decompile the crushmap, add a dummy OSD (with a non-existing
> ID) with y
Hi yall,
Am 29.11.24 um 08:51 schrieb Eugen Block:
rule testrule {
id 6
type erasure
step set_chooseleaf_tries 5
step set_choose_tries 100
step take default class test
step chooseleaf indep 0 type host
step emit
}
Does anyone know
I see the same on a newly deployed 17.2.8 cluster.
all empty perf values.
Den tors 28 nov. 2024 kl 23:45 skrev Marc :
>
>
>
> My ceph osd perf are all 0, do I need to enable module for this?
> osd_perf_query? Where should I find this in manuals? Or do I just need to
> wait?
>
>
> [@ target]# cep
Hi,
> I mean that even if I pin all top dirs (of course without repinning
on next levels) to rank 1 - I see some amount of Reqs on rank 1.
I assume you mean if you pin all top dirs to rank 0, you still see IO on
rank 1? I still can't reproduce that, I waited for 15 minutes or so with
rank 1
I tried to get the counters, then I was pointed to enabling the module:
# ceph osd perf
osd commit_latency(ms) apply_latency(ms)
11 0 0
8 0 0
6 0 0
1 0
Hi,
just to clarify, the public network (in your case 192.168.1.0/24) is
basically for all traffic if you don't have a cluster network defined.
If you do, it will be only used for OSD to OSD communication for
replication, recovery and heartbeats [0]. The communication to the
MONs will hap
Hi Istvan,
Did the PG split involved using more OSDs than before? If so then increasing
these values (apart from the sleep) should not have a negative impact on
clients I/O compared to before the split and should accelerate the whole
process.
Did you reshard the buckets as discussed in the oth
Hi Michel,
This is correct. Don't see anything wrong with that.
Regards,
Frédéric.
- Le 28 Nov 24, à 8:16, Michel Niyoyita mico...@gmail.com a écrit :
> Hello team ,
>
> I am creating new cluster which will be created using CEPHADM, I will use
> 192.168.1.0/24 as public network and 10.10.9
Increased from 9 servers to 11 so let's say 20% capacity and performance added.
This is a different cluster purely rbd.
(For the other topic can't be resharded because in multisite it will disappear
all the data disappear on remote site, need to create new bucket and migrate
data first to a hig
- Le 29 Nov 24, à 11:11, Istvan Szabo, Agoda a
écrit :
> Increased from 9 servers to 11 so let's say 20% capacity and performance
> added.
> This is a different cluster purely rbd.
I see, so big objects. You might want to increase osd_max_trimming_pgs and
eventually osd_pg_max_concurren
Hello,
We are using Ceph as a storage backend for Openstack (Cinder, Nova,
Glance, Manila) and we are replacing old hardware hosting Ceph monitors
(MON,MGR,MDS) to new ones.
I have already added the new ones in production, monitors successfully
joined the quorum and new MGR/MDS are standby.
Which questions do you have? When I first started to deal with crush
rules I was overwhelmed, but with a bit of practice and trial & error
you're going to figure it out.
Maybe this helps a bit (inline comments):
id 6 -> self explanatory
type erasure -> self explanatory
step set_chooseleaf_
Thank you all
On Fri, Nov 29, 2024 at 12:05 PM Frédéric Nass <
frederic.n...@univ-lorraine.fr> wrote:
> Hi Michel,
>
> This is correct. Don't see anything wrong with that.
>
> Regards,
> Frédéric.
>
> - Le 28 Nov 24, à 8:16, Michel Niyoyita mico...@gmail.com a écrit :
>
> > Hello team ,
> >
Hi,
we are already running the "default" rgw pool with some users.
Data is stored in pool:
pool 9 'default.rgw.buckets.data' erasure profile ec-32-profile size 5
min_size 4 crush_rule 1 object_hash rjenkins pg_num 512 pgp_num 512
autoscale_mode on last_change 309346 lfor 0/127784/214408 flags
has
Hi Frederic,
>My question was more about why bluefs would still fail to allocate 4k
chunks after being allowed to do so by
https://tracker.ceph.com/issues/53466 (John's case with v17.2.6 actually)
My hypothesis is that it's facing real "no-space" case not the one I
explained above. But again
Hello,
We have an Alma8.9 (version 4 kernel) quincy (17.2.7) CephFS cluster
with spinners for our bulk data and SSDs for the metadata where we have
a single unfound object in the bulk pool:
[root@ceph-n30 ~]# ceph -s
cluster:
id: fa7cf62b-e261-49cd-b00e-383c36b79ef3
> we are already running the "default" rgw pool with some users.
>
> Data is stored in pool:
> pool 9 'default.rgw.buckets.data' erasure profile ec-32-profile size 5
> min_size 4 crush_rule 1 object_hash rjenkins pg_num 512 pgp_num 512
> autoscale_mode on last_change 309346 lfor 0/127784/214408 fla
The reshard topic is running on quincy 17.2.7, but tested today the reshard,
objects gone.
Istvan
From: Frédéric Nass
Sent: Friday, November 29, 2024 5:17:27 PM
To: Szabo, Istvan (Agoda)
Cc: Ceph Users
Subject: Re: [ceph-users] Snaptriming speed degrade with pg
On Fri, Nov 29, 2024, 5:33 AM Adrien Georget
wrote:
> Hello,
>
> We are using Ceph as a storage backend for Openstack (Cinder, Nova,
> Glance, Manila) and we are replacing old hardware hosting Ceph monitors
> (MON,MGR,MDS) to new ones.
> I have already added the new ones in production, monitors s
Ahoi Eugen,
Am 29.11.24 um 11:31 schrieb Eugen Block:
step set_chooseleaf_tries 5 -> stick to defaults, usually works (number
of max attempts to find suitable OSDs)
Why do we need more than one attempt to find an OSD? Why is the result
different if we walk through a rule more than once?
s
Confirming Tyler's description, we had to do lots of database
manipulation in order to get the new IPs into the connection parameters.
Since you already added the new monitors, there's not much else you can
do. But I would have suggested to rather reinstall the MONs instead of
adding new ones a
Syntax errors on the config?
Try to start manually with -x to be sure
what does the journal log has to say?
https://github.com/nfs-ganesha/nfs-ganesha/issues/730
Release notes:
https://github.com/nfs-ganesha/nfs-ganesha/wiki/ReleaseNotes_5
On Thu, 28 Nov 2024 at 12:35, Marc wrote:
> >
> > In
This is new, and on in the default config. I am currently running without it
just fine. I guess it is storing info for when you have other cluster nodes.
RecoveryBackend = rados_ng;
>
>
> Syntax errors on the config?
> Try to start manually with -x to be sure
> what does the journal log has
- Le 29 Nov 24, à 12:03, Igor Fedotov a écrit :
> Hi Frederic,
>>My question was more about why bluefs would still fail to allocate 4k chunks
>>after being allowed to do so by [ https://tracker.ceph.com/issues/53466 |
> >https://tracker.ceph.com/issues/53466 ] (John's case with v17.2.6 actu
Andre,
see responses inline.
Zitat von Andre Tann :
Ahoi Eugen,
Am 29.11.24 um 11:31 schrieb Eugen Block:
step set_chooseleaf_tries 5 -> stick to defaults, usually works
(number of max attempts to find suitable OSDs)
Why do we need more than one attempt to find an OSD? Why is the
resul
Hi,the traditional solution is to deny anything but deletions and either write the transaction log to another device or even Filesystem or add support deletions without a transaction log together with a force switch.Regards --martinAm 29.11.2024 12:03 schrieb Igor Fedotov :Hi Frederic,
>My ques
Hi everyone,
Stupid questionafter some test I was able to dump a user caps with
ceph auth get --format json
but I'wasn't able to find the other way, something
ceph auth add fubar.json
Is they are any way to add a user (without given a key and with a key).
Regards
--
Albert SHIH 🦫
Absolutely.
You define a placement target and storage class in the zone / zonegroup, commit
the period, and create/modify the users. New buckets they create will then go
to the secondary storage class. Clients can also specify a storage class in
their request headers, and you can also force t
For me, the perf stats are non-zero only for those OSD's that are
currently writing. The others that are idle/reading show zero. (I
have a recovery going on, lots of PG's being moved to two new disks.
The two new ones have stats, all the others show zero.)
-- linas
On Fri, Nov 29, 2024 at 3:10 A
Or, just reassign one existing OSD to the new class.
>> Note that testing this rule with crushtool won't work here since the
>> fake OSD isn't assigned to a hosts.
> But what's the point in having a rule without the corresponding
> devices? You won't be able to create a pool with that rule anywa
>
> Or, just reassign one existing OSD to the new class.
>
> >> Note that testing this rule with crushtool won't work here since the
> >> fake OSD isn't assigned to a hosts.
>
> > But what's the point in having a rule without the corresponding
> > devices? You won't be able to create a pool with
Hi Anthony,
No we dont have any hours set - scrubbing happens at all times. The only thing
we changed from default and kept was increasing osd_max_scrubs to 5 to try and
catch up. Other than that it was just expanding the window of scrubbing
intervals as pgs not deep-scrubbed in time alerts kept
Hi all, sveikas,
Thanks everyone for the tips and trying to help out!
I've eventually raised a bug tracker for the case to get more developers
involved: https://tracker.ceph.com/issues/69078
We tried decreasing osd_scrub_chunk_max from 25 to 15 as per Frédéric
suggestion, but unfortunately did
Hi Frédéric,
Thanks for pointing out! I see we have 25 set for osd_scrub_chunk_max
(default).
I will try reducing it back to 15 and see if that helps this case.
Regards,
Laimis J.
___
ceph-users mailing list -- ceph-users@ceph.io
To unsubscribe send
Trying to create the Block and Block.DB devices for BlueStore. Within the
cephadm shell able to run the vgcreate but get the following errors and we
cannot see the vg device in /dev so doesn't seem to actually create the vg, but
vgs within the cephadm shell shows it, but vgs at the o/s level doe
35 matches
Mail list logo
