gefyrahq / gefyra / 23653504911
74%

DEFAULT BRANCH: main
Ran
Jobs 0
Files 0
Run time –
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
23653504911

push

github

web-flow 
Fix(#GO-1020): BridgeMount stuck in PREPARING when HPA scales original deployment (#924)

* fix: break infinite loop when HPA scales during mount install

When HPA changes the replica count while a GefyraBridgeMount is in
INSTALLING or PREPARING, the mount gets stuck forever: ready() requires
matching pod counts between original and shadow workloads, but the
shadow still has the old count.

Add replica count mismatch detection to prepared() and route to restore
instead of retrying install when a mismatch is found. This sends the
mount back through RESTORING → PREPARING, which re-clones the workload
with the correct replica count.

- Add installing → restoring and preparing → restoring transitions
- Detect replica count divergence in prepared()
- Handler checks prepared() before install(), sends restore on mismatch
- Add unit tests covering HPA scale scenarios

* style(#GO-1020): apply ruff formatting to handler and tests

* fix(#GO-1020): limit prepared() circuit breaker to INSTALLING state only

In PREPARING, pod count differences are expected — shadow pods are still
starting up. The existing @install.cond guard handles this via
TemporaryError retries until pods are ready.

The circuit breaker (prepared() → restore) is only needed in INSTALLING,
where a pod count mismatch means HPA scaled the original workload and
the mount is genuinely stuck.

* fix(#GO-1020): scale shadow to match original on HPA replica drift

  The previous fix (db3e406) removed the explicit prepared() check from
  PREPARING state in the handler, but the @install.cond condition still
  called prepared() — blocking the PREPARING→INSTALLING transition on
  replica count mismatch. With an active HPA the counts never converge,
  leaving the mount stuck in PREPARING forever.

  - Add _scale_shadow_to_match(): patches the shadow deployment to the
    original's current replica count instead of triggering a full restore
    cycle (delete + re-clone).
  - Make _bridge_mount_prepared cond... (continued)
Source Files on build 23653504911
Detailed source file information is not available for this build.