Bug 1810577 - [RFE] Add new MD RAID resource agent (i.e. successor to upstream 'Raid1' agent)
Summary: [RFE] Add new MD RAID resource agent (i.e. successor to upstream 'Raid1' agent)
Keywords:
Status: POST
Alias: None
Product: Red Hat Enterprise Linux 8
Classification: Red Hat
Component: resource-agents
Version: 8.3
Hardware: Unspecified
OS: Unspecified
medium
medium
Target Milestone: rc
: 8.9
Assignee: Oyvind Albrigtsen
QA Contact: cluster-qe
Steven J. Levine
URL:
Whiteboard:
Depends On: 1741644 1810561
Blocks: 2166683
TreeView+ depends on / blocked
 
Reported: 2020-03-05 14:29 UTC by Heinz Mauelshagen
Modified: 2023-08-10 15:41 UTC (History)
19 users (show)

Fixed In Version:
Doc Type: If docs needed, set a value
Doc Text:
Clone Of:
: 2166683 (view as bug list)
Environment:
Last Closed:
Type: Bug
Target Upstream Version:
Embargoed:


Attachments (Terms of Use)


Links
System ID Private Priority Status Summary Last Updated
Red Hat Knowledge Base (Article) 2912911 0 None None None 2022-12-19 18:57:05 UTC
Red Hat Knowledge Base (Solution) 3977931 0 None None None 2020-11-17 19:19:27 UTC

Description Heinz Mauelshagen 2020-03-05 14:29:58 UTC
Description of problem:
The upstream resource-agents package contains a functionally limited 'Raid1' resource agent relative to supporting MD Clustered RAID1/10.  This is the feature bz to track the development of a 'Raid1' cloned pacemaker resource agent with cluster enhancements.  For instance, 'Raid1' has no notion of an MD array being clustered or not, thus can't reject creation of a respective resource properly or automatically adjust to it.

Version-Release number of selected component (if applicable):
All

How reproducible:


Steps to Reproduce:
1.
2.
3.

Actual results:


Expected results:


Additional info:
See bz dependency for initial use/test case examples (subject to be copied and enhanced in this bz later)

Comment 1 Heinz Mauelshagen 2020-03-11 12:02:01 UTC
We are developing the agent -> reassigned.

Comment 2 Heinz Mauelshagen 2020-03-13 16:31:32 UTC
Including test cases from closed 1810561 here.

Test cases:
- active-passive
  - create MD resources for all levels (0/1/4/5/6/10) with force_clones=false
  - check they get started properly
  - put (single host) filesystems (xfs, ext4, gfs2) on top and load with I/O
  - define ordering constraints for ^
  - move them to other node manually
  - disable/enable them
  - add/remove legs to/from raid1
  - fail/fence a node and check if they get started correctly
    on another online node
  - fail access to MD array leg(s) on its active node
    and analyse MD/resource/filesystem behaviour
  - reboot whole cluster and check for resource properly started
  - update the resource option force_clones to true
  - clone the resource (should fail on single host MD arrays but it doesn't)
  - test for data corruption  (even with gfs2 on top)
  - remove all resources in/out of order and check for proper removal

- active-active
  - create clustered MD resources for levels 1 and 10
  - check they get started properly
  - clone them
  - check they get started properly on all other online nodes
  - add/remove legs to/from RAID1
  - try reshaping RAID10
  - disable/enable them
  - fail a node and check if they get started correctly on another node
  - fail access to MD array leg(s) on any/multiple nodes
    and analyse MD/resource/filesystem behaviour
  - remove all resources in/out of order and check for proper removal

Comment 3 Heinz Mauelshagen 2020-04-08 12:09:24 UTC
Resource agent in final internal pre-upstream review by Oyvind Albrigtsen et. al. on the cluster team.
Delaying switiching bz status to POST until the pull upstream requests..

Comment 5 Oyvind Albrigtsen 2020-04-15 14:44:14 UTC
https://github.com/ClusterLabs/resource-agents/pull/1481

Comment 7 Oyvind Albrigtsen 2020-04-22 12:46:58 UTC
Agent merged upstream.

Comment 15 Heinz Mauelshagen 2020-10-26 16:20:28 UTC
As we support clustered MD RAID1 and RAID10 in RHEL8, the mdraid resource agent is mandatory for managing such arrays.


Note You need to log in before you can comment on or make changes to this bug.