[Bug 2146964] Re: [SRU] Prevent masakari HA from creating duplicated notifications

[Seyeong Kim]

@alanbach

Thanks for the update.

This is not mandatory but I would suggest using MR instead of debdiffs
for openstack cases as we can review it easier by releases or series.

This might not be 100% perfect example but please refer to this

https://code.launchpad.net/~seyeongkim/ubuntu/+source/gnocchi/+git/gnocchi
https://code.launchpad.net/~seyeongkim/ubuntu/+source/gnocchi/+git/gnocchi/+merge/505220

Let me review above debdiffs though soon.

Thanks!

-- 
You received this bug notification because you are a member of Ubuntu
OpenStack, which is subscribed to Ubuntu Cloud Archive.
https://bugs.launchpad.net/bugs/2146964

Title:
  [SRU] Prevent masakari HA from creating duplicated notifications

Status in Ubuntu Cloud Archive:
  In Progress
Status in Ubuntu Cloud Archive caracal series:
  In Progress
Status in Ubuntu Cloud Archive dalmatian series:
  Won't Fix
Status in Ubuntu Cloud Archive epoxy series:
  New
Status in Ubuntu Cloud Archive flamingo series:
  In Progress
Status in Ubuntu Cloud Archive gazpacho series:
  In Progress
Status in Ubuntu Cloud Archive ussuri series:
  Won't Fix
Status in Ubuntu Cloud Archive yoga series:
  In Progress
Status in masakari:
  Fix Committed
Status in masakari package in Ubuntu:
  In Progress
Status in masakari source package in Focal:
  Won't Fix
Status in masakari source package in Jammy:
  In Progress
Status in masakari source package in Noble:
  In Progress
Status in masakari source package in Questing:
  Won't Fix
Status in masakari source package in Resolute:
  In Progress
Status in masakari source package in Stonking:
  In Progress

Bug description:
  [ Impact ]

   * Older versions of Masakari are lacking a proper mechanism to prevent
     concurrent record inserts into the database, resulting duplicated
     event processing that can potentially lead to disastrous outcomes.
     These include simultanious server migration attempts due to duplicated
     host notifications that can cause failures and database inconsistencies
     which are not always easy to fix.

   * Affected versios include:

     - Focal/Ussuri 9.0.0-0ubuntu0.20.04.5 / 9.0.0-0ubuntu0.20.04.5~cloud0 (UCA) (Won't Fix - Code too old)
     - Jammy/Yoga 13.0.0-0ubuntu1 / 13.0.0-0ubuntu1~cloud0 (UCA)
     - Jammy/Caracal 17.0.0-0ubuntu1 /	17.0.0-0ubuntu1~cloud0 (UCA)
     - Noble 17.0.0-0ubuntu1
     - Resolute 21.0.0-0ubuntu1
     - Stonking 21.0.0-0ubuntu1

   * This is a known issue and reported by many users also presented in
     LP#2028450.

   * This requires an HA Masakari deployment without a coordinator
     configured.

   * None of the current Charmed Masakari deployments support using a
     coordinator and are all affected.

  [ Test Plan ]

   * Testing requires an OpenStack environment deployed with HA Masakari.
     Masakari deployed in HA mode is a hard requirement as the issue only
     affects HA deployments due to workers competing with each other.

   * The LP#2028450 bug includes a reproducer script (2), which is handy
     to reproduce the issue as well as validate the fix.

   * The reproducer script simulates concurrent insertion situations
     using OpenStack CLI.
     
   * After running the reproducer script, execute the following command:
   
  openstack notification list --sort-column generated_time --insecure | tail -n 30
   
   * Bad Case: Look for COMPUTE_HOST entries that have close timestamps and pay attention
     to the state column. The following is an example of a problematic state with
     two entires created with close proximity affecting the same host and both
     and not ignored:
     
  | 9a06d57e-4fb3-4d44-bea3-c3d6f5ee3ea8 | 2026-04-22T17:41:47.000000 | finished | COMPUTE_HOST | 1cb67d44-4a3f-4855-bf3e-960aec3aa079
  | 60075e3c-1d85-4825-86d6-071d345a98fb | 2026-04-22T17:41:47.000000 | finished | COMPUTE_HOST | 1cb67d44-4a3f-4855-bf3e-960aec3aa079

   * Good Case: There should not be duplicated entries with a close proximity:
   
  | 06a5422f-0ea6-46a1-b6b7-a171158f7b57 | 2026-04-22T18:00:47.000000 | finished | COMPUTE_HOST | 1cb67d44-4a3f-4855-bf3e-960aec3aa079

  
  [ Where problems could occur ]

   * The nature if this change is very simple. In the HA API, the
     create notification function will be modified to introduce an
     artificial pause, significantly reducing the chance of
     concurrent insertion if event records in database.

     This will only be invoked if coordination back-end is not
     configured. Please see (3) for more details:

     if not CONF.coordination.backend_url:
             time.sleep(random.uniform(1, 5))

   * The probable risk here is a delay between 1-5 seconds during
     event creation and further processing. These events happen
     for example when a compute node becomes inaccessible and the
     instances need to be evacuated and launched on a new node.
     In other words the recovery will be delayed for 1-5 seconds.

  [ Other Info ]

   * The proposed changes have been merged upstream (4).

  (1) https://bugs.launchpad.net/masakari/+bug/2028450
  (2) https://launchpadlibrarian.net/849768758/lp2028450-reproducer.bash
  (3) https://review.opendev.org/c/openstack/masakari/+/978343/3/masakari/ha/api.py
  (4) https://review.opendev.org/c/openstack/masakari/+/978343

To manage notifications about this bug go to:
https://bugs.launchpad.net/cloud-archive/+bug/2146964/+subscriptions