Fixes for single-page hash index vacuum.
authorRobert Haas <rhaas@postgresql.org>
Mon, 20 Mar 2017 19:49:09 +0000 (15:49 -0400)
committerRobert Haas <rhaas@postgresql.org>
Mon, 20 Mar 2017 19:49:09 +0000 (15:49 -0400)
Clear LH_PAGE_HAS_DEAD_TUPLES during replay, similar to what gets done
for btree.  Update hashdesc.c for xl_hash_vacuum_one_page.

Oversights in commit 6977b8b7f4dfb40896ff5e2175cad7fdbda862eb spotted
by Amit Kapila.  Patch by Ashutosh Sharma.

Bump WAL version.  The original patch to make hash indexes write-ahead
logged probably should have done this, and the single page vacuuming
patch probably should have done it again, but better late than never.

Discussion: http://postgr.es/m/CAA4eK1Kd=mJ9xreovcsh0qMiAj-QqCphHVQ_Lfau1DR9oVjASQ@mail.gmail.com

src/backend/access/hash/hash.c
src/backend/access/hash/hash_xlog.c
src/backend/access/hash/hashinsert.c
src/backend/access/rmgrdesc/hashdesc.c
src/include/access/hash_xlog.h
src/include/access/xlog_internal.h

index cfcec3475d420effd4eeb87051cb4e7618c4dca7..34cc08f12d2fcb39b3bb555bce359864e3e1c25f 100644 (file)
@@ -790,6 +790,7 @@ hashbucketcleanup(Relation rel, Bucket cur_bucket, Buffer bucket_buf,
        OffsetNumber deletable[MaxOffsetNumber];
        int         ndeletable = 0;
        bool        retain_pin = false;
+       bool        clear_dead_marking = false;
 
        vacuum_delay_point();
 
@@ -877,11 +878,14 @@ hashbucketcleanup(Relation rel, Bucket cur_bucket, Buffer bucket_buf,
            /*
             * Let us mark the page as clean if vacuum removes the DEAD tuples
             * from an index page. We do this by clearing LH_PAGE_HAS_DEAD_TUPLES
-            * flag. Clearing this flag is just a hint; replay won't redo this.
+            * flag.
             */
            if (tuples_removed && *tuples_removed > 0 &&
                opaque->hasho_flag & LH_PAGE_HAS_DEAD_TUPLES)
+           {
                opaque->hasho_flag &= ~LH_PAGE_HAS_DEAD_TUPLES;
+               clear_dead_marking = true;
+           }
 
            MarkBufferDirty(buf);
 
@@ -891,6 +895,7 @@ hashbucketcleanup(Relation rel, Bucket cur_bucket, Buffer bucket_buf,
                xl_hash_delete xlrec;
                XLogRecPtr  recptr;
 
+               xlrec.clear_dead_marking = clear_dead_marking;
                xlrec.is_primary_bucket_page = (buf == bucket_buf) ? true : false;
 
                XLogBeginInsert();
index 8647e8c6adca5cb9dfbda70d87f20c44cf5819b8..ac82092ab2492ca7081f2118077436cb16f975a9 100644 (file)
@@ -859,6 +859,19 @@ hash_xlog_delete(XLogReaderState *record)
                PageIndexMultiDelete(page, unused, unend - unused);
        }
 
+       /*
+        * Mark the page as not containing any LP_DEAD items only if
+        * clear_dead_marking flag is set to true. See comments in
+        * hashbucketcleanup() for details.
+        */
+       if (xldata->clear_dead_marking)
+       {
+           HashPageOpaque pageopaque;
+
+           pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+           pageopaque->hasho_flag &= ~LH_PAGE_HAS_DEAD_TUPLES;
+       }
+
        PageSetLSN(page, lsn);
        MarkBufferDirty(deletebuf);
    }
@@ -1078,6 +1091,7 @@ hash_xlog_vacuum_one_page(XLogReaderState *record)
    Buffer metabuf;
    Page page;
    XLogRedoAction action;
+   HashPageOpaque pageopaque;
 
    xldata = (xl_hash_vacuum_one_page *) XLogRecGetData(record);
 
@@ -1126,6 +1140,13 @@ hash_xlog_vacuum_one_page(XLogReaderState *record)
                PageIndexMultiDelete(page, unused, unend - unused);
        }
 
+       /*
+        * Mark the page as not containing any LP_DEAD items. See comments
+        * in _hash_vacuum_one_page() for details.
+        */
+       pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+       pageopaque->hasho_flag &= ~LH_PAGE_HAS_DEAD_TUPLES;
+
        PageSetLSN(page, lsn);
        MarkBufferDirty(buffer);
    }
index 8b6d0a0ff7821bbbe2717369e21f2108d06a9e65..8640e85a5c6b62960ebd4352c73e92a5bfbfbdfc 100644 (file)
@@ -374,6 +374,14 @@ _hash_vacuum_one_page(Relation rel, Buffer metabuf, Buffer buf,
 
        PageIndexMultiDelete(page, deletable, ndeletable);
 
+       /*
+        * Mark the page as not containing any LP_DEAD items. This is not
+        * certainly true (there might be some that have recently been marked,
+        * but weren't included in our target-item list), but it will almost
+        * always be true and it doesn't seem worth an additional page scan
+        * to check it. Remember that LH_PAGE_HAS_DEAD_TUPLES is only a hint
+        * anyway.
+        */
        pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
        pageopaque->hasho_flag &= ~LH_PAGE_HAS_DEAD_TUPLES;
 
index 5bd5c8dc0103cdf5902f32fd00bbd131d90529df..5f5f4a0255192d9cf11fc638acde9d7115af4ca7 100644 (file)
@@ -96,7 +96,8 @@ hash_desc(StringInfo buf, XLogReaderState *record)
            {
                xl_hash_delete *xlrec = (xl_hash_delete *) rec;
 
-               appendStringInfo(buf, "is_primary %c",
+               appendStringInfo(buf, "clear_dead_marking %c, is_primary %c",
+                                xlrec->clear_dead_marking ? 'T' : 'F',
                                 xlrec->is_primary_bucket_page ? 'T' : 'F');
                break;
            }
@@ -104,6 +105,14 @@ hash_desc(StringInfo buf, XLogReaderState *record)
            {
                xl_hash_update_meta_page *xlrec = (xl_hash_update_meta_page *) rec;
 
+               appendStringInfo(buf, "ntuples %g",
+                                xlrec->ntuples);
+               break;
+           }
+       case XLOG_HASH_VACUUM_ONE_PAGE:
+           {
+               xl_hash_vacuum_one_page *xlrec = (xl_hash_vacuum_one_page *) rec;
+
                appendStringInfo(buf, "ntuples %g",
                                 xlrec->ntuples);
                break;
index dfd923781997e4586b90c492c2dff700d79fc40c..2e64cfa3eafc477bc64550b19c74a81e7731b98e 100644 (file)
@@ -197,6 +197,8 @@ typedef struct xl_hash_squeeze_page
  */
 typedef struct xl_hash_delete
 {
+   bool        clear_dead_marking;     /* TRUE if this operation clears
+                                        * LH_PAGE_HAS_DEAD_TUPLES flag */
    bool        is_primary_bucket_page; /* TRUE if the operation is for
                                         * primary bucket page */
 }  xl_hash_delete;
index b8b15f18752cf460c559bbcb63bf15825febb1a7..7957cab98c62731b64d0dbf6f800589594aa171d 100644 (file)
@@ -31,7 +31,7 @@
 /*
  * Each page of XLOG file has a header like this:
  */
-#define XLOG_PAGE_MAGIC 0xD095 /* can be used as WAL version indicator */
+#define XLOG_PAGE_MAGIC 0xD096 /* can be used as WAL version indicator */
 
 typedef struct XLogPageHeaderData
 {