diff --git a/doc/src/sgml/ref/reindex.sgml b/doc/src/sgml/ref/reindex.sgml index 7222665..051ebd7 100644 --- a/doc/src/sgml/ref/reindex.sgml +++ b/doc/src/sgml/ref/reindex.sgml @@ -21,7 +21,7 @@ PostgreSQL documentation -REINDEX { INDEX | TABLE | DATABASE | SYSTEM } name [ FORCE ] +REINDEX { INDEX | TABLE | DATABASE | SYSTEM } [ CONCURRENTLY ] name [ FORCE ] @@ -68,9 +68,12 @@ REINDEX { INDEX | TABLE | DATABASE | SYSTEM } nam An index build with the CONCURRENTLY option failed, leaving an invalid index. Such indexes are useless but it can be convenient to use REINDEX to rebuild them. Note that - REINDEX will not perform a concurrent build. To build the - index without interfering with production you should drop the index and - reissue the CREATE INDEX CONCURRENTLY command. + REINDEX will perform a concurrent build if + CONCURRENTLY is specified. To build the index without interfering + with production you should drop the index and reissue either the + CREATE INDEX CONCURRENTLY or REINDEX CONCURRENTLY + command. Indexes of toast relations can be rebuilt with REINDEX + CONCURRENTLY. @@ -139,6 +142,21 @@ REINDEX { INDEX | TABLE | DATABASE | SYSTEM } nam + CONCURRENTLY + + + When this option is used, PostgreSQL will rebuild the + index without taking any locks that prevent concurrent inserts, + updates, or deletes on the table; whereas a standard reindex build + locks out writes (but not reads) on the table until it's done. + There are several caveats to be aware of when using this option + — see . + + + + + FORCE @@ -231,6 +249,112 @@ REINDEX { INDEX | TABLE | DATABASE | SYSTEM } nam to be reindexed by separate commands. This is still possible, but redundant. + + + + Rebuilding Indexes Concurrently + + + index + rebuilding concurrently + + + + Rebuilding an index can interfere with regular operation of a database. + Normally PostgreSQL locks the table whose index is rebuilt + against writes and performs the entire index build with a single scan of the + table. Other transactions can still read the table, but if they try to + insert, update, or delete rows in the table they will block until the + index rebuild is finished. This could have a severe effect if the system is + a live production database. Very large tables can take many hours to be + indexed, and even for smaller tables, an index rebuild can lock out writers + for periods that are unacceptably long for a production system. + + + + PostgreSQL supports rebuilding indexes without locking + out writes. This method is invoked by specifying the + CONCURRENTLY option of REINDEX. + When this option is used, PostgreSQL must perform two + scans of the table for each index that needs to be rebuild and in + addition it must wait for all existing transactions that could potentially + use the index to terminate. This method requires more total work than a + standard index rebuild and takes significantly longer to complete as it + needs to wait for unfinished transactions that might modify the index. + However, since it allows normal operations to continue while the index + is rebuilt, this method is useful for rebuilding indexes in a production + environment. Of course, the extra CPU, memory and I/O load imposed by + the index rebuild might slow other operations. + + + + In a concurrent index build, a new index whose storage will replace the one + to be rebuild is actually entered into the system catalogs in one transaction, + then two table scans occur in two more transactions and to make the new + index valid from the other backends. Once this is performed, the old + and fresh indexes are swapped in, and the index used during process is + marked as invalid in a third transaction. Finally two additional + transactions are used to mark the concurrent index as not ready and then + drop it. + + + + If a problem arises while rebuilding the indexes, such as a + uniqueness violation in a unique index, the REINDEX + command will fail but leave behind an invalid new index on top + of the existing one. This index will be ignored for querying purposes + because it might be incomplete; however it will still consume update + overhead. The psql \d command will report + such an index as INVALID: + + +postgres=# \d tab + Table "public.tab" + Column | Type | Modifiers +--------+---------+----------- + col | integer | +Indexes: + "idx" btree (col) + "idx_cct" btree (col) INVALID + + + The recommended recovery method in such cases is to drop the concurrent + index and try again to perform REINDEX CONCURRENTLY. + The concurrent index created during the processing has a name finishing by + the suffix cct. This works as well with indexes of toast relations. + + + + Regular index builds permit other regular index builds on the + same table to occur in parallel, but only one concurrent index build + can occur on a table at a time. In both cases, no other types of schema + modification on the table are allowed meanwhile. Another difference + is that a regular REINDEX TABLE or REINDEX INDEX + command can be performed within a transaction block, but + REINDEX CONCURRENTLY cannot. REINDEX DATABASE is + by default not allowed to run inside a transaction block, so in this case + CONCURRENTLY is not supported. + + + + Invalid indexes of toast relations can be dropped if a failure occurred + during REINDEX CONCURRENTLY. Live indexes of toast relations + cannot be dropped. + + + + REINDEX DATABASE used with CONCURRENTLY + rebuilds concurrently only the non-system relations. System + relations are rebuilt with a non-concurrent context. Toast indexes are + rebuilt concurrently if the relation they depend on is a non-system + relation. + + + + REINDEX SYSTEM does not support CONCURRENTLY + . + + @@ -262,7 +386,17 @@ $ psql broken_db ... broken_db=> REINDEX DATABASE broken_db; broken_db=> \q - + + + + + Rebuild a table concurrently: + + +REINDEX TABLE CONCURRENTLY my_broken_table; + + + diff --git a/src/backend/catalog/index.c b/src/backend/catalog/index.c index 0f3b45f..f72efbb 100644 --- a/src/backend/catalog/index.c +++ b/src/backend/catalog/index.c @@ -43,9 +43,11 @@ #include "catalog/pg_trigger.h" #include "catalog/pg_type.h" #include "catalog/storage.h" +#include "commands/defrem.h" #include "commands/tablecmds.h" #include "commands/trigger.h" #include "executor/executor.h" +#include "mb/pg_wchar.h" #include "miscadmin.h" #include "nodes/makefuncs.h" #include "nodes/nodeFuncs.h" @@ -672,6 +674,10 @@ UpdateIndexRelation(Oid indexoid, * will be marked "invalid" and the caller must take additional steps * to fix it up. * is_internal: if true, post creation hook for new index + * is_reindex: if true, create an index that is used as a duplicate of an + * existing index created during a concurrent operation. This index can + * also be a toast relation. Sufficient locks are normally taken on + * the related relations once this is called during a concurrent operation. * * Returns the OID of the created index. */ @@ -695,7 +701,8 @@ index_create(Relation heapRelation, bool allow_system_table_mods, bool skip_build, bool concurrent, - bool is_internal) + bool is_internal, + bool is_reindex) { Oid heapRelationId = RelationGetRelid(heapRelation); Relation pg_class; @@ -738,19 +745,23 @@ index_create(Relation heapRelation, /* * concurrent index build on a system catalog is unsafe because we tend to - * release locks before committing in catalogs + * release locks before committing in catalogs. If the index is created during + * a REINDEX CONCURRENTLY operation, sufficient locks are already taken. */ if (concurrent && - IsSystemRelation(heapRelation)) + IsSystemRelation(heapRelation) && + !is_reindex) ereport(ERROR, (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), errmsg("concurrent index creation on system catalog tables is not supported"))); /* * This case is currently not supported, but there's no way to ask for it - * in the grammar anyway, so it can't happen. + * in the grammar anyway, so it can't happen. This might be called during a + * conccurrent reindex operation, in this case sufficient locks are already + * taken on the related relations. */ - if (concurrent && is_exclusion) + if (concurrent && is_exclusion && !is_reindex) ereport(ERROR, (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), errmsg_internal("concurrent index creation for exclusion constraints is not supported"))); @@ -1095,6 +1106,427 @@ index_create(Relation heapRelation, return indexRelationId; } + +/* + * index_concurrent_create + * + * Create an index based on the given one that will be used for concurrent + * operations. The index is inserted into catalogs and needs to be built later + * on. This is called during concurrent index processing. The heap relation + * on which is based the index needs to be closed by the caller. + */ +Oid +index_concurrent_create(Relation heapRelation, Oid indOid, char *concurrentName) +{ + Relation indexRelation; + IndexInfo *indexInfo; + Oid concurrentOid = InvalidOid; + List *columnNames = NIL; + List *indexprs = NIL; + ListCell *indexpr_item; + int i; + HeapTuple indexTuple, classTuple; + Datum indclassDatum, colOptionDatum, optionDatum; + oidvector *indclass; + int2vector *indcoloptions; + bool isnull; + bool isconstraint; + bool initdeferred = false; + Oid constraintOid = get_index_constraint(indOid); + + indexRelation = index_open(indOid, RowExclusiveLock); + + /* Concurrent index uses the same index information as former index */ + indexInfo = BuildIndexInfo(indexRelation); + + /* + * Determine if index is initdeferred, this depends on its dependent + * constraint. + */ + if (OidIsValid(constraintOid)) + { + /* Look for the correct value */ + HeapTuple constTuple; + Form_pg_constraint constraint; + + constTuple = SearchSysCache1(CONSTROID, + ObjectIdGetDatum(constraintOid)); + if (!HeapTupleIsValid(constTuple)) + elog(ERROR, "cache lookup failed for constraint %u", + constraintOid); + constraint = (Form_pg_constraint) GETSTRUCT(constTuple); + initdeferred = constraint->condeferred; + + ReleaseSysCache(constTuple); + } + + /* Get expressions associated to this index for compilation of column names */ + indexprs = RelationGetIndexExpressions(indexRelation); + indexpr_item = list_head(indexprs); + + /* Build the list of column names, necessary for index_create */ + for (i = 0; i < indexInfo->ii_NumIndexAttrs; i++) + { + char *origname, *curname; + int i; + char buf[NAMEDATALEN]; + + AttrNumber attnum = indexInfo->ii_KeyAttrNumbers[i]; + + /* Pick up column name depending on attribute type */ + if (attnum != 0) + { + /* + * This is a column attribute, so simply pick column name from + * relation. + */ + Form_pg_attribute attform = heapRelation->rd_att->attrs[attnum - 1];; + origname = pstrdup(NameStr(attform->attname)); + } + else + { + Node *indnode; + /* + * This is the case of an expression, so pick up the expression + * name. + */ + Assert(indexpr_item != NULL); + indnode = (Node *) lfirst(indexpr_item); + indexpr_item = lnext(indexpr_item); + origname = deparse_expression(indnode, + deparse_context_for(RelationGetRelationName(heapRelation), + RelationGetRelid(heapRelation)), + false, false); + } + + /* + * Check if the name picked has any conflict with exising names and + * change it. + */ + curname = origname; + for (i = 1;; i++) + { + ListCell *lc2; + char nbuf[32]; + int nlen; + + foreach(lc2, columnNames) + { + if (strcmp(curname, (char *) lfirst(lc2)) == 0) + break; + } + if (lc2 == NULL) + break; /* found nonconflicting name */ + + sprintf(nbuf, "%d", i); + + /* Ensure generated names are shorter than NAMEDATALEN */ + nlen = pg_mbcliplen(origname, strlen(origname), + NAMEDATALEN - 1 - strlen(nbuf)); + memcpy(buf, origname, nlen); + strcpy(buf + nlen, nbuf); + curname = buf; + } + + /* Append name to existing list */ + columnNames = lappend(columnNames, pstrdup(curname)); + } + + /* + * Index is considered as a constraint if it is PRIMARY KEY or EXCLUSION. + */ + isconstraint = indexRelation->rd_index->indisprimary || + indexRelation->rd_index->indisexclusion; + + /* Get the array of class and column options IDs from index info */ + indexTuple = SearchSysCache1(INDEXRELID, ObjectIdGetDatum(indOid)); + if (!HeapTupleIsValid(indexTuple)) + elog(ERROR, "cache lookup failed for index %u", indOid); + indclassDatum = SysCacheGetAttr(INDEXRELID, indexTuple, + Anum_pg_index_indclass, &isnull); + Assert(!isnull); + indclass = (oidvector *) DatumGetPointer(indclassDatum); + + colOptionDatum = SysCacheGetAttr(INDEXRELID, indexTuple, + Anum_pg_index_indoption, &isnull); + Assert(!isnull); + indcoloptions = (int2vector *) DatumGetPointer(colOptionDatum); + + /* Fetch options of index if any */ + classTuple = SearchSysCache1(RELOID, indOid); + if (!HeapTupleIsValid(classTuple)) + elog(ERROR, "cache lookup failed for relation %u", indOid); + optionDatum = SysCacheGetAttr(RELOID, classTuple, + Anum_pg_class_reloptions, &isnull); + + /* Now create the concurrent index */ + concurrentOid = index_create(heapRelation, + (const char*)concurrentName, + InvalidOid, + InvalidOid, + indexInfo, + columnNames, + indexRelation->rd_rel->relam, + indexRelation->rd_rel->reltablespace, + indexRelation->rd_indcollation, + indclass->values, + indcoloptions->values, + optionDatum, + indexRelation->rd_index->indisprimary, + isconstraint, /* is constraint? */ + !indexRelation->rd_index->indimmediate, /* is deferrable? */ + initdeferred, /* is initially deferred? */ + true, /* allow table to be a system catalog? */ + true, /* skip build? */ + true, /* concurrent? */ + false, /* is_internal */ + true); /* reindex? */ + + /* Close the relations used and clean up */ + index_close(indexRelation, RowExclusiveLock); + ReleaseSysCache(indexTuple); + ReleaseSysCache(classTuple); + + return concurrentOid; +} + + +/* + * index_concurrent_build + * + * Build index for a concurrent operation. Low-level locks are taken when this + * operation is performed to prevent only schema changes. + */ +void +index_concurrent_build(Oid heapOid, + Oid indexOid, + bool isprimary) +{ + Relation rel, + indexRelation; + IndexInfo *indexInfo; + + /* Open and lock the parent heap relation */ + rel = heap_open(heapOid, ShareUpdateExclusiveLock); + + /* And the target index relation */ + indexRelation = index_open(indexOid, RowExclusiveLock); + + /* We have to re-build the IndexInfo struct, since it was lost in commit */ + indexInfo = BuildIndexInfo(indexRelation); + Assert(!indexInfo->ii_ReadyForInserts); + indexInfo->ii_Concurrent = true; + indexInfo->ii_BrokenHotChain = false; + + /* Now build the index */ + index_build(rel, indexRelation, indexInfo, isprimary, false); + + /* Close both the relations, but keep the locks */ + heap_close(rel, NoLock); + index_close(indexRelation, NoLock); +} + + +/* + * index_concurrent_swap + * + * Replace old index by old index in a concurrent context. For the time being + * what is done here is switching the relation relfilenode of the indexes. If + * extra operations are necessary during a concurrent swap, processing should + * be added here. AccessExclusiveLock is taken on the index relations that are + * swapped until the end of the transaction where this function is called. + */ +void +index_concurrent_swap(Oid newIndexOid, Oid oldIndexOid) +{ + Relation oldIndexRel, newIndexRel, pg_class; + HeapTuple oldIndexTuple, newIndexTuple; + Form_pg_class oldIndexForm, newIndexForm; + Oid tmpnode; + + /* + * Take an exclusive lock on the old and new index before swapping them. + */ + oldIndexRel = relation_open(oldIndexOid, AccessExclusiveLock); + newIndexRel = relation_open(newIndexOid, AccessExclusiveLock); + + /* Now swap relfilenode of those indexes */ + pg_class = heap_open(RelationRelationId, RowExclusiveLock); + + oldIndexTuple = SearchSysCacheCopy1(RELOID, + ObjectIdGetDatum(oldIndexOid)); + if (!HeapTupleIsValid(oldIndexTuple)) + elog(ERROR, "could not find tuple for relation %u", oldIndexOid); + newIndexTuple = SearchSysCacheCopy1(RELOID, + ObjectIdGetDatum(newIndexOid)); + if (!HeapTupleIsValid(newIndexTuple)) + elog(ERROR, "could not find tuple for relation %u", newIndexOid); + oldIndexForm = (Form_pg_class) GETSTRUCT(oldIndexTuple); + newIndexForm = (Form_pg_class) GETSTRUCT(newIndexTuple); + + /* Here is where the actual swapping happens */ + tmpnode = oldIndexForm->relfilenode; + oldIndexForm->relfilenode = newIndexForm->relfilenode; + newIndexForm->relfilenode = tmpnode; + + /* Then update the tuples for each relation */ + simple_heap_update(pg_class, &oldIndexTuple->t_self, oldIndexTuple); + simple_heap_update(pg_class, &newIndexTuple->t_self, newIndexTuple); + CatalogUpdateIndexes(pg_class, oldIndexTuple); + CatalogUpdateIndexes(pg_class, newIndexTuple); + + /* Close relations and clean up */ + heap_close(pg_class, RowExclusiveLock); + + /* The lock taken previously is not released until the end of transaction */ + relation_close(oldIndexRel, NoLock); + relation_close(newIndexRel, NoLock); +} + +/* + * index_concurrent_set_dead + * + * Perform the last invalidation stage of DROP INDEX CONCURRENTLY before + * actually dropping the index. After calling this function the index is + * seen by all the backends as dead. + */ +void +index_concurrent_set_dead(Oid indexId, Oid heapId, LOCKTAG *locktag) +{ + Relation heapRelation; + Relation indexRelation; + + /* + * Now we must wait until no running transaction could be using the + * index for a query if necessary. + * + * Note: the reason we use actual lock acquisition here, rather than + * just checking the ProcArray and sleeping, is that deadlock is + * possible if one of the transactions in question is blocked trying + * to acquire an exclusive lock on our table. The lock code will + * detect deadlock and error out properly. + */ + if (locktag) + WaitForVirtualLocks(*locktag, AccessExclusiveLock); + + /* + * No more predicate locks will be acquired on this index, and we're + * about to stop doing inserts into the index which could show + * conflicts with existing predicate locks, so now is the time to move + * them to the heap relation. + */ + heapRelation = heap_open(heapId, ShareUpdateExclusiveLock); + indexRelation = index_open(indexId, ShareUpdateExclusiveLock); + TransferPredicateLocksToHeapRelation(indexRelation); + + /* + * Now we are sure that nobody uses the index for queries; they just + * might have it open for updating it. So now we can unset indisready + * and indislive, then wait till nobody could be using it at all + * anymore. + */ + index_set_state_flags(indexId, INDEX_DROP_SET_DEAD); + + /* + * Invalidate the relcache for the table, so that after this commit + * all sessions will refresh the table's index list. Forgetting just + * the index's relcache entry is not enough. + */ + CacheInvalidateRelcache(heapRelation); + + /* + * Close the relations again, though still holding session lock. + */ + heap_close(heapRelation, NoLock); + index_close(indexRelation, NoLock); +} + +/* + * index_concurrent_clear_valid + * + * Release the valid state of a given index and then release the cache of + * its parent relation. This function should be called when initializing an + * index drop in a concurrent context before setting the index as dead. + */ +void +index_concurrent_clear_valid(Relation heapRelation, Oid indexOid) +{ + /* + * Mark index invalid by updating its pg_index entry + */ + index_set_state_flags(indexOid, INDEX_DROP_CLEAR_VALID); + + /* + * Invalidate the relcache for the table, so that after this commit + * all sessions will refresh any cached plans that might reference the + * index. + */ + CacheInvalidateRelcache(heapRelation); +} + +/* + * index_concurrent_drop + * + * Drop a single index concurrently as the last step of an index concurrent + * process Deletion is done through performDeletion or dependencies of the + * index are not dropped. At this point all the indexes are already considered + * as invalid and dead so they can be dropped without using any concurrent + * options. + */ +void +index_concurrent_drop(Oid indexOid) +{ + Oid constraintOid = get_index_constraint(indexOid); + ObjectAddress object; + Form_pg_index indexForm; + Relation pg_index; + HeapTuple indexTuple; + bool indislive; + + /* + * Check that the index dropped here is not alive, it might be used by + * other backends in this case. + */ + pg_index = heap_open(IndexRelationId, RowExclusiveLock); + + indexTuple = SearchSysCacheCopy1(INDEXRELID, + ObjectIdGetDatum(indexOid)); + if (!HeapTupleIsValid(indexTuple)) + elog(ERROR, "cache lookup failed for index %u", indexOid); + indexForm = (Form_pg_index) GETSTRUCT(indexTuple); + indislive = indexForm->indislive; + + /* Clean up */ + heap_close(pg_index, RowExclusiveLock); + + /* Leave if index is still alive */ + if (indislive) + return; + + /* + * We are sure to have a dead index, so begin the drop process. + * Register constraint or index for drop. + */ + if (OidIsValid(constraintOid)) + { + object.classId = ConstraintRelationId; + object.objectId = constraintOid; + } + else + { + object.classId = RelationRelationId; + object.objectId = indexOid; + } + + object.objectSubId = 0; + + /* Perform deletion for normal and toast indexes */ + performDeletion(&object, + DROP_RESTRICT, + 0); +} + + /* * index_constraint_create * @@ -1324,7 +1756,6 @@ index_drop(Oid indexId, bool concurrent) indexrelid; LOCKTAG heaplocktag; LOCKMODE lockmode; - VirtualTransactionId *old_lockholders; /* * To drop an index safely, we must grab exclusive lock on its parent @@ -1406,17 +1837,8 @@ index_drop(Oid indexId, bool concurrent) (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), errmsg("DROP INDEX CONCURRENTLY must be first action in transaction"))); - /* - * Mark index invalid by updating its pg_index entry - */ - index_set_state_flags(indexId, INDEX_DROP_CLEAR_VALID); - - /* - * Invalidate the relcache for the table, so that after this commit - * all sessions will refresh any cached plans that might reference the - * index. - */ - CacheInvalidateRelcache(userHeapRelation); + /* Mark the index as invalid */ + index_concurrent_clear_valid(userHeapRelation, indexId); /* save lockrelid and locktag for below, then close but keep locks */ heaprelid = userHeapRelation->rd_lockInfo.lockRelId; @@ -1444,63 +1866,8 @@ index_drop(Oid indexId, bool concurrent) CommitTransactionCommand(); StartTransactionCommand(); - /* - * Now we must wait until no running transaction could be using the - * index for a query. To do this, inquire which xacts currently would - * conflict with AccessExclusiveLock on the table -- ie, which ones - * have a lock of any kind on the table. Then wait for each of these - * xacts to commit or abort. Note we do not need to worry about xacts - * that open the table for reading after this point; they will see the - * index as invalid when they open the relation. - * - * Note: the reason we use actual lock acquisition here, rather than - * just checking the ProcArray and sleeping, is that deadlock is - * possible if one of the transactions in question is blocked trying - * to acquire an exclusive lock on our table. The lock code will - * detect deadlock and error out properly. - * - * Note: GetLockConflicts() never reports our own xid, hence we need - * not check for that. Also, prepared xacts are not reported, which - * is fine since they certainly aren't going to do anything more. - */ - old_lockholders = GetLockConflicts(&heaplocktag, AccessExclusiveLock); - - while (VirtualTransactionIdIsValid(*old_lockholders)) - { - VirtualXactLock(*old_lockholders, true); - old_lockholders++; - } - - /* - * No more predicate locks will be acquired on this index, and we're - * about to stop doing inserts into the index which could show - * conflicts with existing predicate locks, so now is the time to move - * them to the heap relation. - */ - userHeapRelation = heap_open(heapId, ShareUpdateExclusiveLock); - userIndexRelation = index_open(indexId, ShareUpdateExclusiveLock); - TransferPredicateLocksToHeapRelation(userIndexRelation); - - /* - * Now we are sure that nobody uses the index for queries; they just - * might have it open for updating it. So now we can unset indisready - * and indislive, then wait till nobody could be using it at all - * anymore. - */ - index_set_state_flags(indexId, INDEX_DROP_SET_DEAD); - - /* - * Invalidate the relcache for the table, so that after this commit - * all sessions will refresh the table's index list. Forgetting just - * the index's relcache entry is not enough. - */ - CacheInvalidateRelcache(userHeapRelation); - - /* - * Close the relations again, though still holding session lock. - */ - heap_close(userHeapRelation, NoLock); - index_close(userIndexRelation, NoLock); + /* Finish invalidation of index and mark it as dead */ + index_concurrent_set_dead(indexId, heapId, &heaplocktag); /* * Again, commit the transaction to make the pg_index update visible @@ -1513,13 +1880,7 @@ index_drop(Oid indexId, bool concurrent) * Wait till every transaction that saw the old index state has * finished. The logic here is the same as above. */ - old_lockholders = GetLockConflicts(&heaplocktag, AccessExclusiveLock); - - while (VirtualTransactionIdIsValid(*old_lockholders)) - { - VirtualXactLock(*old_lockholders, true); - old_lockholders++; - } + WaitForVirtualLocks(heaplocktag, AccessExclusiveLock); /* * Re-open relations to allow us to complete our actions. diff --git a/src/backend/catalog/toasting.c b/src/backend/catalog/toasting.c index 385d64d..0c2971b 100644 --- a/src/backend/catalog/toasting.c +++ b/src/backend/catalog/toasting.c @@ -281,7 +281,7 @@ create_toast_table(Relation rel, Oid toastOid, Oid toastIndexOid, Datum reloptio rel->rd_rel->reltablespace, collationObjectId, classObjectId, coloptions, (Datum) 0, true, false, false, false, - true, false, false, true); + true, false, false, false, false); heap_close(toast_rel, NoLock); diff --git a/src/backend/commands/indexcmds.c b/src/backend/commands/indexcmds.c index f855bef..a12dcb9 100644 --- a/src/backend/commands/indexcmds.c +++ b/src/backend/commands/indexcmds.c @@ -68,8 +68,9 @@ static void ComputeIndexAttrs(IndexInfo *indexInfo, static Oid GetIndexOpClass(List *opclass, Oid attrType, char *accessMethodName, Oid accessMethodId); static char *ChooseIndexName(const char *tabname, Oid namespaceId, - List *colnames, List *exclusionOpNames, - bool primary, bool isconstraint); + List *colnames, List *exclusionOpNames, + bool primary, bool isconstraint, + bool concurrent); static char *ChooseIndexNameAddition(List *colnames); static List *ChooseIndexColumnNames(List *indexElems); static void RangeVarCallbackForReindexIndex(const RangeVar *relation, @@ -311,7 +312,6 @@ DefineIndex(IndexStmt *stmt, Oid tablespaceId; List *indexColNames; Relation rel; - Relation indexRelation; HeapTuple tuple; Form_pg_am accessMethodForm; bool amcanorder; @@ -320,13 +320,9 @@ DefineIndex(IndexStmt *stmt, int16 *coloptions; IndexInfo *indexInfo; int numberOfAttributes; - VirtualTransactionId *old_lockholders; - VirtualTransactionId *old_snapshots; - int n_old_snapshots; LockRelId heaprelid; LOCKTAG heaplocktag; Snapshot snapshot; - int i; /* * count attributes in index @@ -453,7 +449,8 @@ DefineIndex(IndexStmt *stmt, indexColNames, stmt->excludeOpNames, stmt->primary, - stmt->isconstraint); + stmt->isconstraint, + false); /* * look up the access method, verify it can handle the requested features @@ -600,7 +597,7 @@ DefineIndex(IndexStmt *stmt, stmt->isconstraint, stmt->deferrable, stmt->initdeferred, allowSystemTableMods, skip_build || stmt->concurrent, - stmt->concurrent, !check_rights); + stmt->concurrent, !check_rights, false); /* Add any requested comment */ if (stmt->idxcomment != NULL) @@ -663,18 +660,8 @@ DefineIndex(IndexStmt *stmt, * one of the transactions in question is blocked trying to acquire an * exclusive lock on our table. The lock code will detect deadlock and * error out properly. - * - * Note: GetLockConflicts() never reports our own xid, hence we need not - * check for that. Also, prepared xacts are not reported, which is fine - * since they certainly aren't going to do anything more. */ - old_lockholders = GetLockConflicts(&heaplocktag, ShareLock); - - while (VirtualTransactionIdIsValid(*old_lockholders)) - { - VirtualXactLock(*old_lockholders, true); - old_lockholders++; - } + WaitForVirtualLocks(heaplocktag, ShareLock); /* * At this moment we are sure that there are no transactions with the @@ -694,27 +681,13 @@ DefineIndex(IndexStmt *stmt, * HOT-chain or the extension of the chain is HOT-safe for this index. */ - /* Open and lock the parent heap relation */ - rel = heap_openrv(stmt->relation, ShareUpdateExclusiveLock); - - /* And the target index relation */ - indexRelation = index_open(indexRelationId, RowExclusiveLock); - /* Set ActiveSnapshot since functions in the indexes may need it */ PushActiveSnapshot(GetTransactionSnapshot()); - /* We have to re-build the IndexInfo struct, since it was lost in commit */ - indexInfo = BuildIndexInfo(indexRelation); - Assert(!indexInfo->ii_ReadyForInserts); - indexInfo->ii_Concurrent = true; - indexInfo->ii_BrokenHotChain = false; - - /* Now build the index */ - index_build(rel, indexRelation, indexInfo, stmt->primary, false); - - /* Close both the relations, but keep the locks */ - heap_close(rel, NoLock); - index_close(indexRelation, NoLock); + /* Perform concurrent build of index */ + index_concurrent_build(RangeVarGetRelid(stmt->relation, NoLock, false), + indexRelationId, + stmt->primary); /* * Update the pg_index row to mark the index as ready for inserts. Once we @@ -738,13 +711,7 @@ DefineIndex(IndexStmt *stmt, * We once again wait until no transaction can have the table open with * the index marked as read-only for updates. */ - old_lockholders = GetLockConflicts(&heaplocktag, ShareLock); - - while (VirtualTransactionIdIsValid(*old_lockholders)) - { - VirtualXactLock(*old_lockholders, true); - old_lockholders++; - } + WaitForVirtualLocks(heaplocktag, ShareLock); /* * Now take the "reference snapshot" that will be used by validate_index() @@ -773,74 +740,9 @@ DefineIndex(IndexStmt *stmt, * The index is now valid in the sense that it contains all currently * interesting tuples. But since it might not contain tuples deleted just * before the reference snap was taken, we have to wait out any - * transactions that might have older snapshots. Obtain a list of VXIDs - * of such transactions, and wait for them individually. - * - * We can exclude any running transactions that have xmin > the xmin of - * our reference snapshot; their oldest snapshot must be newer than ours. - * We can also exclude any transactions that have xmin = zero, since they - * evidently have no live snapshot at all (and any one they might be in - * process of taking is certainly newer than ours). Transactions in other - * DBs can be ignored too, since they'll never even be able to see this - * index. - * - * We can also exclude autovacuum processes and processes running manual - * lazy VACUUMs, because they won't be fazed by missing index entries - * either. (Manual ANALYZEs, however, can't be excluded because they - * might be within transactions that are going to do arbitrary operations - * later.) - * - * Also, GetCurrentVirtualXIDs never reports our own vxid, so we need not - * check for that. - * - * If a process goes idle-in-transaction with xmin zero, we do not need to - * wait for it anymore, per the above argument. We do not have the - * infrastructure right now to stop waiting if that happens, but we can at - * least avoid the folly of waiting when it is idle at the time we would - * begin to wait. We do this by repeatedly rechecking the output of - * GetCurrentVirtualXIDs. If, during any iteration, a particular vxid - * doesn't show up in the output, we know we can forget about it. + * transactions that might have older snapshots. */ - old_snapshots = GetCurrentVirtualXIDs(snapshot->xmin, true, false, - PROC_IS_AUTOVACUUM | PROC_IN_VACUUM, - &n_old_snapshots); - - for (i = 0; i < n_old_snapshots; i++) - { - if (!VirtualTransactionIdIsValid(old_snapshots[i])) - continue; /* found uninteresting in previous cycle */ - - if (i > 0) - { - /* see if anything's changed ... */ - VirtualTransactionId *newer_snapshots; - int n_newer_snapshots; - int j; - int k; - - newer_snapshots = GetCurrentVirtualXIDs(snapshot->xmin, - true, false, - PROC_IS_AUTOVACUUM | PROC_IN_VACUUM, - &n_newer_snapshots); - for (j = i; j < n_old_snapshots; j++) - { - if (!VirtualTransactionIdIsValid(old_snapshots[j])) - continue; /* found uninteresting in previous cycle */ - for (k = 0; k < n_newer_snapshots; k++) - { - if (VirtualTransactionIdEquals(old_snapshots[j], - newer_snapshots[k])) - break; - } - if (k >= n_newer_snapshots) /* not there anymore */ - SetInvalidVirtualTransactionId(old_snapshots[j]); - } - pfree(newer_snapshots); - } - - if (VirtualTransactionIdIsValid(old_snapshots[i])) - VirtualXactLock(old_snapshots[i], true); - } + WaitForOldSnapshots(snapshot); /* * Index can now be marked valid -- update its pg_index entry @@ -853,7 +755,7 @@ DefineIndex(IndexStmt *stmt, * relcache inval on the parent table to force replanning of cached plans. * Otherwise existing sessions might fail to use the new index where it * would be useful. (Note that our earlier commits did not create reasons - * to replan; so relcache flush on the index itself was sufficient.) + * to replan; relcache flush on the index itself was sufficient.) */ CacheInvalidateRelcacheByRelid(heaprelid.relId); @@ -873,6 +775,521 @@ DefineIndex(IndexStmt *stmt, /* + * ReindexRelationConcurrently + * + * Process REINDEX CONCURRENTLY for given relation Oid. The relation can be + * either an index or a table. If a table is specified, each reindexing step + * is done in parallel with all the table's indexes as well as its dependent + * toast indexes. + */ +bool +ReindexRelationConcurrently(Oid relationOid) +{ + List *concurrentIndexIds = NIL, + *indexIds = NIL, + *parentRelationIds = NIL, + *lockTags = NIL, + *relationLocks = NIL; + ListCell *lc, *lc2; + Snapshot snapshot; + + /* + * Extract the list of indexes that are going to be rebuilt based on the + * list of relation Oids given by caller. For each element in given list, + * If the relkind of given relation Oid is a table, all its valid indexes + * will be rebuilt, including its associated toast table indexes. If + * relkind is an index, this index itself will be rebuilt. The locks taken + * parent relations and involved indexes are kept until this transaction + * is committed to protect against schema changes that might occur until + * the session lock is taken on each relation. + */ + switch (get_rel_relkind(relationOid)) + { + case RELKIND_RELATION: + { + /* + * In the case of a relation, find all its indexes + * including toast indexes. + */ + Relation heapRelation = heap_open(relationOid, + ShareUpdateExclusiveLock); + + /* Track this relation for session locks */ + parentRelationIds = lappend_oid(parentRelationIds, relationOid); + + /* Relation on which is based index cannot be shared */ + if (heapRelation->rd_rel->relisshared) + ereport(ERROR, + (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), + errmsg("concurrent reindex is not supported for shared relations"))); + + /* Add all the valid indexes of relation to list */ + foreach(lc2, RelationGetIndexList(heapRelation)) + { + Oid cellOid = lfirst_oid(lc2); + Relation indexRelation = index_open(cellOid, + ShareUpdateExclusiveLock); + + if (!indexRelation->rd_index->indisvalid) + ereport(WARNING, + (errcode(ERRCODE_INDEX_CORRUPTED), + errmsg("cannot reindex concurrently invalid index \"%s.%s\", skipping", + get_namespace_name(get_rel_namespace(cellOid)), + get_rel_name(cellOid)))); + else + indexIds = lappend_oid(indexIds, cellOid); + + index_close(indexRelation, NoLock); + } + + /* Also add the toast indexes */ + if (OidIsValid(heapRelation->rd_rel->reltoastrelid)) + { + Oid toastOid = heapRelation->rd_rel->reltoastrelid; + Relation toastRelation = heap_open(toastOid, + ShareUpdateExclusiveLock); + + /* Track this relation for session locks */ + parentRelationIds = lappend_oid(parentRelationIds, toastOid); + + foreach(lc2, RelationGetIndexList(toastRelation)) + { + Oid cellOid = lfirst_oid(lc2); + Relation indexRelation = index_open(cellOid, + ShareUpdateExclusiveLock); + + if (!indexRelation->rd_index->indisvalid) + ereport(WARNING, + (errcode(ERRCODE_INDEX_CORRUPTED), + errmsg("cannot reindex concurrently invalid index \"%s.%s\", skipping", + get_namespace_name(get_rel_namespace(cellOid)), + get_rel_name(cellOid)))); + else + indexIds = lappend_oid(indexIds, cellOid); + + index_close(indexRelation, NoLock); + } + + heap_close(toastRelation, NoLock); + } + + heap_close(heapRelation, NoLock); + break; + } + case RELKIND_INDEX: + { + /* + * For an index simply add its Oid to list. Invalid indexes + * cannot be included in list. + */ + Relation indexRelation = index_open(relationOid, ShareUpdateExclusiveLock); + + /* Track the parent relation of this index for session locks */ + parentRelationIds = list_make1_oid(IndexGetRelation(relationOid, false)); + + if (!indexRelation->rd_index->indisvalid) + ereport(WARNING, + (errcode(ERRCODE_INDEX_CORRUPTED), + errmsg("cannot reindex concurrently invalid index \"%s.%s\", skipping", + get_namespace_name(get_rel_namespace(relationOid)), + get_rel_name(relationOid)))); + else + indexIds = list_make1_oid(relationOid); + + index_close(indexRelation, NoLock); + break; + } + default: + /* nothing to do */ + break; + } + + /* Definetely no indexes, so leave */ + if (indexIds == NIL) + return false; + + Assert(parentRelationIds != NIL); + + /* + * Phase 1 of REINDEX CONCURRENTLY + * + * Here begins the process for rebuilding concurrently the indexes. + * We need first to create an index which is based on the same data + * as the former index except that it will be only registered in catalogs + * and will be built after. It is possible to perform all the operations + * on all the indexes at the same time for a parent relation including + * its indexes for toast relation. + */ + + /* Do the concurrent index creation for each index */ + foreach(lc, indexIds) + { + char *concurrentName; + Oid indOid = lfirst_oid(lc); + Oid concurrentOid = InvalidOid; + Relation indexRel, + indexParentRel, + indexConcurrentRel; + LockRelId lockrelid; + + indexRel = index_open(indOid, ShareUpdateExclusiveLock); + /* Open the index parent relation, might be a toast or parent relation */ + indexParentRel = heap_open(indexRel->rd_index->indrelid, + ShareUpdateExclusiveLock); + + /* Choose a relation name for concurrent index */ + concurrentName = ChooseIndexName(get_rel_name(indOid), + get_rel_namespace(indexRel->rd_index->indrelid), + NULL, + false, + false, + false, + true); + + /* Create concurrent index based on given index */ + concurrentOid = index_concurrent_create(indexParentRel, + indOid, + concurrentName); + + /* + * Now open the relation of concurrent index, a lock is also needed on + * it + */ + indexConcurrentRel = index_open(concurrentOid, ShareUpdateExclusiveLock); + + /* Save the concurrent index Oid */ + concurrentIndexIds = lappend_oid(concurrentIndexIds, concurrentOid); + + /* + * Save lockrelid to protect each concurrent relation from drop then + * close relations. The lockrelid on parent relation is not taken here + * to avoid multiple locks taken on the same relation, instead we rely + * on parentRelationIds built earlier. + */ + lockrelid = indexRel->rd_lockInfo.lockRelId; + relationLocks = lappend(relationLocks, &lockrelid); + lockrelid = indexConcurrentRel->rd_lockInfo.lockRelId; + relationLocks = lappend(relationLocks, &lockrelid); + + index_close(indexRel, NoLock); + index_close(indexConcurrentRel, NoLock); + heap_close(indexParentRel, NoLock); + } + + /* + * Save the heap lock for following visibility checks with other backends + * might conflict with this session. + */ + foreach(lc, parentRelationIds) + { + Relation heapRelation = heap_open(lfirst_oid(lc), ShareUpdateExclusiveLock); + LockRelId lockrelid = heapRelation->rd_lockInfo.lockRelId; + LOCKTAG *heaplocktag = (LOCKTAG *) palloc(sizeof(LOCKTAG)); + + /* Add lockrelid of parent relation to the list of locked relations */ + relationLocks = lappend(relationLocks, &lockrelid); + + /* Save the LOCKTAG for this parent relation for the wait phase */ + SET_LOCKTAG_RELATION(*heaplocktag, lockrelid.dbId, lockrelid.relId); + lockTags = lappend(lockTags, heaplocktag); + + /* Close heap relation */ + heap_close(heapRelation, NoLock); + } + + /* + * For a concurrent build, it is necessary to make the catalog entries + * visible to the other transactions before actually building the index. + * This will prevent them from making incompatible HOT updates. The index + * is marked as not ready and invalid so as no other transactions will try + * to use it for INSERT or SELECT. + * + * Before committing, get a session level lock on the relation, the + * concurrent index and its copy to insure that none of them are dropped + * until the operation is done. + */ + foreach(lc, relationLocks) + { + LockRelId lockRel = * (LockRelId *) lfirst(lc); + LockRelationIdForSession(&lockRel, ShareUpdateExclusiveLock); + } + + PopActiveSnapshot(); + CommitTransactionCommand(); + + /* + * Phase 2 of REINDEX CONCURRENTLY + * + * Build concurrent indexes in a separate transaction for each index to + * avoid having open transactions for an unnecessary long time. A + * concurrent build is done for each concurrent index that will replace + * the old indexes. Before doing that, we need to wait on the parent + * relations until no running transactions could have the parent table + * of index open. + */ + + /* Perform a wait on all the session locks */ + StartTransactionCommand(); + WaitForMultipleVirtualLocks(lockTags, ShareLock); + CommitTransactionCommand(); + + /* Get the first element of concurrent index list */ + lc2 = list_head(concurrentIndexIds); + + foreach(lc, indexIds) + { + Relation indexRel; + Oid indOid = lfirst_oid(lc); + Oid concurrentOid = lfirst_oid(lc2); + bool primary; + + /* Move to next concurrent item */ + lc2 = lnext(lc2); + + /* Start new transaction for this index concurrent build */ + StartTransactionCommand(); + + /* Set ActiveSnapshot since functions in the indexes may need it */ + PushActiveSnapshot(GetTransactionSnapshot()); + + /* Index relation has been closed by previous commit, so reopen it */ + indexRel = index_open(indOid, ShareUpdateExclusiveLock); + primary = indexRel->rd_index->indisprimary; + index_close(indexRel, ShareUpdateExclusiveLock); + + /* Perform concurrent build of new index */ + index_concurrent_build(indexRel->rd_index->indrelid, + concurrentOid, + primary); + + /* + * Update the pg_index row of the concurrent index as ready for inserts. + * Once we commit this transaction, any new transactions that open the + * table must insert new entries into the index for insertions and + * non-HOT updates. + */ + index_set_state_flags(concurrentOid, INDEX_CREATE_SET_READY); + + /* we can do away with our snapshot */ + PopActiveSnapshot(); + + /* + * Commit this transaction to make the indisready update visible for + * concurrent index. + */ + CommitTransactionCommand(); + } + + + /* + * Phase 3 of REINDEX CONCURRENTLY + * + * During this phase the concurrent indexes catch up with the INSERT that + * might have occurred in the parent table and are marked as valid once done. + * + * We once again wait until no transaction can have the table open with + * the index marked as read-only for updates. Each index validation is done + * with a separate transaction to avoid opening transaction for an + * unnecessary too long time. + */ + + /* + * Perform a scan of each concurrent index with the heap, then insert + * any missing index entries. + */ + foreach(lc, concurrentIndexIds) + { + Oid indOid = lfirst_oid(lc); + Oid relOid; + + /* Open separate transaction to validate index */ + StartTransactionCommand(); + + /* Get the parent relation Oid */ + relOid = IndexGetRelation(indOid, false); + + /* + * Take the reference snapshot that will be used for the concurrent indexes + * validation. + */ + snapshot = RegisterSnapshot(GetTransactionSnapshot()); + PushActiveSnapshot(snapshot); + + /* Validate index, which might be a toast */ + validate_index(relOid, indOid, snapshot); + + /* + * This concurrent index is now valid as they contain all the tuples + * necessary. However, it might not have taken into account deleted tuples + * before the reference snapshot was taken, so we need to wait for the + * transactions that might have older snapshots than ours. + */ + WaitForOldSnapshots(snapshot); + + /* + * Concurrent index can now be marked as valid -- update pg_index + * entries. + */ + index_set_state_flags(indOid, INDEX_CREATE_SET_VALID); + + /* + * The pg_index update will cause backends to update its entries for the + * concurrent index but it is necessary to do the same thing for cache. + */ + CacheInvalidateRelcacheByRelid(relOid); + + /* we can now do away with our active snapshot */ + PopActiveSnapshot(); + + /* And we can remove the validating snapshot too */ + UnregisterSnapshot(snapshot); + + /* Commit this transaction to make the concurrent index valid */ + CommitTransactionCommand(); + } + + /* + * Phase 4 of REINDEX CONCURRENTLY + * + * Now that the concurrent indexes are valid and can be used, we need to + * swap each concurrent index with its corresponding old index. The old + * index is marked as invalid once this is done, making it not usable + * by other backends once its associated transaction is committed. + */ + + /* Get the first element is concurrent index list */ + lc2 = list_head(concurrentIndexIds); + + /* Swap the indexes and mark the indexes that have the old data as invalid */ + foreach(lc, indexIds) + { + Oid indOid = lfirst_oid(lc); + Oid concurrentOid = lfirst_oid(lc2); + Relation indexRel, indexParentRel; + + /* Move to next concurrent item */ + lc2 = lnext(lc2); + + /* + * Each index needs to be swapped in a separate transaction, so start + * a new one. + */ + StartTransactionCommand(); + + /* + * Mark the cache of associated relation as invalid, open relation + * relations. AccessExclusive Lock is taken here and not a lower lock + * to reduce likelihood of deadlock as ShareUpdateExclusiveLock is + * already taken within session. + */ + indexRel = index_open(indOid, ShareUpdateExclusiveLock); + indexParentRel = heap_open(indexRel->rd_index->indrelid, + ShareUpdateExclusiveLock); + + /* Mark the old index as invalid */ + index_concurrent_clear_valid(indexParentRel, concurrentOid); + + /* Swap old index and its concurrent */ + index_concurrent_swap(concurrentOid, indOid); + + /* + * Invalidate the relcache for the table, so that after this commit + * all sessions will refresh any cached plans that might reference the + * index. + */ + CacheInvalidateRelcache(indexParentRel); + + /* Close relations opened previously for cache invalidation */ + index_close(indexRel, ShareUpdateExclusiveLock); + heap_close(indexParentRel, ShareUpdateExclusiveLock); + + /* Commit this transaction and make old index invalidation visible */ + CommitTransactionCommand(); + } + + /* + * Phase 5 of REINDEX CONCURRENTLY + * + * The concurrent indexes now hold the old relfilenode of the other indexes + * transactions that might use them. Each operation is performed with a + * separate transaction. + */ + + /* Now mark the concurrent indexes as not ready */ + foreach(lc, concurrentIndexIds) + { + Oid indOid = lfirst_oid(lc); + Oid relOid; + + StartTransactionCommand(); + relOid = IndexGetRelation(indOid, false); + + /* + * Finish the index invalidation and set it as dead. It is not + * necessary to wait for virtual locks on the parent relation as it + * is already sure that this session holds sufficient locks.s + */ + index_concurrent_set_dead(indOid, relOid, NULL); + + /* Commit this transaction to make the update visible. */ + CommitTransactionCommand(); + } + + /* + * Phase 6 of REINDEX CONCURRENTLY + * + * Drop the concurrent indexes. This needs to be done through + * performDeletion or related dependencies will not be dropped for the old + * indexes. The internal mechanism of DROP INDEX CONCURRENTLY is not used + * as here the indexes are already considered as dead and invalid, so they + * will not be used by other backends. + */ + foreach(lc, concurrentIndexIds) + { + Oid indexOid = lfirst_oid(lc); + + /* Start transaction to drop this index */ + StartTransactionCommand(); + + /* Get fresh snapshot for next step */ + PushActiveSnapshot(GetTransactionSnapshot()); + + /* + * Open transaction if necessary, for the first index treated its + * transaction has been already opened previously. + */ + index_concurrent_drop(indexOid); + + /* + * For the last index to be treated, do not commit transaction yet. + * This will be done once all the locks on indexes and parent relations + * are released. + */ + if (indexOid != llast_oid(concurrentIndexIds)) + { + /* We can do away with our snapshot */ + PopActiveSnapshot(); + + /* Commit this transaction to make the update visible. */ + CommitTransactionCommand(); + } + } + + /* + * Last thing to do is release the session-level lock on the parent table + * and the indexes of table. + */ + foreach(lc, relationLocks) + { + LockRelId lockRel = * (LockRelId *) lfirst(lc); + UnlockRelationIdForSession(&lockRel, ShareUpdateExclusiveLock); + } + + return true; +} + + +/* * CheckMutability * Test whether given expression is mutable */ @@ -1535,7 +1952,8 @@ ChooseRelationName(const char *name1, const char *name2, static char * ChooseIndexName(const char *tabname, Oid namespaceId, List *colnames, List *exclusionOpNames, - bool primary, bool isconstraint) + bool primary, bool isconstraint, + bool concurrent) { char *indexname; @@ -1561,6 +1979,13 @@ ChooseIndexName(const char *tabname, Oid namespaceId, "key", namespaceId); } + else if (concurrent) + { + indexname = ChooseRelationName(tabname, + NULL, + "cct", + namespaceId); + } else { indexname = ChooseRelationName(tabname, @@ -1673,18 +2098,22 @@ ChooseIndexColumnNames(List *indexElems) * Recreate a specific index. */ Oid -ReindexIndex(RangeVar *indexRelation) +ReindexIndex(RangeVar *indexRelation, bool concurrent) { Oid indOid; Oid heapOid = InvalidOid; - /* lock level used here should match index lock reindex_index() */ - indOid = RangeVarGetRelidExtended(indexRelation, AccessExclusiveLock, - false, false, - RangeVarCallbackForReindexIndex, - (void *) &heapOid); + indOid = RangeVarGetRelidExtended(indexRelation, + concurrent ? ShareUpdateExclusiveLock : AccessExclusiveLock, + false, false, + RangeVarCallbackForReindexIndex, + (void *) &heapOid); - reindex_index(indOid, false); + /* Continue process for concurrent or non-concurrent case */ + if (!concurrent) + reindex_index(indOid, false); + else + ReindexRelationConcurrently(indOid); return indOid; } @@ -1748,18 +2177,33 @@ RangeVarCallbackForReindexIndex(const RangeVar *relation, } } + /* * ReindexTable * Recreate all indexes of a table (and of its toast table, if any) */ Oid -ReindexTable(RangeVar *relation) +ReindexTable(RangeVar *relation, bool concurrent) { Oid heapOid; /* The lock level used here should match reindex_relation(). */ - heapOid = RangeVarGetRelidExtended(relation, ShareLock, false, false, - RangeVarCallbackOwnsTable, NULL); + heapOid = RangeVarGetRelidExtended(relation, + concurrent ? ShareUpdateExclusiveLock : ShareLock, + false, false, + RangeVarCallbackOwnsTable, NULL); + + /* Run through the concurrent process if necessary */ + if (concurrent) + { + if (!ReindexRelationConcurrently(heapOid)) + { + ereport(NOTICE, + (errmsg("table \"%s\" has no indexes", + relation->relname))); + } + return heapOid; + } if (!reindex_relation(heapOid, REINDEX_REL_PROCESS_TOAST)) ereport(NOTICE, @@ -1778,7 +2222,10 @@ ReindexTable(RangeVar *relation) * That means this must not be called within a user transaction block! */ Oid -ReindexDatabase(const char *databaseName, bool do_system, bool do_user) +ReindexDatabase(const char *databaseName, + bool do_system, + bool do_user, + bool concurrent) { Relation relationRelation; HeapScanDesc scan; @@ -1790,6 +2237,15 @@ ReindexDatabase(const char *databaseName, bool do_system, bool do_user) AssertArg(databaseName); + /* + * CONCURRENTLY operation is not allowed for a system, but it is for a + * database. + */ + if (concurrent && !do_user) + ereport(ERROR, + (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), + errmsg("cannot reindex system concurrently"))); + if (strcmp(databaseName, get_database_name(MyDatabaseId)) != 0) ereport(ERROR, (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), @@ -1873,15 +2329,40 @@ ReindexDatabase(const char *databaseName, bool do_system, bool do_user) foreach(l, relids) { Oid relid = lfirst_oid(l); + bool result = false; + bool process_concurrent; StartTransactionCommand(); /* functions in indexes may want a snapshot set */ PushActiveSnapshot(GetTransactionSnapshot()); - if (reindex_relation(relid, REINDEX_REL_PROCESS_TOAST)) + + /* Determine if relation needs to be processed concurrently */ + process_concurrent = concurrent && + !IsSystemNamespace(get_rel_namespace(relid)); + + /* + * Reindex relation with a concurrent or non-concurrent process. + * System relations cannot be reindexed concurrently, but they + * need to be reindexed including pg_class with a normal process + * as they could be corrupted, and concurrent process might also + * use them. This does not include toast relations, which are + * reindexed when their parent relation is processed. + */ + if (process_concurrent) + { + old = MemoryContextSwitchTo(private_context); + result = ReindexRelationConcurrently(relid); + MemoryContextSwitchTo(old); + } + else + result = reindex_relation(relid, REINDEX_REL_PROCESS_TOAST); + + if (result) ereport(NOTICE, - (errmsg("table \"%s.%s\" was reindexed", + (errmsg("table \"%s.%s\" was reindexed%s", get_namespace_name(get_rel_namespace(relid)), - get_rel_name(relid)))); + get_rel_name(relid), + process_concurrent ? " concurrently" : ""))); PopActiveSnapshot(); CommitTransactionCommand(); } diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index 0d6f5c0..0bd67a2 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -904,6 +904,36 @@ RangeVarCallbackForDropRelation(const RangeVar *rel, Oid relOid, Oid oldRelOid, if (classform->relkind != relkind) DropErrorMsgWrongType(rel->relname, classform->relkind, relkind); + /* + * Check the case of a system index that might have been invalidated by a + * failed concurrent process and allow its drop. + */ + if (IsSystemClass(classform) && + relkind == RELKIND_INDEX) + { + HeapTuple locTuple; + Form_pg_index indexform; + bool indisvalid; + + locTuple = SearchSysCache1(INDEXRELID, ObjectIdGetDatum(state->heapOid)); + if (!HeapTupleIsValid(locTuple)) + { + ReleaseSysCache(tuple); + return; + } + + indexform = (Form_pg_index) GETSTRUCT(locTuple); + indisvalid = indexform->indisvalid; + ReleaseSysCache(locTuple); + + /* Leave if index entry is not valid */ + if (!indisvalid) + { + ReleaseSysCache(tuple); + return; + } + } + /* Allow DROP to either table owner or schema owner */ if (!pg_class_ownercheck(relOid, GetUserId()) && !pg_namespace_ownercheck(classform->relnamespace, GetUserId())) diff --git a/src/backend/executor/execUtils.c b/src/backend/executor/execUtils.c index 11be62e..c46bdcc 100644 --- a/src/backend/executor/execUtils.c +++ b/src/backend/executor/execUtils.c @@ -1185,6 +1185,20 @@ check_exclusion_constraint(Relation heap, Relation index, IndexInfo *indexInfo, } /* + * As an invalid index only exists when created in a concurrent context, + * and that this code path cannot be taken by CREATE INDEX CONCURRENTLY + * as this feature is not available for exclusion constraints, this code + * path can only be taken by REINDEX CONCURRENTLY. In this case the same + * index exists in parallel to this one so we can bypass this check as + * it has already been done on the other index existing in parallel. + * If exclusion constraints are supported in the future for CREATE INDEX + * CONCURRENTLY, this should be removed or completed especially for this + * purpose. + */ + if (!index->rd_index->indisvalid) + return true; + + /* * Search the tuples that are in the index for any violations, including * tuples that aren't visible yet. */ diff --git a/src/backend/nodes/copyfuncs.c b/src/backend/nodes/copyfuncs.c index 867b0c0..b93d90c 100644 --- a/src/backend/nodes/copyfuncs.c +++ b/src/backend/nodes/copyfuncs.c @@ -3617,6 +3617,7 @@ _copyReindexStmt(const ReindexStmt *from) COPY_STRING_FIELD(name); COPY_SCALAR_FIELD(do_system); COPY_SCALAR_FIELD(do_user); + COPY_SCALAR_FIELD(concurrent); return newnode; } diff --git a/src/backend/nodes/equalfuncs.c b/src/backend/nodes/equalfuncs.c index 085cd5b..2687bf0 100644 --- a/src/backend/nodes/equalfuncs.c +++ b/src/backend/nodes/equalfuncs.c @@ -1853,6 +1853,7 @@ _equalReindexStmt(const ReindexStmt *a, const ReindexStmt *b) COMPARE_STRING_FIELD(name); COMPARE_SCALAR_FIELD(do_system); COMPARE_SCALAR_FIELD(do_user); + COMPARE_SCALAR_FIELD(concurrent); return true; } diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y index 0787d2f..f087219 100644 --- a/src/backend/parser/gram.y +++ b/src/backend/parser/gram.y @@ -6806,29 +6806,32 @@ opt_if_exists: IF_P EXISTS { $$ = TRUE; } *****************************************************************************/ ReindexStmt: - REINDEX reindex_type qualified_name opt_force + REINDEX reindex_type opt_concurrently qualified_name opt_force { ReindexStmt *n = makeNode(ReindexStmt); n->kind = $2; - n->relation = $3; + n->concurrent = $3; + n->relation = $4; n->name = NULL; $$ = (Node *)n; } - | REINDEX SYSTEM_P name opt_force + | REINDEX SYSTEM_P opt_concurrently name opt_force { ReindexStmt *n = makeNode(ReindexStmt); n->kind = OBJECT_DATABASE; - n->name = $3; + n->concurrent = $3; + n->name = $4; n->relation = NULL; n->do_system = true; n->do_user = false; $$ = (Node *)n; } - | REINDEX DATABASE name opt_force + | REINDEX DATABASE opt_concurrently name opt_force { ReindexStmt *n = makeNode(ReindexStmt); n->kind = OBJECT_DATABASE; - n->name = $3; + n->concurrent = $3; + n->name = $4; n->relation = NULL; n->do_system = true; n->do_user = true; diff --git a/src/backend/storage/ipc/procarray.c b/src/backend/storage/ipc/procarray.c index 4308128..1662a6e 100644 --- a/src/backend/storage/ipc/procarray.c +++ b/src/backend/storage/ipc/procarray.c @@ -2528,6 +2528,152 @@ XidCacheRemoveRunningXids(TransactionId xid, LWLockRelease(ProcArrayLock); } + +/* + * WaitForMultipleVirtualLocks + * + * Wait until no transactions hold the relation related to lock those locks. + * To do this, inquire which xacts currently would conflict with each lock on + * the table referred by the respective LOCKTAG -- ie, which ones have a lock + * that permits writing the relation. Then wait for each of these xacts to + * commit or abort. + * + * To do this, inquire which xacts currently would conflict with lockmode + * on the relation. + * + * Note: GetLockConflicts() never reports our own xid, hence we need not + * check for that. Also, prepared xacts are not reported, which is fine + * since they certainly aren't going to do anything more. + */ +void +WaitForMultipleVirtualLocks(List *locktags, LOCKMODE lockmode) +{ + VirtualTransactionId **old_lockholders; + int i, count = 0; + ListCell *lc; + + /* Leave if no locks to wait for */ + if (list_length(locktags) == 0) + return; + + old_lockholders = (VirtualTransactionId **) + palloc(list_length(locktags) * sizeof(VirtualTransactionId *)); + + /* Collect the transactions we need to wait on for each relation lock */ + foreach(lc, locktags) + { + LOCKTAG *locktag = lfirst(lc); + old_lockholders[count++] = GetLockConflicts(locktag, lockmode); + } + + /* Finally wait for each transaction to complete */ + for (i = 0; i < count; i++) + { + VirtualTransactionId *lockholders = old_lockholders[i]; + + while (VirtualTransactionIdIsValid(*lockholders)) + { + VirtualXactLock(*lockholders, true); + lockholders++; + } + } + + pfree(old_lockholders); +} + + +/* + * WaitForVirtualLocks + * + * Similar to WaitForMultipleVirtualLocks, but for a single lock. + */ +void +WaitForVirtualLocks(LOCKTAG heaplocktag, LOCKMODE lockmode) +{ + WaitForMultipleVirtualLocks(list_make1(&heaplocktag), lockmode); +} + + +/* + * WaitForOldSnapshots + * + * Wait for transactions that might have older snapshot than the given one, + * because is might not contain tuples deleted just before it has been taken. + * Obtain a list of VXIDs of such transactions, and wait for them + * individually. + * + * We can exclude any running transactions that have xmin > the xmin of + * our reference snapshot; their oldest snapshot must be newer than ours. + * We can also exclude any transactions that have xmin = zero, since they + * evidently have no live snapshot at all (and any one they might be in + * process of taking is certainly newer than ours). Transactions in other + * DBs can be ignored too, since they'll never even be able to see this + * index. + * + * We can also exclude autovacuum processes and processes running manual + * lazy VACUUMs, because they won't be fazed by missing index entries + * either. (Manual ANALYZEs, however, can't be excluded because they + * might be within transactions that are going to do arbitrary operations + * later.) + * + * Also, GetCurrentVirtualXIDs never reports our own vxid, so we need not + * check for that. + * + * If a process goes idle-in-transaction with xmin zero, we do not need to + * wait for it anymore, per the above argument. We do not have the + * infrastructure right now to stop waiting if that happens, but we can at + * least avoid the folly of waiting when it is idle at the time we would + * begin to wait. We do this by repeatedly rechecking the output of + * GetCurrentVirtualXIDs. If, during any iteration, a particular vxid + * doesn't show up in the output, we know we can forget about it. + */ +void +WaitForOldSnapshots(Snapshot snapshot) +{ + int i, n_old_snapshots; + VirtualTransactionId *old_snapshots; + + old_snapshots = GetCurrentVirtualXIDs(snapshot->xmin, true, false, + PROC_IS_AUTOVACUUM | PROC_IN_VACUUM, + &n_old_snapshots); + + for (i = 0; i < n_old_snapshots; i++) + { + if (!VirtualTransactionIdIsValid(old_snapshots[i])) + continue; /* found uninteresting in previous cycle */ + + if (i > 0) + { + /* see if anything's changed ... */ + VirtualTransactionId *newer_snapshots; + int n_newer_snapshots, j, k; + + newer_snapshots = GetCurrentVirtualXIDs(snapshot->xmin, + true, false, + PROC_IS_AUTOVACUUM | PROC_IN_VACUUM, + &n_newer_snapshots); + for (j = i; j < n_old_snapshots; j++) + { + if (!VirtualTransactionIdIsValid(old_snapshots[j])) + continue; /* found uninteresting in previous cycle */ + for (k = 0; k < n_newer_snapshots; k++) + { + if (VirtualTransactionIdEquals(old_snapshots[j], + newer_snapshots[k])) + break; + } + if (k >= n_newer_snapshots) /* not there anymore */ + SetInvalidVirtualTransactionId(old_snapshots[j]); + } + pfree(newer_snapshots); + } + + if (VirtualTransactionIdIsValid(old_snapshots[i])) + VirtualXactLock(old_snapshots[i], true); + } +} + + #ifdef XIDCACHE_DEBUG /* diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c index a1c03f1..6a0341b 100644 --- a/src/backend/tcop/utility.c +++ b/src/backend/tcop/utility.c @@ -1292,16 +1292,20 @@ standard_ProcessUtility(Node *parsetree, { ReindexStmt *stmt = (ReindexStmt *) parsetree; + if (stmt->concurrent) + PreventTransactionChain(isTopLevel, + "REINDEX CONCURRENTLY"); + /* we choose to allow this during "read only" transactions */ PreventCommandDuringRecovery("REINDEX"); switch (stmt->kind) { case OBJECT_INDEX: - ReindexIndex(stmt->relation); + ReindexIndex(stmt->relation, stmt->concurrent); break; case OBJECT_TABLE: case OBJECT_MATVIEW: - ReindexTable(stmt->relation); + ReindexTable(stmt->relation, stmt->concurrent); break; case OBJECT_DATABASE: @@ -1313,8 +1317,8 @@ standard_ProcessUtility(Node *parsetree, */ PreventTransactionChain(isTopLevel, "REINDEX DATABASE"); - ReindexDatabase(stmt->name, - stmt->do_system, stmt->do_user); + ReindexDatabase(stmt->name, stmt->do_system, + stmt->do_user, stmt->concurrent); break; default: elog(ERROR, "unrecognized object type: %d", diff --git a/src/include/catalog/index.h b/src/include/catalog/index.h index fb323f7..db2a531 100644 --- a/src/include/catalog/index.h +++ b/src/include/catalog/index.h @@ -60,7 +60,26 @@ extern Oid index_create(Relation heapRelation, bool allow_system_table_mods, bool skip_build, bool concurrent, - bool is_internal); + bool is_internal, + bool is_reindex); + +extern Oid index_concurrent_create(Relation heapRelation, + Oid indOid, + char *concurrentName); + +extern void index_concurrent_build(Oid heapOid, + Oid indexOid, + bool isprimary); + +extern void index_concurrent_swap(Oid newIndexOid, Oid oldIndexOid); + +extern void index_concurrent_set_dead(Oid indexId, + Oid heapId, + LOCKTAG *locktag); + +extern void index_concurrent_clear_valid(Relation heapRelation, Oid indexOid); + +extern void index_concurrent_drop(Oid indexOid); extern void index_constraint_create(Relation heapRelation, Oid indexRelationId, diff --git a/src/include/commands/defrem.h b/src/include/commands/defrem.h index 62515b2..54137c6 100644 --- a/src/include/commands/defrem.h +++ b/src/include/commands/defrem.h @@ -26,10 +26,11 @@ extern Oid DefineIndex(IndexStmt *stmt, bool check_rights, bool skip_build, bool quiet); -extern Oid ReindexIndex(RangeVar *indexRelation); -extern Oid ReindexTable(RangeVar *relation); +extern Oid ReindexIndex(RangeVar *indexRelation, bool concurrent); +extern Oid ReindexTable(RangeVar *relation, bool concurrent); extern Oid ReindexDatabase(const char *databaseName, - bool do_system, bool do_user); + bool do_system, bool do_user, bool concurrent); +extern bool ReindexRelationConcurrently(Oid relOid); extern char *makeObjectName(const char *name1, const char *name2, const char *label); extern char *ChooseRelationName(const char *name1, const char *name2, diff --git a/src/include/nodes/parsenodes.h b/src/include/nodes/parsenodes.h index 2229ef0..bb3ae47 100644 --- a/src/include/nodes/parsenodes.h +++ b/src/include/nodes/parsenodes.h @@ -2538,6 +2538,7 @@ typedef struct ReindexStmt const char *name; /* name of database to reindex */ bool do_system; /* include system tables in database case */ bool do_user; /* include user tables in database case */ + bool concurrent; /* reindex concurrently? */ } ReindexStmt; /* ---------------------- diff --git a/src/include/storage/procarray.h b/src/include/storage/procarray.h index d5fdfea..d4a0981 100644 --- a/src/include/storage/procarray.h +++ b/src/include/storage/procarray.h @@ -76,4 +76,8 @@ extern void XidCacheRemoveRunningXids(TransactionId xid, int nxids, const TransactionId *xids, TransactionId latestXid); +extern void WaitForMultipleVirtualLocks(List *locktags, LOCKMODE lockmode); +extern void WaitForVirtualLocks(LOCKTAG heaplocktag, LOCKMODE lockmode); +extern void WaitForOldSnapshots(Snapshot snapshot); + #endif /* PROCARRAY_H */ diff --git a/src/test/regress/expected/create_index.out b/src/test/regress/expected/create_index.out index 2ae991e..88ec81a 100644 --- a/src/test/regress/expected/create_index.out +++ b/src/test/regress/expected/create_index.out @@ -2721,3 +2721,54 @@ ORDER BY thousand; 1 | 1001 (2 rows) +-- +-- Check behavior of REINDEX and REINDEX CONCURRENTLY +-- +CREATE TABLE concur_reindex_tab (c1 int); +-- REINDEX +REINDEX TABLE concur_reindex_tab; -- notice +NOTICE: table "concur_reindex_tab" has no indexes +REINDEX TABLE CONCURRENTLY concur_reindex_tab; -- notice +NOTICE: table "concur_reindex_tab" has no indexes +ALTER TABLE concur_reindex_tab ADD COLUMN c2 text; -- add toast index +-- Normal index with integer column +CREATE UNIQUE INDEX concur_reindex_ind1 ON concur_reindex_tab(c1); +-- Normal index with text column +CREATE INDEX concur_reindex_ind2 ON concur_reindex_tab(c2); +-- UNIQUE index with expression +CREATE UNIQUE INDEX concur_reindex_ind3 ON concur_reindex_tab(abs(c1)); +-- Duplicate column names +CREATE INDEX concur_reindex_ind4 ON concur_reindex_tab(c1, c1, c2); +-- Create table for check on foreign key dependence switch with indexes swapped +ALTER TABLE concur_reindex_tab ADD PRIMARY KEY USING INDEX concur_reindex_ind1; +CREATE TABLE concur_reindex_tab2 (c1 int REFERENCES concur_reindex_tab); +INSERT INTO concur_reindex_tab VALUES (1, 'a'); +INSERT INTO concur_reindex_tab VALUES (2, 'a'); +REINDEX INDEX CONCURRENTLY concur_reindex_ind1; +REINDEX TABLE CONCURRENTLY concur_reindex_tab; +-- Check errors +-- Cannot run inside a transaction block +BEGIN; +REINDEX TABLE CONCURRENTLY concur_reindex_tab; +ERROR: REINDEX CONCURRENTLY cannot run inside a transaction block +COMMIT; +REINDEX TABLE CONCURRENTLY pg_database; -- no shared relation +ERROR: concurrent reindex is not supported for shared relations +REINDEX SYSTEM CONCURRENTLY postgres; -- not allowed for SYSTEM +ERROR: cannot reindex system concurrently +-- Check the relation status, there should not be invalid indexes +\d concur_reindex_tab +Table "public.concur_reindex_tab" + Column | Type | Modifiers +--------+---------+----------- + c1 | integer | not null + c2 | text | +Indexes: + "concur_reindex_ind1" PRIMARY KEY, btree (c1) + "concur_reindex_ind3" UNIQUE, btree (abs(c1)) + "concur_reindex_ind2" btree (c2) + "concur_reindex_ind4" btree (c1, c1, c2) +Referenced by: + TABLE "concur_reindex_tab2" CONSTRAINT "concur_reindex_tab2_c1_fkey" FOREIGN KEY (c1) REFERENCES concur_reindex_tab(c1) + +DROP TABLE concur_reindex_tab, concur_reindex_tab2; diff --git a/src/test/regress/sql/create_index.sql b/src/test/regress/sql/create_index.sql index 914e7a5..a0b2ae2 100644 --- a/src/test/regress/sql/create_index.sql +++ b/src/test/regress/sql/create_index.sql @@ -912,3 +912,39 @@ ORDER BY thousand; SELECT thousand, tenthous FROM tenk1 WHERE thousand < 2 AND tenthous IN (1001,3000) ORDER BY thousand; + +-- +-- Check behavior of REINDEX and REINDEX CONCURRENTLY +-- +CREATE TABLE concur_reindex_tab (c1 int); +-- REINDEX +REINDEX TABLE concur_reindex_tab; -- notice +REINDEX TABLE CONCURRENTLY concur_reindex_tab; -- notice +ALTER TABLE concur_reindex_tab ADD COLUMN c2 text; -- add toast index +-- Normal index with integer column +CREATE UNIQUE INDEX concur_reindex_ind1 ON concur_reindex_tab(c1); +-- Normal index with text column +CREATE INDEX concur_reindex_ind2 ON concur_reindex_tab(c2); +-- UNIQUE index with expression +CREATE UNIQUE INDEX concur_reindex_ind3 ON concur_reindex_tab(abs(c1)); +-- Duplicate column names +CREATE INDEX concur_reindex_ind4 ON concur_reindex_tab(c1, c1, c2); +-- Create table for check on foreign key dependence switch with indexes swapped +ALTER TABLE concur_reindex_tab ADD PRIMARY KEY USING INDEX concur_reindex_ind1; +CREATE TABLE concur_reindex_tab2 (c1 int REFERENCES concur_reindex_tab); +INSERT INTO concur_reindex_tab VALUES (1, 'a'); +INSERT INTO concur_reindex_tab VALUES (2, 'a'); +REINDEX INDEX CONCURRENTLY concur_reindex_ind1; +REINDEX TABLE CONCURRENTLY concur_reindex_tab; + +-- Check errors +-- Cannot run inside a transaction block +BEGIN; +REINDEX TABLE CONCURRENTLY concur_reindex_tab; +COMMIT; +REINDEX TABLE CONCURRENTLY pg_database; -- no shared relation +REINDEX SYSTEM CONCURRENTLY postgres; -- not allowed for SYSTEM + +-- Check the relation status, there should not be invalid indexes +\d concur_reindex_tab +DROP TABLE concur_reindex_tab, concur_reindex_tab2;