Refresh v3 implementation #1115

marceloneppel · 2025-09-30T20:38:11Z

Issue

The charm is not yet supporting refresh through refresh v3.

Solution

Added refresh v3 implementation based on the developer docs and also based on the implementations from PostgreSQL VM and Kafka K8s.

Use set_unit_status to set the unit status and properly set statuses during refresh.

The old refresh/upgrade logic/files were removed.

Correctly reload pgBackRest TLS server configuration with SIGHUP to avoid interrupting the data replication between the primary and the replica (more details in the comments on this PR).

After this PR is approved, 2a806f7 must be reverted, then the PR should me merged, to properly generate the tags in the repo (this is the same as stated at canonical/postgresql-operator#866 (comment)).

Checklist

I have added or updated any relevant documentation.
I have cleaned any remaining cloud resources from my accounts.

Signed-off-by: Marcelo Henrique Neppel <[email protected]>

codecov · 2025-09-30T20:39:17Z

Codecov Report

❌ Patch coverage is 42.43902% with 118 lines in your changes missing coverage. Please review.
✅ Project coverage is 65.98%. Comparing base (48d4eb4) to head (86860f1).
⚠️ Report is 6 commits behind head on 16/edge.

Files with missing lines	Patch %	Lines
src/charm.py	39.53%	64 Missing and 14 partials ⚠️
src/refresh.py	41.93%	17 Missing and 1 partial ⚠️
src/relations/async_replication.py	47.05%	8 Missing and 1 partial ⚠️
src/relations/postgresql_provider.py	10.00%	9 Missing ⚠️
src/backups.py	86.66%	1 Missing and 1 partial ⚠️
src/relations/logical_replication.py	0.00%	2 Missing ⚠️

❌ Your project check has failed because the head coverage (65.98%) is below the target coverage (70.00%). You can increase the head coverage or adjust the target coverage.

Additional details and impacted files

@@             Coverage Diff             @@
##           16/edge    #1115      +/-   ##
===========================================
- Coverage    67.44%   65.98%   -1.47%     
===========================================
  Files           18       18              
  Lines         3947     3904      -43     
  Branches       564      568       +4     
===========================================
- Hits          2662     2576      -86     
- Misses        1112     1143      +31     
- Partials       173      185      +12

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

Signed-off-by: Marcelo Henrique Neppel <[email protected]>

… tag Signed-off-by: Marcelo Henrique Neppel <[email protected]>

Signed-off-by: Marcelo Henrique Neppel <[email protected]>

…entation Signed-off-by: Marcelo Henrique Neppel <[email protected]>

…check for primary on unit zero Signed-off-by: Marcelo Henrique Neppel <[email protected]>

Signed-off-by: Marcelo Henrique Neppel <[email protected]>

…efresh is not ready (#1156) * Reorder call to refresh logic and don't immediatelly exit hook when refresh is not ready Signed-off-by: Marcelo Henrique Neppel <[email protected]> * Log more info when creating a backup and send SIGHUP when the pgBackRest is already active Signed-off-by: Marcelo Henrique Neppel <[email protected]> * Change log level of messages to debug Signed-off-by: Marcelo Henrique Neppel <[email protected]> * Update unit test Signed-off-by: Marcelo Henrique Neppel <[email protected]> * Fix integration test job collection Signed-off-by: Marcelo Henrique Neppel <[email protected]> --------- Signed-off-by: Marcelo Henrique Neppel <[email protected]>

marceloneppel · 2025-11-20T11:52:17Z

.github/workflows/ci.yaml

      - '**.md'
      - .github/renovate.json5
      - 'docs/**'
+      - 'terraform/**'


Added to match the rule we have in the VM charm.

marceloneppel · 2025-11-20T11:53:05Z

.github/workflows/integration_test.yaml

+                  [pathlib.Path.home() / "go/bin/spread", "-list", "github-ci"],
+                  capture_output=True,
+                  check=True,
+                  text=True,


Update to list tests now that we're using charmcraft v4.

marceloneppel · 2025-11-20T12:01:17Z

src/backups.py

+                "--log-level-file=debug",
+                "--log-subprocess",


More logging added (this helped to identify why the GCP PITR test was failing after refresh V3 was added in this PR: a SIGTERM appeared in the logs).

marceloneppel · 2025-11-20T12:07:56Z

src/backups.py

+            logger.debug("Sending SIGHUP to pgBackRest TLS server to reload configuration")
+            self.container.pebble.send_signal(
+                signal.SIGHUP, services=[self.charm.pgbackrest_server_service]
+            )


Always restarting the service was causing a SIGTERM to be sent to the pgBackRest TLS server in the middle of the process where the replica gets the latest data from the primary to create a backup. This started to happen because of some more deferred events happening. A SIGHUP is enough to update the TLS server when the certificates change.

marceloneppel · 2025-11-20T12:11:39Z

tests/integration/ha_tests/test_upgrade.py

+        # ops_test.model.deploy(
+        #     DATABASE_APP_NAME,
+        #     num_units=3,
+        #     channel="16/edge",
+        #     trust=True,
+        #     config={"profile": "testing"},
+        #     base=CHARM_BASE_NOBLE,
+        # ),


Commented to be restored after the first release with refresh v3 is available.

marceloneppel · 2025-11-20T12:12:05Z

tests/integration/ha_tests/test_upgrade_from_stable.py

+        # ops_test.model.deploy(
+        #     DATABASE_APP_NAME,
+        #     num_units=3,
+        #     channel="16/stable",
+        #     trust=True,
+        #     config={"profile": "testing"},
+        #     base=CHARM_BASE_NOBLE,
+        # ),


Commented to be restored after the first release with refresh v3 is available.

marceloneppel · 2025-11-20T12:13:14Z

tests/integration/helpers.py

            assert standbys == len(ops_test.model.applications[app_name].units) - 1


+async def switchover_to_unit_zero(ops_test: OpsTest) -> None:


Used to reduce the number of switchovers to the minimum (as we expect since the previous refresh/upgrade implementation).

marceloneppel · 2025-11-20T12:14:04Z

tests/integration/test_trust.py

-UNTRUST_ERROR_MESSAGE = f"Insufficient permissions, try: `juju trust {APP_NAME} --scope=cluster`"
+UNTRUST_ERROR_MESSAGE = (
+    f"Run `juju trust {APP_NAME} --scope=cluster`. Needed for in-place refreshes"
+)


The message changes a bit with refresh v3.

Add first part of refresh v3 implementation

dcdc4dd

Signed-off-by: Marcelo Henrique Neppel <[email protected]>

github-actions bot added the Libraries: Out of sync label Sep 30, 2025

marceloneppel added the enhancement New feature, UI change, or workload upgrade label Oct 1, 2025

marceloneppel added 3 commits October 1, 2025 17:26

Add second part of refresh v3 implementation

02651eb

Signed-off-by: Marcelo Henrique Neppel <[email protected]>

Merge remote-tracking branch 'origin/16/edge' into refresh-v3

cd935f2

Signed-off-by: Marcelo Henrique Neppel <[email protected]>

Update poetry.lock

0e073d0

Signed-off-by: Marcelo Henrique Neppel <[email protected]>

github-actions bot added Libraries: OK and removed Libraries: Out of sync labels Oct 2, 2025

marceloneppel added 2 commits October 2, 2025 13:09

Revert (for testing) Set charm refresh compatibility version from git…

086bc7c

… tag Signed-off-by: Marcelo Henrique Neppel <[email protected]>

Merge remote-tracking branch 'origin/16/edge' into refresh-v3

13fe1c2

Signed-off-by: Marcelo Henrique Neppel <[email protected]>

github-actions bot added Libraries: Out of sync and removed Libraries: OK labels Oct 17, 2025

marceloneppel added 17 commits October 27, 2025 10:57

Update logic refresh to start workload at the right time

6482969

Signed-off-by: Marcelo Henrique Neppel <[email protected]>

Wait some seconds before calling resume-refresh

c40d26e

Signed-off-by: Marcelo Henrique Neppel <[email protected]>

Fix both integration tests and make them handle incompatible upgrades

7fe482a

Signed-off-by: Marcelo Henrique Neppel <[email protected]>

Fix unit tests

e53aa35

Signed-off-by: Marcelo Henrique Neppel <[email protected]>

Remove partition check

9b9b03a

Signed-off-by: Marcelo Henrique Neppel <[email protected]>

Remove charm refresh compatibility versioning from the git tag implem…

2a806f7

…entation Signed-off-by: Marcelo Henrique Neppel <[email protected]>

Remove commented code, fix scale down and pre-refresh check, and add …

38e4870

…check for primary on unit zero Signed-off-by: Marcelo Henrique Neppel <[email protected]>

Remove debug error messages

9c1a026

Signed-off-by: Marcelo Henrique Neppel <[email protected]>

Switchover to unit zero when needed on upgrade tests

18c51c4

Signed-off-by: Marcelo Henrique Neppel <[email protected]>

Fix trust test

7a9c052

Signed-off-by: Marcelo Henrique Neppel <[email protected]>

Fix redis-k8s channel

bc4729f

Signed-off-by: Marcelo Henrique Neppel <[email protected]>

Remove commented code

d3a753e

Signed-off-by: Marcelo Henrique Neppel <[email protected]>

Remove commented code and fix log calls

dfe5b0a

Signed-off-by: Marcelo Henrique Neppel <[email protected]>

Fix redis-k8s base

06c15fe

Signed-off-by: Marcelo Henrique Neppel <[email protected]>

Increase tests timeouts

c7cdacd

Signed-off-by: Marcelo Henrique Neppel <[email protected]>

Increase one more test timeout

ed74668

Signed-off-by: Marcelo Henrique Neppel <[email protected]>

Increase test timeout

1cfc44d

Signed-off-by: Marcelo Henrique Neppel <[email protected]>

Merge remote-tracking branch 'origin/16/edge' into refresh-v3

46eac97

Signed-off-by: Marcelo Henrique Neppel <[email protected]>

carlcsaposs-canonical mentioned this pull request Nov 11, 2025

Update canonical/data-platform-workflows action to v37 (16/edge) #1146

Open

1 task

marceloneppel commented Nov 20, 2025

View reviewed changes

marceloneppel marked this pull request as ready for review November 20, 2025 12:39

marceloneppel requested review from a team, dragomirp, onkolahmet and taurus-forever and removed request for a team November 20, 2025 12:39

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Refresh v3 implementation #1115

Refresh v3 implementation #1115

Uh oh!

marceloneppel commented Sep 30, 2025 •

edited

Loading

Uh oh!

codecov bot commented Sep 30, 2025 •

edited

Loading

Uh oh!

marceloneppel Nov 20, 2025

Uh oh!

marceloneppel Nov 20, 2025

Uh oh!

marceloneppel Nov 20, 2025

Uh oh!

marceloneppel Nov 20, 2025

Uh oh!

marceloneppel Nov 20, 2025

Uh oh!

marceloneppel Nov 20, 2025

Uh oh!

marceloneppel Nov 20, 2025

Uh oh!

marceloneppel Nov 20, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

		assert standbys == len(ops_test.model.applications[app_name].units) - 1


		async def switchover_to_unit_zero(ops_test: OpsTest) -> None:

Refresh v3 implementation #1115

Are you sure you want to change the base?

Refresh v3 implementation #1115

Uh oh!

Conversation

marceloneppel commented Sep 30, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Issue

Solution

Checklist

Uh oh!

codecov bot commented Sep 30, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

marceloneppel commented Sep 30, 2025 •

edited

Loading

codecov bot commented Sep 30, 2025 •

edited

Loading