You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Is your feature request related to a problem? Please describe.
When PostgreSQL upgrade fails, Epiphany performs automatic rollback (Ansible rescue block) during which the following operations are done:
remove PostgreSQL related packages (new version)
remove new version data directory
restore repmgr symlinks
start and enable old PostgreSQL service
start and enable old repmgr service
When operator wants to debug/analyze a failure, there is no new binaries, so reproducing an issue (re-run the failed command) is difficult.
Describe the solution you'd like
The rollback creates a file (can be script) with instructions how to uninstall new binaries instead removing them automatically.
Describe alternatives you've considered
No automatic rollback, all commands in a file.
Additional context
The upgrade is run based on the condition: when: not pg_data_dir.stat.exists or pg_data_dir_files.matched == 0.
So the rollback must remove files from data dir.
DoD checklist
Changelog updated
COMPONENTS.md updated / doesn't need to be updated
Schema updated / doesn't need to be updated
Feature has automated tests
Automated tests passed (QA pipelines)
apply
upgrade
Idempotency tested
Documentation added / updated / doesn't need to be updated
All conversations in PR resolved
Solution meets requirements and is done according to design doc
Usage compliant with license
Backport tasks created / doesn't need to be backported
The text was updated successfully, but these errors were encountered:
✔️ Tested apply and upgrade from v1.0.x to v1.3.0dev with rollback procedure,
Single-node installation and cluster deployment with replication enabled have been checked.
Is your feature request related to a problem? Please describe.
When PostgreSQL upgrade fails, Epiphany performs automatic rollback (Ansible rescue block) during which the following operations are done:
When operator wants to debug/analyze a failure, there is no new binaries, so reproducing an issue (re-run the failed command) is difficult.
Describe the solution you'd like
The rollback creates a file (can be script) with instructions how to uninstall new binaries instead removing them automatically.
Describe alternatives you've considered
No automatic rollback, all commands in a file.
Additional context
The upgrade is run based on the condition:
when: not pg_data_dir.stat.exists or pg_data_dir_files.matched == 0
.So the rollback must remove files from data dir.
DoD checklist
The text was updated successfully, but these errors were encountered: