Pool update and pool creation does not honor subFailureDomain

[ideepika]

Is this a bug report or feature request?

• Bug Report

Deviation from expected behavior:
Parameters like subFailureDomain and CrushRoot are not honored when updating crush rule(this is observed on new pool creation as well as crush rule update using yaml edit), the feature to support crush rule change was brought by this PR: #12263 but if subfailureDomain is specified, it is completely ignored and instead a new rule get's created.

Eg:

+ failureDomain: zone
 replicated:
- size: 3
+ size: 4
+ replicasPerFailureDomain: 2
+ subFailureDomain: host
+ crushRoot: default
+ deviceClass: nvme

2 crush rules get's created:

2024-06-03 11:51:25.237613 I | cephclient: creating a new crush rule for changed deviceClass ("default~hdd"-->"hdd") on crush rule "test-crush-bug-az-4"
2024-06-03 11:51:25.237647 I | cephclient: updating pool "test-crush-bug-az-4" failure domain from "zone" to "zone" with new crush rule "test-crush-bug-az-ab-4_zone_hdd"

2 rules get's created:

   {
        "rule_id": 4,
        "rule_name": "test-crush-bug-az-ab-4_zone_hdd",
        "type": 1,
        "steps": [
            {
                "op": "take",
                "item": -2,
                "item_name": "default~hdd"
            },
            {
                "op": "chooseleaf_firstn",
                "num": 0,
                "type": "zone"
            },
            {
                "op": "emit"
            }
        ]
    },

    {
        "rule_id": 157,
        "rule_name": "test-crush-bug-az-ab-4",
        "type": 1,
        "steps": [
            {
                "op": "take",
                "item": -2,
                "item_name": "default~hdd"
            },
            {
                "op": "choose_firstn",
                "num": 0,
                "type": "zone"
            },
            {
                "op": "chooseleaf_firstn",
                "num": 2,
                "type": "host"
            },
            {
                "op": "emit"
            }
        ]
    }
]

of which 2nd one is correct, also crush rule update should honor subFailureDomain

[ideepika]

@subhamkrai

[travisn]

SubfailureDomain is a field currently only supported for stretch clusters. Are you wanting to use it and support updates for some other scenario, or is this for stretch? I just wouldn't expect the subfailure domain to change in stretch clusters.

[ideepika]

SubfailureDomain is a field currently only supported for stretch clusters. Are you wanting to use it and support updates for some other scenario, or is this for stretch? I just wouldn't expect the subfailure domain to change in stretch clusters.

yes, for advanced cases, it was succcesfully created but I suppose on OSD restart or when operator rescans the configuration, it replaces the subdomain crush rule with the one with just failuredomain, maybe related to #14300

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed in a week if no further activity occurs. Thank you for your contributions.

[github-actions]

This issue has been automatically closed due to inactivity. Please re-open if this still requires investigation.