public inbox for pve-devel@lists.proxmox.com
 help / color / mirror / Atom feed
From: Nicolas Frey <n.frey@proxmox.com>
To: pve-devel@lists.proxmox.com
Subject: Re: [PATCH pve-docs 1/1] ceph: add warning about mixing device-specific with device-unspecific CRUSH rules when using autoscaler
Date: Wed, 22 Apr 2026 09:30:35 +0200	[thread overview]
Message-ID: <2698684c-ee19-4689-8da5-2557d09dedfa@proxmox.com> (raw)
In-Reply-To: <878qafikbb.fsf@proxmox.com>

On 4/22/26 8:23 AM, Kefu Chai wrote:
> Nicolas Frey <n.frey@proxmox.com> writes:
> 
> Hi Nicolas, thanks for this improvement. A few nits:
> 
>> Suggested-by: Friedrich Weber <f.weber@proxmox.com>
>> Signed-off-by: Nicolas Frey <n.frey@proxmox.com>
>> ---
>>  pveceph.adoc | 3 +++
>>  1 file changed, 3 insertions(+)
>>
>> diff --git a/pveceph.adoc b/pveceph.adoc
>> index 2aae6d6..dfa5d95 100644
>> --- a/pveceph.adoc
>> +++ b/pveceph.adoc
>> @@ -916,6 +916,9 @@ TIP: If the pool already contains objects, these must be moved accordingly.
>>  Depending on your setup, this may introduce a big performance impact on your
>>  cluster. As an alternative, you can create a new pool and move disks separately.
>>
>> +WARNING: When using the autoscaler, all pools must either exclusively be assigned
>> +device-specific or device-unspecific CRUSH rules. Mixing them across pools will
>>
> 
> device-specific / device-unspecific is not Ceph's terminology. Ceph's
> terms are "CRUSH rules that specify a device class", or simply "device
> class rules".
> 

Thanks! I was not aware of the terminology used by ceph here

>> +prevent the autoscaler from functioning.
> 
> this might overstates the problem caused by overlapped roots. The
> autoscaler keeps running: it just skips pools whose CRUSH rules's OSD
> sets overlap, and the pg_num is not adjustted for those pools. In other
> words, non-overlapping pools continue to scale.

Ah, thanks for the explanation!

> 
> So, probably we can be more specific here? like:
> 
> When using the PG autoscaler, all pools in the cluster must use
> CRUSH rules of the same kind, either all specifying a device class,
> or all without one. Otherwise the autoscaler will skip the affected
> pools and their `pg_num` will not be adjusted.
> 
> what do you think?
> 

That sounds great! I'll send a v2 with your suggestions

>>
>>  Ceph Client
>>  -----------
>> --
>> 2.47.3





  reply	other threads:[~2026-04-22  7:31 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-04-21  8:05 Nicolas Frey
2026-04-22  6:24 ` Kefu Chai
2026-04-22  7:30   ` Nicolas Frey [this message]
2026-04-22  7:55 ` superseded: " Nicolas Frey

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=2698684c-ee19-4689-8da5-2557d09dedfa@proxmox.com \
    --to=n.frey@proxmox.com \
    --cc=pve-devel@lists.proxmox.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
Service provided by Proxmox Server Solutions GmbH | Privacy | Legal