[PATCH manager] fix #8163: ceph: pool create: pass the CRUSH rule on creation

[PATCH manager] fix #8163: ceph: pool create: pass the CRUSH rule on creation

[Kefu Chai]

create_pool() only passes the pool name and pg_num to "osd pool create"
and sets the CRUSH rule afterwards with "osd pool set". Ceph checks
mon_max_pg_per_osd when a pool is created, against the rule it is
created with, but not when the rule of an existing pool changes. So a
pool with a non-default rule is checked against the OSDs and pools of
the default rule. It is refused if those are near the limit, even if it
would fit on its own rule, and it is never checked against its own
rule.

Pass the rule to "osd pool create" for replicated pools. Erasure-coded
pools stay as they are, since Ceph creates their rule from the profile.

Keep setting the size afterwards. Ceph checks the limit again when the
size grows, and with this change that check uses the pool's own rule.
Passing the size on creation would also make it fail in stretch mode,
unless the size matches mon_stretch_pool_size.

Signed-off-by: Kefu Chai <k.chai@proxmox.com>
---
 PVE/Ceph/Tools.pm | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/PVE/Ceph/Tools.pm b/PVE/Ceph/Tools.pm
index 1eed2a2d..94177709 100644
--- a/PVE/Ceph/Tools.pm
+++ b/PVE/Ceph/Tools.pm
@@ -349,6 +349,10 @@ sub create_pool {
     $mon_params->{erasure_code_profile} = extract_param($param, 'erasure_code_profile')
         if $param->{erasure_code_profile};
 
+    # Ceph checks mon_max_pg_per_osd against the rule given on create, not on a later rule change
+    $mon_params->{rule} = extract_param($param, 'crush_rule')
+        if $param->{crush_rule} && ($mon_params->{pool_type} // '') ne 'erasure';
+
     $rados->mon_command($mon_params);
 
     set_pool($pool, $param);
-- 
2.47.3

applied: [PATCH manager] fix #8163: ceph: pool create: pass the CRUSH rule on creation

[Thomas Lamprecht]

On Thu, 15 Oct 2026 15:19:43 +0800, Kefu Chai wrote:
> create_pool() only passes the pool name and pg_num to "osd pool create"
> and sets the CRUSH rule afterwards with "osd pool set". Ceph checks
> mon_max_pg_per_osd when a pool is created, against the rule it is
> created with, but not when the rule of an existing pool changes. So a
> pool with a non-default rule is checked against the OSDs and pools of
> the default rule. It is refused if those are near the limit, even if it
> would fit on its own rule, and it is never checked against its own
> rule.
> 
> [...]

Applied, thanks!

[1/1] fix #8163: ceph: pool create: pass the CRUSH rule on creation
      https://git.proxmox.com/?p=pve-manager.git;a=commitdiff;h=5a1d4293a5b1d55bb29fbe19f84043dfe869e257