Columnar: fix wraparound bug. #5962
Conversation
jeff-davis
force-pushed
the
columnar-fix-wraparound
branch
2 times, most recently
from
1fb9e72 to
c15c8a2
Compare
Codecov Report
@@ Coverage Diff @@
## master #5962 +/- ##
==========================================
- Coverage 92.58% 92.58% -0.01%
==========================================
Files 237 237
Lines 48102 48108 +6
==========================================
+ Hits 44536 44540 +4
- Misses 3566 3568 +2 |
jeff-davis
force-pushed
the
columnar-fix-wraparound
branch
from
c15c8a2 to
fb22e91
Compare
| vac_open_indexes(rel, RowExclusiveLock, &nindexes, &indrels); | ||
| vac_close_indexes(nindexes, indrels, NoLock); | ||
|
|
There was a problem hiding this comment.
Should we do someting with indRels ? Since this code currently just opens and closes those indexes.
There was a problem hiding this comment.
vac_open_indexes seems() to be skipping invalid indexes, so might be nice to add a test where we vacuum a relation that has an invalid index ?
There was a problem hiding this comment.
I changed to use RelationGetIndexList() instead, which is lighter-weight.
It doesn't look like it's skipping indexes where !indisvalid... the comment over vac_open_indexes() says:
* We consider an index vacuumable if it is marked insertable (indisready).
* If it isn't, probably a CREATE INDEX CONCURRENTLY command failed early in
* execution, and what we have is too corrupt to be processable. We will
* vacuum even if the index isn't indisvalid; this is important because in a
* unique index, uniqueness checks will be performed anyway and had better not
* hit dangling index pointers.
If you mean try to test the case where !indisready, that seems like overkill, unless there's a case I'm missing.
| vac_update_relstats(rel, new_rel_pages, new_live_tuples, | ||
| new_rel_allvisible, nindexes > 0, | ||
| newRelFrozenXid, newRelminMxid, false); | ||
| #endif |
There was a problem hiding this comment.
should we call pgstat_report_vacuum at some point ?
There was a problem hiding this comment.
Yes, done.
We could also improve the logging/instrumentation output, but I'll leave that for another PR.
jeff-davis
force-pushed
the
columnar-fix-wraparound
branch
from
fb22e91 to
bc528ee
Compare
columnar_vacuum_rel() now advances relfrozenxid. Fixes #5958.
jeff-davis
force-pushed
the
columnar-fix-wraparound
branch
from
bc528ee to
a2a947c
Compare
columnar_vacuum_rel() now advances relfrozenxid.
Fixes #5958.
DESCRIPTION: Fix columnar freezing/wraparound bug.