On Tue, May 3, 2016 at 11:48 AM, Robert Haas <robertmh...@gmail.com> wrote:
> OK, I see now: the basic idea here is that we can't prune based on the
> newer XID unless the page LSN is guaranteed to advance whenever data
> is removed. Currently, we attempt to limit bloat in non-unlogged,
> non-catalog tables. You're saying we can instead attempt to limit
> bloat only in non-unlogged, non-catalog tables without hash indexes,
> and that will fix this issue. Am I right?
As a first cut, something like the attached.
--
Kevin Grittner
EDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
diff --git a/src/backend/access/hash/hash.c b/src/backend/access/hash/hash.c
index 4fecece..49a6c81 100644
--- a/src/backend/access/hash/hash.c
+++ b/src/backend/access/hash/hash.c
@@ -279,7 +279,6 @@ hashgettuple(IndexScanDesc scan, ScanDirection dir)
buf = so->hashso_curbuf;
Assert(BufferIsValid(buf));
page = BufferGetPage(buf);
- TestForOldSnapshot(scan->xs_snapshot, rel, page);
maxoffnum = PageGetMaxOffsetNumber(page);
for (offnum = ItemPointerGetOffsetNumber(current);
offnum <= maxoffnum;
diff --git a/src/backend/access/hash/hashsearch.c b/src/backend/access/hash/hashsearch.c
index eb8c9cd..4825558 100644
--- a/src/backend/access/hash/hashsearch.c
+++ b/src/backend/access/hash/hashsearch.c
@@ -189,7 +189,6 @@ _hash_first(IndexScanDesc scan, ScanDirection dir)
/* Read the metapage */
metabuf = _hash_getbuf(rel, HASH_METAPAGE, HASH_READ, LH_META_PAGE);
page = BufferGetPage(metabuf);
- TestForOldSnapshot(scan->xs_snapshot, rel, page);
metap = HashPageGetMeta(page);
/*
@@ -243,7 +242,6 @@ _hash_first(IndexScanDesc scan, ScanDirection dir)
/* Fetch the primary bucket page for the bucket */
buf = _hash_getbuf(rel, blkno, HASH_READ, LH_BUCKET_PAGE);
page = BufferGetPage(buf);
- TestForOldSnapshot(scan->xs_snapshot, rel, page);
opaque = (HashPageOpaque) PageGetSpecialPointer(page);
Assert(opaque->hasho_bucket == bucket);
@@ -350,7 +348,6 @@ _hash_step(IndexScanDesc scan, Buffer *bufP, ScanDirection dir)
_hash_readnext(rel, &buf, &page, &opaque);
if (BufferIsValid(buf))
{
- TestForOldSnapshot(scan->xs_snapshot, rel, page);
maxoff = PageGetMaxOffsetNumber(page);
offnum = _hash_binsearch(page, so->hashso_sk_hash);
}
@@ -392,7 +389,6 @@ _hash_step(IndexScanDesc scan, Buffer *bufP, ScanDirection dir)
_hash_readprev(rel, &buf, &page, &opaque);
if (BufferIsValid(buf))
{
- TestForOldSnapshot(scan->xs_snapshot, rel, page);
maxoff = PageGetMaxOffsetNumber(page);
offnum = _hash_binsearch_last(page, so->hashso_sk_hash);
}
diff --git a/src/backend/utils/cache/relcache.c b/src/backend/utils/cache/relcache.c
index 432feef..79cc3df 100644
--- a/src/backend/utils/cache/relcache.c
+++ b/src/backend/utils/cache/relcache.c
@@ -5313,6 +5313,52 @@ RelationIdIsInInitFile(Oid relationId)
}
/*
+ * Tells whether any index for the relation is unlogged.
+ *
+ * Any index using the hash AM is implicitly unlogged.
+ *
+ * Note: There doesn't seem to be any way to have an unlogged index attached
+ * to a permanent table except to create a hash index, but it seems best to
+ * keep this general so that it returns sensible results even when they seem
+ * obvious (like for an unlogged table) and to handle possible future unlogged
+ * indexes on permanent tables.
+ */
+bool
+RelationHasUnloggedIndex(Relation rel)
+{
+ List *indexoidlist;
+ ListCell *indexoidscan;
+ bool result = false;
+
+ indexoidlist = RelationGetIndexList(rel);
+
+ foreach(indexoidscan, indexoidlist)
+ {
+ Oid indexoid = lfirst_oid(indexoidscan);
+ HeapTuple tp;
+ Form_pg_class reltup;
+
+ tp = SearchSysCache1(RELOID, ObjectIdGetDatum(indexoid));
+ if (!HeapTupleIsValid(tp))
+ elog(ERROR, "cache lookup failed for relation %u", indexoid);
+ reltup = (Form_pg_class) GETSTRUCT(tp);
+
+ if (reltup->relpersistence == RELPERSISTENCE_UNLOGGED
+ || reltup->relam == HASH_AM_OID)
+ result = true;
+
+ ReleaseSysCache(tp);
+
+ if (result == true)
+ break;
+ }
+
+ list_free(indexoidlist);
+
+ return result;
+}
+
+/*
* Invalidate (remove) the init file during commit of a transaction that
* changed one or more of the relation cache entries that are kept in the
* local init file.
diff --git a/src/backend/utils/time/snapmgr.c b/src/backend/utils/time/snapmgr.c
index 0a9a231..e1551a3 100644
--- a/src/backend/utils/time/snapmgr.c
+++ b/src/backend/utils/time/snapmgr.c
@@ -1590,7 +1590,8 @@ TransactionIdLimitedForOldSnapshots(TransactionId recentXmin,
&& old_snapshot_threshold >= 0
&& RelationNeedsWAL(relation)
&& !IsCatalogRelation(relation)
- && !RelationIsAccessibleInLogicalDecoding(relation))
+ && !RelationIsAccessibleInLogicalDecoding(relation)
+ && !RelationHasUnloggedIndex(relation))
{
int64 ts = GetSnapshotCurrentTimestamp();
TransactionId xlimit = recentXmin;
diff --git a/src/include/utils/rel.h b/src/include/utils/rel.h
index b5d82d6..a0ba417 100644
--- a/src/include/utils/rel.h
+++ b/src/include/utils/rel.h
@@ -505,5 +505,6 @@ typedef struct ViewOptions
/* routines in utils/cache/relcache.c */
extern void RelationIncrementReferenceCount(Relation rel);
extern void RelationDecrementReferenceCount(Relation rel);
+extern bool RelationHasUnloggedIndex(Relation rel);
#endif /* REL_H */
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
