Re: Upgrading cephadm cluster
Ok, thanks for the confirmation. Just as a followup, our Ceph nodes are running on Debian + Docker -- currently Debian 12.4. I assume that installing any OS-level updates should not affect anything Ceph-wise, as long as there are no issues with Docker? In other words, upgrading the underlying OS and Ceph are basically two very separate(d) things? Regards, Alan On Tue, 2025-07-22 at 18:34 -0400, Anthony D'Atri wrote:
Yes.
On Jul 22, 2025, at 6:33 PM, Alan Murrell <Alan@t-net.ca> wrote:
Hello,
I just wanted to check quickly, but feel I am being a little *too* cautious...
We have a Ceph cluster (installed via cephadm) that is currently version 18.2.2. The "Upgrade" tab gives me an option to upgrade to 18.2.7 (its default selection), or I can upgrade to another version where I can choose "earlier" upgrades (from 18.2.3 to the offered 18.2.7)
Is it generally OK and safe to "jump" upgrades like this (i.e. go straight from 18.2.2 straight to 18.2.7) or would it be recommended to upgrade in order (18.2.2 -> 18.2.3 -> 18.2.4 -> etc.)?
I assume once I am on latest 18.2.7, there might be an option to upgrade to 19.2.x (not that I would necessarily want to do that straight away)?
Thanks! :-)
Regards,
Alan _______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io
Mostly. This is one benefit of containerization. Note that downgrading isn’t tested and should be avoided. In some cases a downgrade within the same major release will work, but not always. Upgrading to a +2 release is supported, e.g, from Quincy straight to Squid. Ideally from the latest maintenance release of the existing deployment.
On Jul 22, 2025, at 6:41 PM, Alan Murrell <Alan@t-net.ca> wrote:
Ok, thanks for the confirmation.
Just as a followup, our Ceph nodes are running on Debian + Docker -- currently Debian 12.4. I assume that installing any OS-level updates should not affect anything Ceph-wise, as long as there are no issues with Docker? In other words, upgrading the underlying OS and Ceph are basically two very separate(d) things?
Regards,
Alan
On Tue, 2025-07-22 at 18:34 -0400, Anthony D'Atri wrote: Yes.
On Jul 22, 2025, at 6:33 PM, Alan Murrell <Alan@t-net.ca> wrote:
Hello,
I just wanted to check quickly, but feel I am being a little *too* cautious...
We have a Ceph cluster (installed via cephadm) that is currently version 18.2.2. The "Upgrade" tab gives me an option to upgrade to 18.2.7 (its default selection), or I can upgrade to another version where I can choose "earlier" upgrades (from 18.2.3 to the offered 18.2.7)
Is it generally OK and safe to "jump" upgrades like this (i.e. go straight from 18.2.2 straight to 18.2.7) or would it be recommended to upgrade in order (18.2.2 -> 18.2.3 -> 18.2.4 -> etc.)?
I assume once I am on latest 18.2.7, there might be an option to upgrade to 19.2.x (not that I would necessarily want to do that straight away)?
Thanks! :-)
Regards,
Alan _______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io
ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io
"In other words, upgrading the underlying OS and Ceph are basically two very separate(d) things?" Unless you have to reboot a node after OS upgrade. In this case send a node into maintenance mode (ceph orch host maintenance enter <host> --force) before rebooting. Sent via the Samsung Galaxy S25 Ultra, an AT&T 5G smartphone Get Outlook for Android<https://aka.ms/AAb9ysg> ________________________________ From: Alan Murrell <alan@t-net.ca> Sent: Tuesday, July 22, 2025 6:39:04 PM To: ceph-users@ceph.io <ceph-users@ceph.io> Subject: [ceph-users] Re: Upgrading cephadm cluster Ok, thanks for the confirmation. Just as a followup, our Ceph nodes are running on Debian + Docker -- currently Debian 12.4. I assume that installing any OS-level updates should not affect anything Ceph-wise, as long as there are no issues with Docker? In other words, upgrading the underlying OS and Ceph are basically two very separate(d) things? Regards, Alan On Tue, 2025-07-22 at 18:34 -0400, Anthony D'Atri wrote:
Yes.
On Jul 22, 2025, at 6:33 PM, Alan Murrell <Alan@t-net.ca> wrote:
Hello,
I just wanted to check quickly, but feel I am being a little *too* cautious...
We have a Ceph cluster (installed via cephadm) that is currently version 18.2.2. The "Upgrade" tab gives me an option to upgrade to 18.2.7 (its default selection), or I can upgrade to another version where I can choose "earlier" upgrades (from 18.2.3 to the offered 18.2.7)
Is it generally OK and safe to "jump" upgrades like this (i.e. go straight from 18.2.2 straight to 18.2.7) or would it be recommended to upgrade in order (18.2.2 -> 18.2.3 -> 18.2.4 -> etc.)?
I assume once I am on latest 18.2.7, there might be an option to upgrade to 19.2.x (not that I would necessarily want to do that straight away)?
Thanks! :-)
Regards,
Alan _______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io
ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io
Alan, I'd say yes OS and Ceph upgrades are two very separated operations. Thus I personally advise (it's what I'm doing) not to do them at the same time. It makes easier to identify the culprit in case of problems.I generally upgrade the OS before Ceph. Best regards, Michel Sent from my mobile Le 23 juillet 2025 01:46:25 Vladimir Sigunov <vladimir.sigunov@gmail.com> a écrit :
"In other words, upgrading the underlying OS and Ceph are basically two very separate(d) things?"
Unless you have to reboot a node after OS upgrade. In this case send a node into maintenance mode (ceph orch host maintenance enter <host> --force) before rebooting.
Sent via the Samsung Galaxy S25 Ultra, an AT&T 5G smartphone Get Outlook for Android<https://aka.ms/AAb9ysg> ________________________________ From: Alan Murrell <alan@t-net.ca> Sent: Tuesday, July 22, 2025 6:39:04 PM To: ceph-users@ceph.io <ceph-users@ceph.io> Subject: [ceph-users] Re: Upgrading cephadm cluster
Ok, thanks for the confirmation.
Just as a followup, our Ceph nodes are running on Debian + Docker -- currently Debian 12.4. I assume that installing any OS-level updates should not affect anything Ceph-wise, as long as there are no issues with Docker? In other words, upgrading the underlying OS and Ceph are basically two very separate(d) things?
Regards,
Alan
On Tue, 2025-07-22 at 18:34 -0400, Anthony D'Atri wrote:
Yes.
On Jul 22, 2025, at 6:33 PM, Alan Murrell <Alan@t-net.ca> wrote:
Hello,
I just wanted to check quickly, but feel I am being a little *too* cautious...
We have a Ceph cluster (installed via cephadm) that is currently version 18.2.2. The "Upgrade" tab gives me an option to upgrade to 18.2.7 (its default selection), or I can upgrade to another version where I can choose "earlier" upgrades (from 18.2.3 to the offered 18.2.7)
Is it generally OK and safe to "jump" upgrades like this (i.e. go straight from 18.2.2 straight to 18.2.7) or would it be recommended to upgrade in order (18.2.2 -> 18.2.3 -> 18.2.4 -> etc.)?
I assume once I am on latest 18.2.7, there might be an option to upgrade to 19.2.x (not that I would necessarily want to do that straight away)?
Thanks! :-)
Regards,
Alan _______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io
ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io _______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io
Den ons 23 juli 2025 kl 08:24 skrev Michel Jouvin <michel.jouvin@ijclab.in2p3.fr>:
I'd say yes OS and Ceph upgrades are two very separated operations. Thus I personally advise (it's what I'm doing) not to do them at the same time. It makes easier to identify the culprit in case of problems.I generally upgrade the OS before Ceph.
Agreed, I also recommend you to do one at a time, and as said, often you would do the OS first, to an OS that supports both current ceph version and the version you want to move to, in case you change major version. -- May the most significant bit of your life be positive.
Thank you everyone, for your responses and advice. VERY much appreciated.
Unless you have to reboot a node after OS upgrade. In this case send a node into maintenance mode (ceph orch host maintenance enter <host> --force) before rebooting.
Ah, I will add that one in to my Ceph notes. If I have to reboot a node, I usually just set "no out" for the OSDs, but this one seems better (and PROPER) -- and looking at the documentation for it, seems that it does a "no out" for the OSDs as part of the maintenance process anyway. And thank you for the advice of updating the OS on each node first before upgrading Ceph (I will probably wait a few days after the OS upgrade before upgrading Ceph just to be sure there are no issues with the OS upgrade) Regards, Alan
participants (5)
-
Alan Murrell
-
Anthony D'Atri
-
Janne Johansson
-
Michel Jouvin
-
Vladimir Sigunov