Hi people,
I currently try to add ~30 OSDs to our cluster and wanted to use the
gentle-rerweight script for that.
I use ceph-colume lvm prepare --data /dev/sdX to create the osd and want to
start it without weighting it in.
systemctl start ceph-osd@OSD starts the OSD with full weight.
Is this possible?
--
Die Selbsthilfegruppe "UTF-8-Probleme" trifft sich diesmal abweichend im
groüen Saal.
Hi all.
I have some questions about how MDS cluster works when crash or operation
failure occur.
1. I read ceph documentation and code that each MDS has its own journal and
some directory operations like rename use distributed transaction mechanism
with ceph-defined events (e.g. EPeerUpdate). What I wonder is when one of
MDSs is crashed and then reboots, if some metadata in other MDS's journal
is not flushed yet, is rebooted MDS read stale metadata?
2. While rename operation is performing on 2 MDSs and one of them is
crashed, Does the other MDS rollback modified in-memory metadata ?
Thanks,
kyujin
Hello,
Our podman backed cluster currently is on version 15.2.8, the ceph and
cephadm packages on the hosts are 15.2.10.
When we check the package version with 'ceph orch upgrade check
--ceph-version 15.2.10' it tells us that it failed to pull the image on
a host.
Inspecting the log we see that it pulls the image successfully but then
fails inspecting it[0].
I can run the command manually without problems when I remove the 'json
' from it.
So far I was unable to find the issue myself, maybe someone here knows
why this happens and how to fix it.
Kind regards,
Julian Fölsch
[0] https://agdsn.me/~paktosan/ceph/octopus_upgrade/cephadm_log.txt
--
Julian Fölsch
Arbeitsgemeinschaft Dresdner Studentennetz (AG DSN)
Telefon: +49 351 271816 69
Mobil: +49 152 22915871
Fax: +49 351 46469685
Email: julian.foelsch(a)agdsn.de
Studierendenrat der TU Dresden
Helmholtzstr. 10
01069 Dresden
"rbd status" will tell you which IP addresses have the image open.
"rbd lock ls" will show you who owns the lock.
On Wed, Mar 24, 2021 at 9:07 AM Edgelong Voodu <1070443499cs(a)gmail.com> wrote:
>
> Hi,
> As info mentioned in the log file
> There still a lock owner alive. The question is how to find out the alive owner and what is the root cause about this. Why cannot acquire lock to the owner?
> Thank you so much
>
> Jason Dillaman <jdillama(a)redhat.com>于2021年3月24日 周三20:55写道：
>>
>> It sounds like this is a non-primary mirrored image, which means it's
>> read-only and cannot be modified. A quick "rbd info" will tell you the
>> mirror state. Instead, you would need to force-promote it to primary
>> via "rbd mirror image promote --force" before attempting to modify the
>> image.
>>
>> On Wed, Mar 24, 2021 at 3:44 AM Edgelong Voodu <1070443499cs(a)gmail.com> wrote:
>> >
>> > hi , dillaman:
>> > i was confused nowadays, when i execute the rbd feature disable <image> journaling,
>> > the result of this often shows me :
>> > rbd: failed to update image features: (30) Read-only file system
>> > i use the rbd-mirror for replicate data between two clouster. i thought the rbd-mirror could be the reason. but i have no idea about it. could you give some advise ?
>> > some log when execute rbd command also send you.
>>
>>
>>
>> --
>> Jason
>>
> --
> 发自移动版 Gmail
--
Jason
That means you have another librbd client that "owns" the image that
you will need to shut down before you can make that change.
On Wed, Mar 24, 2021 at 9:01 AM Edgelong Voodu <1070443499cs(a)gmail.com> wrote:
>
> hi，
> Thank you for the clarification.
> It’s already promoted and not mirrored ,a primary image
> I have run rbd —debug-rbd=30 and collect a log file
> Which shows locker owner still alive and unable get the lock, return -EAGAIN
> I’ll send you log later
>
> Thank you so much
>
>
> Jason Dillaman <jdillama(a)redhat.com>于2021年3月24日 周三20:55写道：
>>
>> It sounds like this is a non-primary mirrored image, which means it's
>> read-only and cannot be modified. A quick "rbd info" will tell you the
>> mirror state. Instead, you would need to force-promote it to primary
>> via "rbd mirror image promote --force" before attempting to modify the
>> image.
>>
>> On Wed, Mar 24, 2021 at 3:44 AM Edgelong Voodu <1070443499cs(a)gmail.com> wrote:
>> >
>> > hi , dillaman:
>> > i was confused nowadays, when i execute the rbd feature disable <image> journaling,
>> > the result of this often shows me :
>> > rbd: failed to update image features: (30) Read-only file system
>> > i use the rbd-mirror for replicate data between two clouster. i thought the rbd-mirror could be the reason. but i have no idea about it. could you give some advise ?
>> > some log when execute rbd command also send you.
>>
>>
>>
>> --
>> Jason
>>
> --
> 发自移动版 Gmail
--
Jason
It sounds like this is a non-primary mirrored image, which means it's
read-only and cannot be modified. A quick "rbd info" will tell you the
mirror state. Instead, you would need to force-promote it to primary
via "rbd mirror image promote --force" before attempting to modify the
image.
On Wed, Mar 24, 2021 at 3:44 AM Edgelong Voodu <1070443499cs(a)gmail.com> wrote:
>
> hi , dillaman:
> i was confused nowadays, when i execute the rbd feature disable <image> journaling,
> the result of this often shows me :
> rbd: failed to update image features: (30) Read-only file system
> i use the rbd-mirror for replicate data between two clouster. i thought the rbd-mirror could be the reason. but i have no idea about it. could you give some advise ?
> some log when execute rbd command also send you.
--
Jason
Hello people,
I am still struggeling with the balancer
(https://www.mail-archive.com/ceph-users@ceph.io/msg09124.html)
Now I've read some more and might think that I do not have enough PGs.
Currently I have 84OSDs and 1024PGs for the main pool (3008 total). I
have the autoscaler enabled, but I doesn't tell me to increase the
PGs.
What do you think?
--
Die Selbsthilfegruppe "UTF-8-Probleme" trifft sich diesmal abweichend
im groüen Saal.
Hi all,
We have a few subdirs with an rctime in the future.
# getfattr -n ceph.dir.rctime session
# file: session
ceph.dir.rctime="2576387188.090"
I can't find any subdir or item in that directory with that rctime, so
I presume that there was previously a file and that rctime cannot go
backwards [1]
Is there any way to fix these rctimes so they show the latest ctime of
the subtree?
Also -- are we still relying on the client clock to set the rctime /
ctime of a file? Would it make sense to limit ctime/rctime for any
update to the current time on the MDS ?
Best Regards,
Dan
[1] https://github.com/ceph/ceph/pull/24023/commits/920ef964311a61fcc6c0d6671b7…
Hi everyone:
I posted to the list on Friday morning (UK time), but apparently my email
is still in moderation (I have an email from the list bot telling me that
it's held for moderation but no updates).
Since this is a bit urgent - we have ~3PB of storage offline - I'm posting
again.
To save retyping the whole thing, I will direct you to a copy of the email
I wrote on Friday:
http://aoanla.pythonanywhere.com/Logs/EmailToCephUsers.txt
(Since that was sent, we did successfully add big SSDs to the MON hosts so
they don't fill up their disks with store.db s).
I would appreciate any advice - assuming this also doesn't get stuck in
moderation queues.
--
Sam Skipsey (he/him, they/them)
Hello,
Based on other discussions in this list I have concluded that I need to add
NVMe to my OSD nodes and expand the NVMe (DB/WAL) for each OSD. Is there a
way to do this without destroying and rebuilding each OSD (after
safe removal from the cluster, of course)? Is there a way to use
ceph-bluestore-tool for this? Is it as simple as lvextend?
Why more NVMe? Frequent DB spillovers, and the recommendation that the
NVMe should be 40GB for every TB of HDD. When I did my initial setup I
thought that 124GB of NVMe for a 12TB HDD would be sufficient, but by the
above metric it should be more like 480GB of NVMe.
Thanks.
-Dave
--
Dave Hall
Binghamton University
kdhall(a)binghamton.edu
