jangjang0401 opened a new pull request, #1743:
URL: https://github.com/apache/cloudberry/pull/1743
With hot_standby=on, mirrors accept SQL connections and return PQPING_OK
from pg_isready. The previous code unconditionally mapped PQPING_OK +
role=mirror to "Acting as Primary", causing gpstate -s to show a spurious
warning on every mirror.
Fix this in clsSystemState.__buildGpStateData() by cross-checking
pg_stat_replication on the primary: if the mirror has an active WAL
receiver connection in streaming or catchup state, it is a legitimate
hot standby and the status is corrected to "Up". Only fall through to
"Acting as Primary" when no such replication connection exists, meaning
the segment truly promoted itself to primary.
This approach reuses the existing primary connection already established
by _add_replication_info(), so no additional database connections are
required. _add_replication_info() is updated to return the raw
replication state string to make this information available to the
caller.
### What does this PR do?
Fixes a false positive in `gpstate -s` where every mirror segment is
reported as "Acting as Primary" when `hot_standby=on` is enabled.
Root cause:
- With `hot_standby=on`, mirrors accept read-only SQL connections and
`pg_isready` returns `PQPING_OK(0)` instead of `PQPING_MIRROR_READY(64)`.
- The legacy logic in `gpgetstatususingtransition.py` assumed
`PQPING_OK + role=mirror` could only mean "the mirror was promoted to
primary", which is no longer true under hot standby.
Fix:
- In `clsSystemState.__buildGpStateData()`, cross-check the mirror's
status against `pg_stat_replication` (already queried on the primary
by `_add_replication_info()`).
- If the mirror has an active WAL receiver connection (`streaming` or
`catchup`), it is a legitimate hot standby → corrected to `"Up"`.
- If no such replication connection exists, the mirror genuinely
promoted itself → `"Acting as Primary"` is preserved.
### Type of Change
- [x] Bug fix (non-breaking change)
- [ ] New feature (non-breaking change)
- [ ] Breaking change (fix or feature with breaking changes)
- [ ] Documentation update
### Breaking Changes
None. The fix only changes how the status string is derived in
`gpstate`; cluster configuration, replication behavior, and segment
behavior are unchanged.
### Test Plan
Tested on a 5-VM cluster (1 coordinator + 1 standby coordinator + 3
segment hosts, 6 primary / 6 mirror segments) with `hot_standby=on`:
**Before the fix:**
- Every mirror reported `Segment status = Acting as Primary`
- `gpstate -s` produced spurious warnings on every mirror
- `gp_segment_configuration`, replication state, and `gpstate -m` all
showed the cluster as healthy — only the `gpstate -s` output was wrong
**After the fix:**
- Every mirror reports `Segment status = Up`
- No "Acting as Primary" entries appear on healthy mirrors
- No additional database connections are opened
### Impact
**Performance:**
No additional database connections. The fix consumes data that
`_add_replication_info()` already collects from the primary; only the
return value of that function is changed.
**User-facing changes:**
`gpstate -s` no longer reports false "Acting as Primary" warnings for
healthy hot-standby mirrors. Genuine promotion is still detected and
reported as before.
**Dependencies:**
None.
### Checklist
- [x] Followed [contribution
guide](https://cloudberry.apache.org/contribute/code)
- [ ] Added/updated documentation
- [x] Reviewed code for security implications
- [ ] Requested review from [cloudberry
committers](https://github.com/orgs/apache/teams/cloudberry-committers)
### Additional Context
An alternative approach considered was to call `pg_is_in_recovery()`
directly on each mirror inside `_get_segment_status()` in
`gpgetstatususingtransition.py`. That approach was rejected because:
1. It opens one additional DB connection per mirror on every `gpstate`
invocation, which is undesirable in unhealthy-cluster scenarios
where `gpstate` is run most frequently.
2. Mirror-side recovery state is a weaker source of truth than the
primary's view of its replication connections.
The `pg_stat_replication` approach reuses an existing connection and is
authoritative from the primary's perspective.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
