On Mon, Sep 27, 2021 at 12:23 PM Dilip Kumar <dilipbal...@gmail.com> wrote:
>
>
> Open question:
> - Scan pg_class vs scan directories
> - Whether to retain the old created database mechanism as option or not.
I have done some code improvement in 0001 and 0002.
--
Regards,
Dilip Kumar
EnterpriseDB: http://www.enterprisedb.com
From fce1a87e25d20bf4b1a85c6cc535db42b5bdfc73 Mon Sep 17 00:00:00 2001
From: Dilip Kumar <dilipkumar@localhost.localdomain>
Date: Wed, 1 Sep 2021 14:06:29 +0530
Subject: [PATCH v4 1/6] Refactor relmap load and relmap write functions
Currently, write_relmap_file and load_relmap_file are tightly
coupled with shared_map and local_map. As part of the higher
level patch set we need remap read/write interfaces that are
not dependent upon shared_map and local_map, and we should be
able to pass map memory as an external parameter instead.
---
src/backend/utils/cache/relmapper.c | 163 ++++++++++++++++++++++--------------
1 file changed, 99 insertions(+), 64 deletions(-)
diff --git a/src/backend/utils/cache/relmapper.c b/src/backend/utils/cache/relmapper.c
index a6e38ad..bb39632 100644
--- a/src/backend/utils/cache/relmapper.c
+++ b/src/backend/utils/cache/relmapper.c
@@ -136,6 +136,12 @@ static void apply_map_update(RelMapFile *map, Oid relationId, Oid fileNode,
bool add_okay);
static void merge_map_updates(RelMapFile *map, const RelMapFile *updates,
bool add_okay);
+static void read_relmap_file(char *mapfilename, RelMapFile *map,
+ bool lock_held);
+static void write_relmap_file_internal(char *mapfilename, RelMapFile *newmap,
+ bool write_wal, bool send_sinval,
+ bool preserve_files, Oid dbid, Oid tsid,
+ const char *dbpath);
static void load_relmap_file(bool shared, bool lock_held);
static void write_relmap_file(bool shared, RelMapFile *newmap,
bool write_wal, bool send_sinval, bool preserve_files,
@@ -687,36 +693,19 @@ RestoreRelationMap(char *startAddress)
}
/*
- * load_relmap_file -- load data from the shared or local map file
+ * read_relmap_file -- read data from given mapfilename file.
*
* Because the map file is essential for access to core system catalogs,
* failure to read it is a fatal error.
- *
- * Note that the local case requires DatabasePath to be set up.
*/
static void
-load_relmap_file(bool shared, bool lock_held)
+read_relmap_file(char *mapfilename, RelMapFile *map, bool lock_held)
{
- RelMapFile *map;
- char mapfilename[MAXPGPATH];
pg_crc32c crc;
int fd;
int r;
- if (shared)
- {
- snprintf(mapfilename, sizeof(mapfilename), "global/%s",
- RELMAPPER_FILENAME);
- map = &shared_map;
- }
- else
- {
- snprintf(mapfilename, sizeof(mapfilename), "%s/%s",
- DatabasePath, RELMAPPER_FILENAME);
- map = &local_map;
- }
-
- /* Read data ... */
+ /* Open the relmap file for reading. */
fd = OpenTransientFile(mapfilename, O_RDONLY | PG_BINARY);
if (fd < 0)
ereport(FATAL,
@@ -779,62 +768,50 @@ load_relmap_file(bool shared, bool lock_held)
}
/*
- * Write out a new shared or local map file with the given contents.
- *
- * The magic number and CRC are automatically updated in *newmap. On
- * success, we copy the data to the appropriate permanent static variable.
- *
- * If write_wal is true then an appropriate WAL message is emitted.
- * (It will be false for bootstrap and WAL replay cases.)
- *
- * If send_sinval is true then a SI invalidation message is sent.
- * (This should be true except in bootstrap case.)
- *
- * If preserve_files is true then the storage manager is warned not to
- * delete the files listed in the map.
+ * load_relmap_file -- load data from the shared or local map file
*
- * Because this may be called during WAL replay when MyDatabaseId,
- * DatabasePath, etc aren't valid, we require the caller to pass in suitable
- * values. The caller is also responsible for being sure no concurrent
- * map update could be happening.
+ * Note that the local case requires DatabasePath to be set up.
*/
static void
-write_relmap_file(bool shared, RelMapFile *newmap,
- bool write_wal, bool send_sinval, bool preserve_files,
- Oid dbid, Oid tsid, const char *dbpath)
+load_relmap_file(bool shared, bool lock_held)
{
- int fd;
- RelMapFile *realmap;
+ RelMapFile *map;
char mapfilename[MAXPGPATH];
- /*
- * Fill in the overhead fields and update CRC.
- */
- newmap->magic = RELMAPPER_FILEMAGIC;
- if (newmap->num_mappings < 0 || newmap->num_mappings > MAX_MAPPINGS)
- elog(ERROR, "attempt to write bogus relation mapping");
-
- INIT_CRC32C(newmap->crc);
- COMP_CRC32C(newmap->crc, (char *) newmap, offsetof(RelMapFile, crc));
- FIN_CRC32C(newmap->crc);
-
- /*
- * Open the target file. We prefer to do this before entering the
- * critical section, so that an open() failure need not force PANIC.
- */
if (shared)
{
snprintf(mapfilename, sizeof(mapfilename), "global/%s",
RELMAPPER_FILENAME);
- realmap = &shared_map;
+ map = &shared_map;
}
else
{
snprintf(mapfilename, sizeof(mapfilename), "%s/%s",
- dbpath, RELMAPPER_FILENAME);
- realmap = &local_map;
+ DatabasePath, RELMAPPER_FILENAME);
+ map = &local_map;
}
+ /* Read data ... */
+ read_relmap_file(mapfilename, map, lock_held);
+}
+
+/*
+ * Helper function for write_relmap_file, Read comments atop write_relmap_file
+ * for more details. The CRC should be computed by the caller and stored in
+ * the newmap.
+ */
+static void
+write_relmap_file_internal(char *mapfilename, RelMapFile *newmap,
+ bool write_wal, bool send_sinval,
+ bool preserve_files, Oid dbid, Oid tsid,
+ const char *dbpath)
+{
+ int fd;
+
+ /*
+ * Open the target file. We prefer to do this before entering the
+ * critical section, so that an open() failure need not force PANIC.
+ */
fd = OpenTransientFile(mapfilename, O_WRONLY | O_CREAT | PG_BINARY);
if (fd < 0)
ereport(ERROR,
@@ -934,6 +911,68 @@ write_relmap_file(bool shared, RelMapFile *newmap,
}
}
+ /* Critical section done */
+ if (write_wal)
+ END_CRIT_SECTION();
+}
+
+/*
+ * Write out a new shared or local map file with the given contents.
+ *
+ * The magic number and CRC are automatically updated in *newmap. On
+ * success, we copy the data to the appropriate permanent static variable.
+ *
+ * If write_wal is true then an appropriate WAL message is emitted.
+ * (It will be false for bootstrap and WAL replay cases.)
+ *
+ * If send_sinval is true then a SI invalidation message is sent.
+ * (This should be true except in bootstrap case.)
+ *
+ * If preserve_files is true then the storage manager is warned not to
+ * delete the files listed in the map.
+ *
+ * Because this may be called during WAL replay when MyDatabaseId,
+ * DatabasePath, etc aren't valid, we require the caller to pass in suitable
+ * values. The caller is also responsible for being sure no concurrent
+ * map update could be happening.
+ */
+static void
+write_relmap_file(bool shared, RelMapFile *newmap,
+ bool write_wal, bool send_sinval, bool preserve_files,
+ Oid dbid, Oid tsid, const char *dbpath)
+{
+ RelMapFile *realmap;
+ char mapfilename[MAXPGPATH];
+
+ /*
+ * Fill in the overhead fields and update CRC.
+ */
+ newmap->magic = RELMAPPER_FILEMAGIC;
+ if (newmap->num_mappings < 0 || newmap->num_mappings > MAX_MAPPINGS)
+ elog(ERROR, "attempt to write bogus relation mapping");
+
+ INIT_CRC32C(newmap->crc);
+ COMP_CRC32C(newmap->crc, (char *) newmap, offsetof(RelMapFile, crc));
+ FIN_CRC32C(newmap->crc);
+
+ if (shared)
+ {
+ snprintf(mapfilename, sizeof(mapfilename), "global/%s",
+ RELMAPPER_FILENAME);
+ realmap = &shared_map;
+ }
+ else
+ {
+ snprintf(mapfilename, sizeof(mapfilename), "%s/%s",
+ dbpath, RELMAPPER_FILENAME);
+ realmap = &local_map;
+ }
+
+ /* Write the map to the relmap file. */
+ write_relmap_file_internal(mapfilename, newmap, write_wal,
+ send_sinval, preserve_files, dbid, tsid,
+ dbpath);
+
/*
* Success, update permanent copy. During bootstrap, we might be working
* on the permanent copy itself, in which case skip the memcpy() to avoid
@@ -943,10 +982,6 @@ write_relmap_file(bool shared, RelMapFile *newmap,
memcpy(realmap, newmap, sizeof(RelMapFile));
else
Assert(!send_sinval); /* must be bootstrapping */
-
- /* Critical section done */
- if (write_wal)
- END_CRIT_SECTION();
}
/*
--
1.8.3.1
From 27743d5cf737dd10e54c4231e3d2c5cace3b87ea Mon Sep 17 00:00:00 2001
From: Dilip Kumar <dilipkumar@localhost.localdomain>
Date: Fri, 24 Sep 2021 18:13:25 +0530
Subject: [PATCH v4 3/6] Refactor index_copy_data
Make separate interface for copying relation storage, this will
be used by later patch for copying the database relations.
---
src/backend/commands/tablecmds.c | 59 ++++++++++++++++++++++++----------------
src/include/commands/tablecmds.h | 5 ++++
2 files changed, 40 insertions(+), 24 deletions(-)
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index ff97b61..426d1b0 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -14189,21 +14189,13 @@ AlterTableMoveAll(AlterTableMoveAllStmt *stmt)
return new_tablespaceoid;
}
-static void
-index_copy_data(Relation rel, RelFileNode newrnode)
+/*
+ * Copy source smgr all fork's data to the destination smgr.
+ */
+void
+RelationCopyAllFork(SMgrRelation src_smgr, SMgrRelation dst_smgr,
+ char relpersistence, copy_relation_storage copy_storage)
{
- SMgrRelation dstrel;
-
- dstrel = smgropen(newrnode, rel->rd_backend);
-
- /*
- * Since we copy the file directly without looking at the shared buffers,
- * we'd better first flush out any pages of the source relation that are
- * in shared buffers. We assume no new changes will be made while we are
- * holding exclusive lock on the rel.
- */
- FlushRelationBuffers(rel);
-
/*
* Create and copy all forks of the relation, and schedule unlinking of
* old physical files.
@@ -14211,32 +14203,51 @@ index_copy_data(Relation rel, RelFileNode newrnode)
* NOTE: any conflict in relfilenode value will be caught in
* RelationCreateStorage().
*/
- RelationCreateStorage(newrnode, rel->rd_rel->relpersistence);
+ RelationCreateStorage(dst_smgr->smgr_rnode.node, relpersistence);
/* copy main fork */
- RelationCopyStorage(RelationGetSmgr(rel), dstrel, MAIN_FORKNUM,
- rel->rd_rel->relpersistence);
+ copy_storage(src_smgr, dst_smgr, MAIN_FORKNUM, relpersistence);
/* copy those extra forks that exist */
for (ForkNumber forkNum = MAIN_FORKNUM + 1;
forkNum <= MAX_FORKNUM; forkNum++)
{
- if (smgrexists(RelationGetSmgr(rel), forkNum))
+ if (smgrexists(src_smgr, forkNum))
{
- smgrcreate(dstrel, forkNum, false);
+ smgrcreate(dst_smgr, forkNum, false);
/*
* WAL log creation if the relation is persistent, or this is the
* init fork of an unlogged relation.
*/
- if (RelationIsPermanent(rel) ||
- (rel->rd_rel->relpersistence == RELPERSISTENCE_UNLOGGED &&
+ if (relpersistence == RELPERSISTENCE_PERMANENT ||
+ (relpersistence == RELPERSISTENCE_UNLOGGED &&
forkNum == INIT_FORKNUM))
- log_smgrcreate(&newrnode, forkNum);
- RelationCopyStorage(RelationGetSmgr(rel), dstrel, forkNum,
- rel->rd_rel->relpersistence);
+ log_smgrcreate(&dst_smgr->smgr_rnode.node, forkNum);
+
+ /* Copy a fork's data, block by block. */
+ copy_storage(src_smgr, dst_smgr, forkNum, relpersistence);
}
}
+}
+
+static void
+index_copy_data(Relation rel, RelFileNode newrnode)
+{
+ SMgrRelation dstrel;
+
+ dstrel = smgropen(newrnode, rel->rd_backend);
+
+ /*
+ * Since we copy the file directly without looking at the shared buffers,
+ * we'd better first flush out any pages of the source relation that are
+ * in shared buffers. We assume no new changes will be made while we are
+ * holding exclusive lock on the rel.
+ */
+ FlushRelationBuffers(rel);
+
+ RelationCopyAllFork(RelationGetSmgr(rel), dstrel,
+ rel->rd_rel->relpersistence, RelationCopyStorage);
/* drop old relation, and close new one */
RelationDropStorage(rel);
diff --git a/src/include/commands/tablecmds.h b/src/include/commands/tablecmds.h
index 336549c..e0e0aa5 100644
--- a/src/include/commands/tablecmds.h
+++ b/src/include/commands/tablecmds.h
@@ -19,10 +19,13 @@
#include "catalog/objectaddress.h"
#include "nodes/parsenodes.h"
#include "storage/lock.h"
+#include "storage/smgr.h"
#include "utils/relcache.h"
struct AlterTableUtilityContext; /* avoid including tcop/utility.h here */
+typedef void (*copy_relation_storage) (SMgrRelation src, SMgrRelation dst,
+ ForkNumber forkNum, char relpersistence);
extern ObjectAddress DefineRelation(CreateStmt *stmt, char relkind, Oid ownerId,
ObjectAddress *typaddress, const char *queryString);
@@ -42,6 +45,8 @@ extern void AlterTableInternal(Oid relid, List *cmds, bool recurse);
extern Oid AlterTableMoveAll(AlterTableMoveAllStmt *stmt);
+extern void RelationCopyAllFork(SMgrRelation src_smgr, SMgrRelation dst_smgr,
+ char relpersistence, copy_relation_storage copy_storage);
extern ObjectAddress AlterTableNamespace(AlterObjectSchemaStmt *stmt,
Oid *oldschema);
--
1.8.3.1
From 5f174c31b9218c0f30d30ef9102ca3268fa40094 Mon Sep 17 00:00:00 2001
From: Dilip Kumar <dilipkumar@localhost.localdomain>
Date: Fri, 24 Sep 2021 18:23:39 +0530
Subject: [PATCH v4 4/6] Extend bufmgr interfaces
Extend ReadBufferWithoutRelcache interface to take relpersistence as
and input and extend DropDatabaseBuffers to take tablespace oid as
input.
---
src/backend/storage/buffer/bufmgr.c | 24 +++++++++++-------------
src/include/storage/bufmgr.h | 5 +++--
2 files changed, 14 insertions(+), 15 deletions(-)
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index e88e4e9..ed54c34 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -770,24 +770,17 @@ ReadBufferExtended(Relation reln, ForkNumber forkNum, BlockNumber blockNum,
/*
* ReadBufferWithoutRelcache -- like ReadBufferExtended, but doesn't require
* a relcache entry for the relation.
- *
- * NB: At present, this function may only be used on permanent relations, which
- * is OK, because we only use it during XLOG replay. If in the future we
- * want to use it on temporary or unlogged relations, we could pass additional
- * parameters.
*/
Buffer
ReadBufferWithoutRelcache(RelFileNode rnode, ForkNumber forkNum,
BlockNumber blockNum, ReadBufferMode mode,
- BufferAccessStrategy strategy)
+ BufferAccessStrategy strategy, char relpersistence)
{
bool hit;
SMgrRelation smgr = smgropen(rnode, InvalidBackendId);
- Assert(InRecovery);
-
- return ReadBuffer_common(smgr, RELPERSISTENCE_PERMANENT, forkNum, blockNum,
+ return ReadBuffer_common(smgr, relpersistence, forkNum, blockNum,
mode, strategy, &hit);
}
@@ -797,7 +790,7 @@ ReadBufferWithoutRelcache(RelFileNode rnode, ForkNumber forkNum,
*
* *hit is set to true if the request was satisfied from shared buffer cache.
*/
-static Buffer
+Buffer
ReadBuffer_common(SMgrRelation smgr, char relpersistence, ForkNumber forkNum,
BlockNumber blockNum, ReadBufferMode mode,
BufferAccessStrategy strategy, bool *hit)
@@ -3402,10 +3395,13 @@ FindAndDropRelFileNodeBuffers(RelFileNode rnode, ForkNumber forkNum,
* database, to avoid trying to flush data to disk when the directory
* tree no longer exists. Implementation is pretty similar to
* DropRelFileNodeBuffers() which is for destroying just one relation.
+ *
+ * If a valid tablespace oid is passed then it will compare the tablespace
+ * oid as well otherwise just the db oid.
* --------------------------------------------------------------------
*/
void
-DropDatabaseBuffers(Oid dbid)
+DropDatabaseBuffers(Oid dbid, Oid tbsid)
{
int i;
@@ -3423,11 +3419,13 @@ DropDatabaseBuffers(Oid dbid)
* As in DropRelFileNodeBuffers, an unlocked precheck should be safe
* and saves some cycles.
*/
- if (bufHdr->tag.rnode.dbNode != dbid)
+ if (bufHdr->tag.rnode.dbNode != dbid ||
+ (OidIsValid(tbsid) && bufHdr->tag.rnode.spcNode != tbsid))
continue;
buf_state = LockBufHdr(bufHdr);
- if (bufHdr->tag.rnode.dbNode == dbid)
+ if (bufHdr->tag.rnode.dbNode == dbid &&
+ (!OidIsValid(tbsid) || bufHdr->tag.rnode.spcNode == tbsid))
InvalidateBuffer(bufHdr); /* releases spinlock */
else
UnlockBufHdr(bufHdr, buf_state);
diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h
index cfce23e..237c6a9 100644
--- a/src/include/storage/bufmgr.h
+++ b/src/include/storage/bufmgr.h
@@ -184,7 +184,8 @@ extern Buffer ReadBufferExtended(Relation reln, ForkNumber forkNum,
BufferAccessStrategy strategy);
extern Buffer ReadBufferWithoutRelcache(RelFileNode rnode,
ForkNumber forkNum, BlockNumber blockNum,
- ReadBufferMode mode, BufferAccessStrategy strategy);
+ ReadBufferMode mode, BufferAccessStrategy strategy,
+ char relpersistence);
extern void ReleaseBuffer(Buffer buffer);
extern void UnlockReleaseBuffer(Buffer buffer);
extern void MarkBufferDirty(Buffer buffer);
@@ -207,7 +208,7 @@ extern void FlushDatabaseBuffers(Oid dbid);
extern void DropRelFileNodeBuffers(struct SMgrRelationData *smgr_reln, ForkNumber *forkNum,
int nforks, BlockNumber *firstDelBlock);
extern void DropRelFileNodesAllBuffers(struct SMgrRelationData **smgr_reln, int nnodes);
-extern void DropDatabaseBuffers(Oid dbid);
+extern void DropDatabaseBuffers(Oid dbid, Oid tbsid);
#define RelationGetNumberOfBlocks(reln) \
RelationGetNumberOfBlocksInFork(reln, MAIN_FORKNUM)
--
1.8.3.1
From 28e9dc7185b0553a5604ac49ac0092ffad2306b7 Mon Sep 17 00:00:00 2001
From: Dilip Kumar <dilipkumar@localhost.localdomain>
Date: Fri, 24 Sep 2021 18:29:17 +0530
Subject: [PATCH v4 5/6] New interface to lock relation id
Same as LockRelationOid, but instead of rel oid it will take
LockRelId object as an input. So instead of using MyDatabaseId it
will use the dboid passed in the LockRelId object. So this will
provide an option to lock the relation even if we are not connected
to the database.
---
src/backend/storage/lmgr/lmgr.c | 28 ++++++++++++++++++++++++++++
src/include/storage/lmgr.h | 1 +
2 files changed, 29 insertions(+)
diff --git a/src/backend/storage/lmgr/lmgr.c b/src/backend/storage/lmgr/lmgr.c
index cdf2266..4a321aa 100644
--- a/src/backend/storage/lmgr/lmgr.c
+++ b/src/backend/storage/lmgr/lmgr.c
@@ -176,6 +176,34 @@ ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode)
}
/*
+ * LockRelationId
+ *
+ * Lock, given a LockRelId. Same as LockRelationOid but take LockRelId as an
+ * input.
+ */
+void
+LockRelationId(LockRelId *relid, LOCKMODE lockmode)
+{
+ LOCKTAG tag;
+ LOCALLOCK *locallock;
+ LockAcquireResult res;
+
+ SET_LOCKTAG_RELATION(tag, relid->dbId, relid->relId);
+
+ res = LockAcquireExtended(&tag, lockmode, false, false, true, &locallock);
+
+ /*
+ * Now that we have the lock, check for invalidation messages; see notes
+ * in LockRelationOid.
+ */
+ if (res != LOCKACQUIRE_ALREADY_CLEAR)
+ {
+ AcceptInvalidationMessages();
+ MarkLockClear(locallock);
+ }
+}
+
+/*
* UnlockRelationId
*
* Unlock, given a LockRelId. This is preferred over UnlockRelationOid
diff --git a/src/include/storage/lmgr.h b/src/include/storage/lmgr.h
index b009559..092ee93 100644
--- a/src/include/storage/lmgr.h
+++ b/src/include/storage/lmgr.h
@@ -38,6 +38,7 @@ extern void RelationInitLockInfo(Relation relation);
/* Lock a relation */
extern void LockRelationOid(Oid relid, LOCKMODE lockmode);
+extern void LockRelationId(LockRelId *relid, LOCKMODE lockmode);
extern bool ConditionalLockRelationOid(Oid relid, LOCKMODE lockmode);
extern void UnlockRelationId(LockRelId *relid, LOCKMODE lockmode);
extern void UnlockRelationOid(Oid relid, LOCKMODE lockmode);
--
1.8.3.1
From 8b113b9d59ef9aa75c8ed3862b7cf8058da157ee Mon Sep 17 00:00:00 2001
From: dilipkumar <dilipbal...@gmail.com>
Date: Mon, 4 Oct 2021 13:50:44 +0530
Subject: [PATCH v4 2/6] Extend relmap interfaces
Support new interfaces in relmapper, 1) Support copying the
relmap file from one database path to the other database path.
2) Like RelationMapOidToFilenode, provide another interface
which do the same but instead of getting it for the database
we are connected to it will get it for the input database
path.
These interfaces are required for next patch for supporting the
wal logged created database.
---
src/backend/utils/cache/relmapper.c | 122 +++++++++++++++++++++++++++++++-----
src/include/utils/relmapper.h | 6 +-
2 files changed, 112 insertions(+), 16 deletions(-)
diff --git a/src/backend/utils/cache/relmapper.c b/src/backend/utils/cache/relmapper.c
index bb39632..51f361c 100644
--- a/src/backend/utils/cache/relmapper.c
+++ b/src/backend/utils/cache/relmapper.c
@@ -141,7 +141,7 @@ static void read_relmap_file(char *mapfilename, RelMapFile *map,
static void write_relmap_file_internal(char *mapfilename, RelMapFile *newmap,
bool write_wal, bool send_sinval,
bool preserve_files, Oid dbid, Oid tsid,
- const char *dbpath);
+ const char *dbpath, bool create);
static void load_relmap_file(bool shared, bool lock_held);
static void write_relmap_file(bool shared, RelMapFile *newmap,
bool write_wal, bool send_sinval, bool preserve_files,
@@ -256,6 +256,36 @@ RelationMapFilenodeToOid(Oid filenode, bool shared)
}
/*
+ * RelationMapOidToFilenodeForDatabase
+ *
+ * Find relfilenode for the given relation id in the dbpath. Returns
+ * InvalidOid if the relationId is not found in the relmap.
+ */
+Oid
+RelationMapOidToFilenodeForDatabase(char *dbpath, Oid relationId)
+{
+ RelMapFile map;
+ int i;
+ char mapfilename[MAXPGPATH];
+
+ /* Relmap file path for the given dbpath. */
+ snprintf(mapfilename, sizeof(mapfilename), "%s/%s",
+ dbpath, RELMAPPER_FILENAME);
+
+ /* Read the relmap file from the source database. */
+ read_relmap_file(mapfilename, &map, false);
+
+ /* Iterate over the relmap entries to find the input relation oid. */
+ for (i = 0; i < map.num_mappings; i++)
+ {
+ if (relationId == map.mappings[i].mapoid)
+ return map.mappings[i].mapfilenode;
+ }
+
+ return InvalidOid;
+}
+
+/*
* RelationMapUpdateMap
*
* Install a new relfilenode mapping for the specified relation.
@@ -693,7 +723,43 @@ RestoreRelationMap(char *startAddress)
}
/*
- * read_relmap_file -- read data from given mapfilename file.
+ * CopyRelationMap
+ *
+ * Copy relmapfile from source db path to the destination db path and WAL log
+ * the operation. This function is only called during the create database, so
+ * the destination database is not yet visible to anyone else, thus we don't
+ * need to acquire the relmap lock while updating the destination relmap.
+ */
+void
+CopyRelationMap(Oid dbid, Oid tsid, char *srcdbpath, char *dstdbpath)
+{
+ RelMapFile map;
+ char mapfilename[MAXPGPATH];
+
+ /* Relmap file path of the source database. */
+ snprintf(mapfilename, sizeof(mapfilename), "%s/%s",
+ srcdbpath, RELMAPPER_FILENAME);
+
+ /* Read the relmap file from the source database. */
+ read_relmap_file(mapfilename, &map, false);
+
+ /* Relmap file path of the destination database. */
+ snprintf(mapfilename, sizeof(mapfilename), "%s/%s",
+ dstdbpath, RELMAPPER_FILENAME);
+
+ /*
+ * Write map contents into the destination database's relmap file.
+ * write_relmap_file_internal, expects that the CRC should have been
+ * computed and stored in the input map. But, since we have read this map
+ * from the source database and directly writing to the destination file
+ * without updating it so we don't need to recompute it.
+ */
+ write_relmap_file_internal(mapfilename, &map, true, false, true, dbid,
+ tsid, dstdbpath, true);
+}
+
+/*
+ * read_relmap_file - read data from given mapfilename file.
*
* Because the map file is essential for access to core system catalogs,
* failure to read it is a fatal error.
@@ -796,15 +862,18 @@ load_relmap_file(bool shared, bool lock_held)
}
/*
- * Helper function for write_relmap_file, Read comments atop write_relmap_file
- * for more details. The CRC should be computed by the caller and stored in
- * the newmap.
+ * Helper function for write_relmap_file and CopyRelationMap, Read comments
+ * atop write_relmap_file for more details. The CRC should be computed by the
+ * caller and stored in the newmap.
+ *
+ * Pass the create = true, if we are copying the relmap file during CREATE
+ * DATABASE command.
*/
static void
write_relmap_file_internal(char *mapfilename, RelMapFile *newmap,
bool write_wal, bool send_sinval,
bool preserve_files, Oid dbid, Oid tsid,
- const char *dbpath)
+ const char *dbpath, bool create)
{
int fd;
@@ -830,6 +899,7 @@ write_relmap_file_internal(char *mapfilename, RelMapFile *newmap,
xlrec.dbid = dbid;
xlrec.tsid = tsid;
xlrec.nbytes = sizeof(RelMapFile);
+ xlrec.create = create;
XLogBeginInsert();
XLogRegisterData((char *) (&xlrec), MinSizeOfRelmapUpdate);
@@ -971,7 +1041,7 @@ write_relmap_file(bool shared, RelMapFile *newmap,
/* Write the map to the relmap file. */
write_relmap_file_internal(mapfilename, newmap, write_wal,
send_sinval, preserve_files, dbid, tsid,
- dbpath);
+ dbpath, false);
/*
* Success, update permanent copy. During bootstrap, we might be working
@@ -1063,15 +1133,37 @@ relmap_redo(XLogReaderState *record)
* Write out the new map and send sinval, but of course don't write a
* new WAL entry. There's no surrounding transaction to tell to
* preserve files, either.
- *
- * There shouldn't be anyone else updating relmaps during WAL replay,
- * but grab the lock to interlock against load_relmap_file().
*/
- LWLockAcquire(RelationMappingLock, LW_EXCLUSIVE);
- write_relmap_file((xlrec->dbid == InvalidOid), &newmap,
- false, true, false,
- xlrec->dbid, xlrec->tsid, dbpath);
- LWLockRelease(RelationMappingLock);
+ if (!xlrec->create)
+ {
+ /*
+ * There shouldn't be anyone else updating relmaps during WAL
+ * replay, but grab the lock to interlock against
+ * load_relmap_file().
+ */
+ LWLockAcquire(RelationMappingLock, LW_EXCLUSIVE);
+ write_relmap_file((xlrec->dbid == InvalidOid), &newmap,
+ false, true, false,
+ xlrec->dbid, xlrec->tsid, dbpath);
+ LWLockRelease(RelationMappingLock);
+ }
+ else
+ {
+ char mapfilename[MAXPGPATH];
+
+ /* Construct the mapfilename. */
+ snprintf(mapfilename, sizeof(mapfilename), "%s/%s",
+ dbpath, RELMAPPER_FILENAME);
+
+ /*
+ * We don't need to take relmap lock because this wal is logged
+ * while creating a new database, so there could be no one else
+ * reading/writing the relmap file.
+ */
+ write_relmap_file_internal(mapfilename, &newmap, false, false,
+ false, xlrec->dbid, xlrec->tsid, dbpath,
+ true);
+ }
pfree(dbpath);
}
diff --git a/src/include/utils/relmapper.h b/src/include/utils/relmapper.h
index c0d14da..4165f09 100644
--- a/src/include/utils/relmapper.h
+++ b/src/include/utils/relmapper.h
@@ -29,6 +29,7 @@ typedef struct xl_relmap_update
Oid dbid; /* database ID, or 0 for shared map */
Oid tsid; /* database's tablespace, or pg_global */
int32 nbytes; /* size of relmap data */
+ bool create; /* true if creating new relmap */
char data[FLEXIBLE_ARRAY_MEMBER];
} xl_relmap_update;
@@ -39,6 +40,8 @@ extern Oid RelationMapOidToFilenode(Oid relationId, bool shared);
extern Oid RelationMapFilenodeToOid(Oid relationId, bool shared);
+extern Oid RelationMapOidToFilenodeForDatabase(char *dbpath, Oid relationId);
+
extern void RelationMapUpdateMap(Oid relationId, Oid fileNode, bool shared,
bool immediate);
@@ -62,7 +65,8 @@ extern void RelationMapInitializePhase3(void);
extern Size EstimateRelationMapSpace(void);
extern void SerializeRelationMap(Size maxSize, char *startAddress);
extern void RestoreRelationMap(char *startAddress);
-
+extern void CopyRelationMap(Oid dbid, Oid tsid, char *srcdbpath,
+ char *dstdbpath);
extern void relmap_redo(XLogReaderState *record);
extern void relmap_desc(StringInfo buf, XLogReaderState *record);
extern const char *relmap_identify(uint8 info);
--
1.8.3.1
From b236bf6517b799c1ba7a3c74bfb50505b9d7accd Mon Sep 17 00:00:00 2001
From: Dilip Kumar <dilipkumar@localhost.localdomain>
Date: Fri, 24 Sep 2021 18:33:01 +0530
Subject: [PATCH v4 6/6] WAL logged CREATE DATABASE
Currently, CREATE DATABASE forces a checkpoint, then copies all the files,
then forces another checkpoint. The comments in the createdb() function
explain the reasons for this. The attached patch fixes this problem by making
create database completely WAL logged so that we can avoid the checkpoints.
This can also be useful for supporting the TDE. For example, if we need different
encryption for the source and the target database then we can not re-encrypt the
page data if we copy the whole directory. But with this patch, we are copying
page by page so we have an opportunity to re-encrypt the page before copying that
to the target database.
---
src/backend/access/rmgrdesc/dbasedesc.c | 3 +-
src/backend/access/transam/xlogutils.c | 12 +-
src/backend/commands/dbcommands.c | 704 +++++++++++++++++++++-----------
src/bin/pg_rewind/parsexlog.c | 1 +
src/include/commands/dbcommands_xlog.h | 3 -
5 files changed, 479 insertions(+), 244 deletions(-)
diff --git a/src/backend/access/rmgrdesc/dbasedesc.c b/src/backend/access/rmgrdesc/dbasedesc.c
index 2660984..5010f72 100644
--- a/src/backend/access/rmgrdesc/dbasedesc.c
+++ b/src/backend/access/rmgrdesc/dbasedesc.c
@@ -28,8 +28,7 @@ dbase_desc(StringInfo buf, XLogReaderState *record)
{
xl_dbase_create_rec *xlrec = (xl_dbase_create_rec *) rec;
- appendStringInfo(buf, "copy dir %u/%u to %u/%u",
- xlrec->src_tablespace_id, xlrec->src_db_id,
+ appendStringInfo(buf, "create dir %u/%u",
xlrec->tablespace_id, xlrec->db_id);
}
else if (info == XLOG_DBASE_DROP)
diff --git a/src/backend/access/transam/xlogutils.c b/src/backend/access/transam/xlogutils.c
index 88a1bfd..a7a8b79 100644
--- a/src/backend/access/transam/xlogutils.c
+++ b/src/backend/access/transam/xlogutils.c
@@ -483,8 +483,8 @@ XLogReadBufferExtended(RelFileNode rnode, ForkNumber forknum,
if (blkno < lastblock)
{
/* page exists in file */
- buffer = ReadBufferWithoutRelcache(rnode, forknum, blkno,
- mode, NULL);
+ buffer = ReadBufferWithoutRelcache(rnode, forknum, blkno, mode, NULL,
+ RELPERSISTENCE_PERMANENT);
}
else
{
@@ -508,8 +508,8 @@ XLogReadBufferExtended(RelFileNode rnode, ForkNumber forknum,
LockBuffer(buffer, BUFFER_LOCK_UNLOCK);
ReleaseBuffer(buffer);
}
- buffer = ReadBufferWithoutRelcache(rnode, forknum,
- P_NEW, mode, NULL);
+ buffer = ReadBufferWithoutRelcache(rnode, forknum, P_NEW, mode,
+ NULL, RELPERSISTENCE_PERMANENT);
}
while (BufferGetBlockNumber(buffer) < blkno);
/* Handle the corner case that P_NEW returns non-consecutive pages */
@@ -518,8 +518,8 @@ XLogReadBufferExtended(RelFileNode rnode, ForkNumber forknum,
if (mode == RBM_ZERO_AND_LOCK || mode == RBM_ZERO_AND_CLEANUP_LOCK)
LockBuffer(buffer, BUFFER_LOCK_UNLOCK);
ReleaseBuffer(buffer);
- buffer = ReadBufferWithoutRelcache(rnode, forknum, blkno,
- mode, NULL);
+ buffer = ReadBufferWithoutRelcache(rnode, forknum, blkno, mode,
+ NULL, RELPERSISTENCE_PERMANENT);
}
}
diff --git a/src/backend/commands/dbcommands.c b/src/backend/commands/dbcommands.c
index 029fab4..2b70d4d 100644
--- a/src/backend/commands/dbcommands.c
+++ b/src/backend/commands/dbcommands.c
@@ -45,13 +45,13 @@
#include "commands/dbcommands_xlog.h"
#include "commands/defrem.h"
#include "commands/seclabel.h"
+#include "commands/tablecmds.h"
#include "commands/tablespace.h"
#include "mb/pg_wchar.h"
#include "miscadmin.h"
#include "pgstat.h"
#include "postmaster/bgwriter.h"
#include "replication/slot.h"
-#include "storage/copydir.h"
#include "storage/fd.h"
#include "storage/ipc.h"
#include "storage/lmgr.h"
@@ -62,6 +62,7 @@
#include "utils/builtins.h"
#include "utils/fmgroids.h"
#include "utils/pg_locale.h"
+#include "utils/relmapper.h"
#include "utils/snapmgr.h"
#include "utils/syscache.h"
@@ -77,6 +78,19 @@ typedef struct
Oid dest_tsoid; /* tablespace we are trying to move to */
} movedb_failure_params;
+/*
+ * When creating a database, we scan the pg_class of the source database to
+ * identify all the relations to be copied. The structure is used for storing
+ * information about each relation of the source database.
+ */
+typedef struct CreateDBRelInfo
+{
+ RelFileNode rnode; /* physical relation identifier */
+ Oid reloid; /* relation oid */
+ char relpersistence; /* relation's persistence level */
+} CreateDBRelInfo;
+
+
/* non-export function prototypes */
static void createdb_failure_callback(int code, Datum arg);
static void movedb(const char *dbname, const char *tblspcname);
@@ -91,6 +105,426 @@ static bool have_createdb_privilege(void);
static void remove_dbtablespaces(Oid db_id);
static bool check_db_file_conflict(Oid db_id);
static int errdetail_busy_db(int notherbackends, int npreparedxacts);
+static void CreateDirAndVersionFile(char *dbpath, Oid dbid, Oid tsid,
+ bool isRedo);
+static List *GetDatabaseRelationList(Oid srctbid, Oid srcdbid, char *srcpath);
+void RelationCopyStorageUsingBuffer(SMgrRelation src, SMgrRelation dst,
+ ForkNumber forkNum, char relpersistence);
+static void CopyDatabase(Oid src_dboid, Oid dboid, Oid src_tsid, Oid dst_tsid);
+
+/*
+ * CreateDirAndVersionFile - Create database directory and write out the
+ * PG_VERSION file in the database path.
+ *
+ * If isRedo is true, it's okay for the database directory to exist already.
+ *
+ * We can directly write PG_MAJORVERSION in the version file instead of copying
+ * from the source database file because these two must be the same.
+ */
+static void
+CreateDirAndVersionFile(char *dbpath, Oid dbid, Oid tsid, bool isRedo)
+{
+ int fd;
+ int nbytes = strlen(PG_MAJORVERSION);
+ char versionfile[MAXPGPATH];
+
+ /* If we are not in WAL replay then write the WAL. */
+ if (!isRedo)
+ {
+ xl_dbase_create_rec xlrec;
+ XLogRecPtr lsn;
+
+ /* Now errors are fatal ... */
+ START_CRIT_SECTION();
+
+ xlrec.db_id = dbid;
+ xlrec.tablespace_id = tsid;
+
+ XLogBeginInsert();
+ XLogRegisterData((char *) (&xlrec), sizeof(xl_dbase_create_rec));
+
+ lsn = XLogInsert(RM_DBASE_ID, XLOG_DBASE_CREATE);
+
+ /* As always, WAL must hit the disk before the data update does. */
+ XLogFlush(lsn);
+ }
+
+ /* Create database directory. */
+ if (MakePGDirectory(dbpath) < 0)
+ {
+ /* Failure other than already exists or not in WAL replay? */
+ if (errno != EEXIST || !isRedo)
+ ereport(ERROR,
+ (errcode_for_file_access(),
+ errmsg("could not create directory \"%s\": %m", dbpath)));
+ }
+
+ /*
+ * Create PG_VERSION file in the database path. If the file already exists
+ * and we are in WAL replay then try again to open it in the write mode.
+ */
+ snprintf(versionfile, sizeof(versionfile), "%s/%s", dbpath, "PG_VERSION");
+
+ fd = OpenTransientFile(versionfile, O_RDWR | O_CREAT | O_EXCL | PG_BINARY);
+ if (fd < 0 && errno == EEXIST && isRedo)
+ fd = OpenTransientFile(versionfile, O_RDWR | PG_BINARY);
+
+ if (fd < 0)
+ ereport(ERROR,
+ (errcode_for_file_access(),
+ errmsg("could not create file \"%s\": %m", versionfile)));
+
+ /* Write PG_MAJORVERSION in the PG_VERSION file. */
+ pgstat_report_wait_start(WAIT_EVENT_COPY_FILE_WRITE);
+ errno = 0;
+ if ((int) write(fd, (char *) PG_MAJORVERSION, nbytes) != nbytes)
+ {
+ /* If write didn't set errno, assume problem is no disk space. */
+ if (errno == 0)
+ errno = ENOSPC;
+ ereport(ERROR,
+ (errcode_for_file_access(),
+ errmsg("could not write to file \"%s\": %m", versionfile)));
+ }
+ pgstat_report_wait_end();
+
+ /* Close the version file. */
+ CloseTransientFile(fd);
+
+ /* Critical section done. */
+ if (!isRedo)
+ END_CRIT_SECTION();
+}
+
+/*
+ * GetDatabaseRelationList - Get relfilenode list to be copied.
+ *
+ * Iterate over each block of the pg_class relation. From there, we will check
+ * all the visible tuples in order to get a list of all the valid relfilenodes
+ * in the source database that should be copied to the target database.
+ */
+static List *
+GetDatabaseRelationList(Oid tbid, Oid dbid, char *srcpath)
+{
+ SMgrRelation rd_smgr;
+ RelFileNode rnode;
+ BlockNumber nblocks;
+ BlockNumber blkno;
+ OffsetNumber offnum;
+ OffsetNumber maxoff;
+ Buffer buf;
+ Oid relfilenode;
+ Page page;
+ List *rnodelist = NIL;
+ HeapTupleData tuple;
+ Form_pg_class classForm;
+ LockRelId relid;
+ BufferAccessStrategy bstrategy;
+
+ /* Get pg_class relfilenode. */
+ relfilenode = RelationMapOidToFilenodeForDatabase(srcpath,
+ RelationRelationId);
+ /*
+ * We are going to read the buffers associated with the pg_class relation.
+ * Thus, acquire the relation level lock before start scanning. As we are
+ * not connected to the database, we cannot use relation_open directly, so
+ * we have to lock using relation id.
+ */
+ relid.dbId = dbid;
+ relid.relId = RelationRelationId;
+ LockRelationId(&relid, AccessShareLock);
+
+ /* Prepare a relnode for pg_class relation. */
+ rnode.spcNode = tbid;
+ rnode.dbNode = dbid;
+ rnode.relNode = relfilenode;
+
+ /*
+ * We are not connected to the source database so open the pg_class
+ * relation at the smgr level and get the block count.
+ */
+ rd_smgr = smgropen(rnode, InvalidBackendId);
+ nblocks = smgrnblocks(rd_smgr, MAIN_FORKNUM);
+
+ /*
+ * We're going to read the whole pg_class so better to use bulk-read buffer
+ * access strategy.
+ */
+ bstrategy = GetAccessStrategy(BAS_BULKREAD);
+
+ /* Iterate over each block on the pg_class relation. */
+ for (blkno = 0; blkno < nblocks; blkno++)
+ {
+ /*
+ * We are not connected to the source database so directly use the lower
+ * level bufmgr interface which operates on the rnode.
+ */
+ buf = ReadBufferWithoutRelcache(rnode, MAIN_FORKNUM, blkno,
+ RBM_NORMAL, bstrategy,
+ RELPERSISTENCE_PERMANENT);
+
+ LockBuffer(buf, BUFFER_LOCK_SHARE);
+ page = BufferGetPage(buf);
+ if (PageIsNew(page) || PageIsEmpty(page))
+ continue;
+
+ maxoff = PageGetMaxOffsetNumber(page);
+
+ /* Iterate over each tuple on the page. */
+ for (offnum = FirstOffsetNumber;
+ offnum <= maxoff;
+ offnum = OffsetNumberNext(offnum))
+ {
+ ItemId itemid;
+
+ itemid = PageGetItemId(page, offnum);
+
+ /* Nothing to do if slot is empty or already dead. */
+ if (!ItemIdIsUsed(itemid) || ItemIdIsDead(itemid) ||
+ ItemIdIsRedirected(itemid))
+ continue;
+
+ Assert(ItemIdIsNormal(itemid));
+ ItemPointerSet(&(tuple.t_self), blkno, offnum);
+
+ tuple.t_data = (HeapTupleHeader) PageGetItem(page, itemid);
+ tuple.t_len = ItemIdGetLength(itemid);
+ tuple.t_tableOid = RelationRelationId;
+
+ /*
+ * If the tuple is visible then add its relfilenode info to the
+ * list.
+ */
+ if (HeapTupleSatisfiesVisibility(&tuple, GetActiveSnapshot(), buf))
+ {
+ Oid relfilenode = InvalidOid;
+ CreateDBRelInfo *relinfo;
+
+ classForm = (Form_pg_class) GETSTRUCT(&tuple);
+
+ /* We don't need to copy the shared objects to the target. */
+ if (classForm->reltablespace == GLOBALTABLESPACE_OID)
+ continue;
+
+ /*
+ * If the object doesn't have the storage then nothing to be
+ * done for that object so just ignore it.
+ */
+ if (!RELKIND_HAS_STORAGE(classForm->relkind))
+ continue;
+
+ /*
+ * If relfilenode is valid then directly use it. Otherwise,
+ * consult the relmapper for the mapped relation.
+ *
+ * XXX We can optimize RelationMapOidToFileenodeForDatabase API
+ * so that instead of reading the relmap file every time, it can
+ * save it in a temporary variable and use it for subsequent
+ * calls. Then later reset it once we're done or at the
+ * transaction end.
+ */
+ if (OidIsValid(classForm->relfilenode))
+ relfilenode = classForm->relfilenode;
+ else
+ relfilenode = RelationMapOidToFilenodeForDatabase(srcpath,
+ classForm->oid);
+
+ /* We must have a valid relfilenode oid. */
+ Assert(OidIsValid(relfilenode));
+
+ /* Prepare a rel info element and add it to the list. */
+ relinfo = (CreateDBRelInfo *) palloc(sizeof(CreateDBRelInfo));
+ if (OidIsValid(classForm->reltablespace))
+ relinfo->rnode.spcNode = classForm->reltablespace;
+ else
+ relinfo->rnode.spcNode = tbid;
+
+ relinfo->rnode.dbNode = dbid;
+ relinfo->rnode.relNode = relfilenode;
+ relinfo->reloid = classForm->oid;
+ relinfo->relpersistence = classForm->relpersistence;
+
+ if (rnodelist == NULL)
+ rnodelist = list_make1(relinfo);
+ else
+ rnodelist = lappend(rnodelist, relinfo);
+ }
+ }
+
+ /* Release the buffer lock. */
+ UnlockReleaseBuffer(buf);
+ }
+
+ /* Release the lock. */
+ UnlockRelationId(&relid, AccessShareLock);
+
+ return rnodelist;
+}
+
+/*
+ * RelationCopyStorageUsingBuffer - Copy fork's data using bufmgr.
+ *
+ * Same as RelationCopyStorage but instead of using smgrread and smgrextend this
+ * will copy using bufmgr apis.
+ */
+void
+RelationCopyStorageUsingBuffer(SMgrRelation src, SMgrRelation dst,
+ ForkNumber forkNum, char relpersistence)
+{
+ Buffer srcBuf;
+ Buffer dstBuf;
+ Page srcPage;
+ Page dstPage;
+ bool use_wal;
+ bool copying_initfork;
+ BlockNumber nblocks;
+ BlockNumber blkno;
+ BufferAccessStrategy bstrategy_src;
+ BufferAccessStrategy bstrategy_dst;
+
+ /* Refer comments in RelationCopyStorage. */
+ copying_initfork = relpersistence == RELPERSISTENCE_UNLOGGED &&
+ forkNum == INIT_FORKNUM;
+ use_wal = XLogIsNeeded() &&
+ (relpersistence == RELPERSISTENCE_PERMANENT || copying_initfork);
+
+ /* Get number of blocks in the source relation. */
+ nblocks = smgrnblocks(src, forkNum);
+
+ /*
+ * We are going to copy whole relation from the source to the destination
+ * so use BAS_BULKREAD strategy for the source relation and BAS_BULKWRITE
+ * strategy for the destination relation.
+ */
+ bstrategy_src = GetAccessStrategy(BAS_BULKREAD);
+ bstrategy_dst = GetAccessStrategy(BAS_BULKWRITE);
+
+ /* Iterate over each block of the source relation file. */
+ for (blkno = 0; blkno < nblocks; blkno++)
+ {
+ /* If we got a cancel signal during the copy of the data, quit */
+ CHECK_FOR_INTERRUPTS();
+
+ /* Read block from source relation. */
+ srcBuf = ReadBufferWithoutRelcache(src->smgr_rnode.node, forkNum,
+ blkno, RBM_NORMAL, bstrategy_src,
+ relpersistence);
+ srcPage = BufferGetPage(srcBuf);
+ if (PageIsNew(srcPage) || PageIsEmpty(srcPage))
+ {
+ ReleaseBuffer(srcBuf);
+ continue;
+ }
+
+ /* Use P_NEW to extend the relation. */
+ dstBuf = ReadBufferWithoutRelcache(dst->smgr_rnode.node, forkNum,
+ P_NEW, RBM_NORMAL, bstrategy_dst,
+ relpersistence);
+ LockBuffer(dstBuf, BUFFER_LOCK_EXCLUSIVE);
+
+ START_CRIT_SECTION();
+
+ /* Initialize the page and write the data. */
+ dstPage = BufferGetPage(dstBuf);
+ PageInit(dstPage, BufferGetPageSize(dstBuf), 0);
+ memcpy(dstPage, srcPage, BLCKSZ);
+ MarkBufferDirty(dstBuf);
+
+ /* WAL-log the copied page. */
+ if (use_wal)
+ log_newpage_buffer(dstBuf, true);
+
+ END_CRIT_SECTION();
+
+ UnlockReleaseBuffer(dstBuf);
+ ReleaseBuffer(srcBuf);
+ }
+}
+
+/*
+ * CopyDatabase - Copy source database to the target database.
+ *
+ * Create target database directory and copy data files from the source database
+ * to the target database, block by block and WAL log all the operations.
+ */
+static void
+CopyDatabase(Oid src_dboid, Oid dst_dboid, Oid src_tsid, Oid dst_tsid)
+{
+ char *srcpath;
+ char *dstpath;
+ List *rnodelist = NULL;
+ ListCell *cell;
+ LockRelId relid;
+ RelFileNode srcrnode;
+ RelFileNode dstrnode;
+ CreateDBRelInfo *relinfo;
+
+ /* Get the source database path. */
+ srcpath = GetDatabasePath(src_dboid, src_tsid);
+
+ /* Get the destination database path. */
+ dstpath = GetDatabasePath(dst_dboid, dst_tsid);
+
+ /* Create database directory and write PG_VERSION file. */
+ CreateDirAndVersionFile(dstpath, dst_dboid, dst_tsid, false);
+
+ /* Copy relmap file from source database to the destination database. */
+ CopyRelationMap(dst_dboid, dst_tsid, srcpath, dstpath);
+
+ /* Get list of all valid relnode from the source database. */
+ rnodelist = GetDatabaseRelationList(src_tsid, src_dboid, srcpath);
+ Assert(rnodelist != NIL);
+
+ /*
+ * Database id is common for all the relation so set it before entering to
+ * the loop.
+ */
+ relid.dbId = src_dboid;
+
+ /*
+ * Iterate over each relfilenode and copy the relation data block by block
+ * from source database to the destination database.
+ */
+ foreach(cell, rnodelist)
+ {
+ SMgrRelation src_smgr;
+ SMgrRelation dst_smgr;
+
+ relinfo = lfirst(cell);
+ srcrnode = relinfo->rnode;
+
+ /*
+ * If the relation is from the default tablespace then we need to
+ * create it in the destinations db's default tablespace. Otherwise,
+ * we need to create in the same tablespace as it is in the source
+ * database.
+ */
+ if (srcrnode.spcNode != src_tsid)
+ dstrnode.spcNode = srcrnode.spcNode;
+ else
+ dstrnode.spcNode = dst_tsid;
+
+ dstrnode.dbNode = dst_dboid;
+ dstrnode.relNode = srcrnode.relNode;
+
+ /* Acquire the lock on relation before start copying. */
+ relid.relId = relinfo->reloid;
+ LockRelationId(&relid, AccessShareLock);
+
+ /* Open the source and the destination relation at smgr level. */
+ src_smgr = smgropen(srcrnode, InvalidBackendId);
+ dst_smgr = smgropen(dstrnode, InvalidBackendId);
+
+ /* Copy relation storage from source to the destination. */
+ RelationCopyAllFork(src_smgr, dst_smgr, relinfo->relpersistence,
+ RelationCopyStorageUsingBuffer);
+
+ /* Release the lock. */
+ UnlockRelationId(&relid, AccessShareLock);
+ }
+
+ list_free_deep(rnodelist);
+}
/*
@@ -99,8 +533,6 @@ static int errdetail_busy_db(int notherbackends, int npreparedxacts);
Oid
createdb(ParseState *pstate, const CreatedbStmt *stmt)
{
- TableScanDesc scan;
- Relation rel;
Oid src_dboid;
Oid src_owner;
int src_encoding = -1;
@@ -563,139 +995,27 @@ createdb(ParseState *pstate, const CreatedbStmt *stmt)
InvokeObjectPostCreateHook(DatabaseRelationId, dboid, 0);
/*
- * Force a checkpoint before starting the copy. This will force all dirty
- * buffers, including those of unlogged tables, out to disk, to ensure
- * source database is up-to-date on disk for the copy.
- * FlushDatabaseBuffers() would suffice for that, but we also want to
- * process any pending unlink requests. Otherwise, if a checkpoint
- * happened while we're copying files, a file might be deleted just when
- * we're about to copy it, causing the lstat() call in copydir() to fail
- * with ENOENT.
- */
- RequestCheckpoint(CHECKPOINT_IMMEDIATE | CHECKPOINT_FORCE | CHECKPOINT_WAIT
- | CHECKPOINT_FLUSH_ALL);
-
- /*
- * Once we start copying subdirectories, we need to be able to clean 'em
- * up if we fail. Use an ENSURE block to make sure this happens. (This
- * is not a 100% solution, because of the possibility of failure during
- * transaction commit after we leave this routine, but it should handle
- * most scenarios.)
+ * Once we start copying files from the source database, we need to be able
+ * to clean 'em up if we fail. Use an ENSURE block to make sure this
+ * happens. (This is not a 100% solution, because of the possibility of
+ * failure during transaction commit after we leave this routine, but it
+ * should handle most scenarios.)
*/
fparms.src_dboid = src_dboid;
fparms.dest_dboid = dboid;
PG_ENSURE_ERROR_CLEANUP(createdb_failure_callback,
PointerGetDatum(&fparms));
{
- /*
- * Iterate through all tablespaces of the template database, and copy
- * each one to the new database.
- */
- rel = table_open(TableSpaceRelationId, AccessShareLock);
- scan = table_beginscan_catalog(rel, 0, NULL);
- while ((tuple = heap_getnext(scan, ForwardScanDirection)) != NULL)
- {
- Form_pg_tablespace spaceform = (Form_pg_tablespace) GETSTRUCT(tuple);
- Oid srctablespace = spaceform->oid;
- Oid dsttablespace;
- char *srcpath;
- char *dstpath;
- struct stat st;
-
- /* No need to copy global tablespace */
- if (srctablespace == GLOBALTABLESPACE_OID)
- continue;
-
- srcpath = GetDatabasePath(src_dboid, srctablespace);
-
- if (stat(srcpath, &st) < 0 || !S_ISDIR(st.st_mode) ||
- directory_is_empty(srcpath))
- {
- /* Assume we can ignore it */
- pfree(srcpath);
- continue;
- }
-
- if (srctablespace == src_deftablespace)
- dsttablespace = dst_deftablespace;
- else
- dsttablespace = srctablespace;
-
- dstpath = GetDatabasePath(dboid, dsttablespace);
-
- /*
- * Copy this subdirectory to the new location
- *
- * We don't need to copy subdirectories
- */
- copydir(srcpath, dstpath, false);
-
- /* Record the filesystem change in XLOG */
- {
- xl_dbase_create_rec xlrec;
-
- xlrec.db_id = dboid;
- xlrec.tablespace_id = dsttablespace;
- xlrec.src_db_id = src_dboid;
- xlrec.src_tablespace_id = srctablespace;
-
- XLogBeginInsert();
- XLogRegisterData((char *) &xlrec, sizeof(xl_dbase_create_rec));
-
- (void) XLogInsert(RM_DBASE_ID,
- XLOG_DBASE_CREATE | XLR_SPECIAL_REL_UPDATE);
- }
- }
- table_endscan(scan);
- table_close(rel, AccessShareLock);
-
- /*
- * We force a checkpoint before committing. This effectively means
- * that committed XLOG_DBASE_CREATE operations will never need to be
- * replayed (at least not in ordinary crash recovery; we still have to
- * make the XLOG entry for the benefit of PITR operations). This
- * avoids two nasty scenarios:
- *
- * #1: When PITR is off, we don't XLOG the contents of newly created
- * indexes; therefore the drop-and-recreate-whole-directory behavior
- * of DBASE_CREATE replay would lose such indexes.
- *
- * #2: Since we have to recopy the source database during DBASE_CREATE
- * replay, we run the risk of copying changes in it that were
- * committed after the original CREATE DATABASE command but before the
- * system crash that led to the replay. This is at least unexpected
- * and at worst could lead to inconsistencies, eg duplicate table
- * names.
- *
- * (Both of these were real bugs in releases 8.0 through 8.0.3.)
- *
- * In PITR replay, the first of these isn't an issue, and the second
- * is only a risk if the CREATE DATABASE and subsequent template
- * database change both occur while a base backup is being taken.
- * There doesn't seem to be much we can do about that except document
- * it as a limitation.
- *
- * Perhaps if we ever implement CREATE DATABASE in a less cheesy way,
- * we can avoid this.
- */
- RequestCheckpoint(CHECKPOINT_IMMEDIATE | CHECKPOINT_FORCE | CHECKPOINT_WAIT);
-
- /*
- * Close pg_database, but keep lock till commit.
- */
- table_close(pg_database_rel, NoLock);
-
- /*
- * Force synchronous commit, thus minimizing the window between
- * creation of the database files and committal of the transaction. If
- * we crash before committing, we'll have a DB that's taking up disk
- * space but is not in pg_database, which is not good.
- */
- ForceSyncCommit();
+ CopyDatabase(src_dboid, dboid, src_deftablespace, dst_deftablespace);
}
PG_END_ENSURE_ERROR_CLEANUP(createdb_failure_callback,
PointerGetDatum(&fparms));
+ /*
+ * Close pg_database, but keep lock till commit.
+ */
+ table_close(pg_database_rel, NoLock);
+
return dboid;
}
@@ -938,7 +1258,7 @@ dropdb(const char *dbname, bool missing_ok, bool force)
* is important to ensure that no remaining backend tries to write out a
* dirty buffer to the dead database later...
*/
- DropDatabaseBuffers(db_id);
+ DropDatabaseBuffers(db_id, InvalidOid);
/*
* Tell the stats collector to forget it immediately, too.
@@ -1196,37 +1516,6 @@ movedb(const char *dbname, const char *tblspcname)
dst_dbpath = GetDatabasePath(db_id, dst_tblspcoid);
/*
- * Force a checkpoint before proceeding. This will force all dirty
- * buffers, including those of unlogged tables, out to disk, to ensure
- * source database is up-to-date on disk for the copy.
- * FlushDatabaseBuffers() would suffice for that, but we also want to
- * process any pending unlink requests. Otherwise, the check for existing
- * files in the target directory might fail unnecessarily, not to mention
- * that the copy might fail due to source files getting deleted under it.
- * On Windows, this also ensures that background procs don't hold any open
- * files, which would cause rmdir() to fail.
- */
- RequestCheckpoint(CHECKPOINT_IMMEDIATE | CHECKPOINT_FORCE | CHECKPOINT_WAIT
- | CHECKPOINT_FLUSH_ALL);
-
- /*
- * Now drop all buffers holding data of the target database; they should
- * no longer be dirty so DropDatabaseBuffers is safe.
- *
- * It might seem that we could just let these buffers age out of shared
- * buffers naturally, since they should not get referenced anymore. The
- * problem with that is that if the user later moves the database back to
- * its original tablespace, any still-surviving buffers would appear to
- * contain valid data again --- but they'd be missing any changes made in
- * the database while it was in the new tablespace. In any case, freeing
- * buffers that should never be used again seems worth the cycles.
- *
- * Note: it'd be sufficient to get rid of buffers matching db_id and
- * src_tblspcoid, but bufmgr.c presently provides no API for that.
- */
- DropDatabaseBuffers(db_id);
-
- /*
* Check for existence of files in the target directory, i.e., objects of
* this database that are already in the target tablespace. We can't
* allow the move in such a case, because we would need to change those
@@ -1261,38 +1550,16 @@ movedb(const char *dbname, const char *tblspcname)
}
/*
- * Use an ENSURE block to make sure we remove the debris if the copy fails
- * (eg, due to out-of-disk-space). This is not a 100% solution, because
- * of the possibility of failure during transaction commit, but it should
- * handle most scenarios.
+ * Use an ENSURE block to make sure we remove the debris if the copy fails.
+ * This is not a 100% solution, because of the possibility of failure
+ * during transaction commit, but it should handle most scenarios.
*/
fparms.dest_dboid = db_id;
fparms.dest_tsoid = dst_tblspcoid;
PG_ENSURE_ERROR_CLEANUP(movedb_failure_callback,
PointerGetDatum(&fparms));
{
- /*
- * Copy files from the old tablespace to the new one
- */
- copydir(src_dbpath, dst_dbpath, false);
-
- /*
- * Record the filesystem change in XLOG
- */
- {
- xl_dbase_create_rec xlrec;
-
- xlrec.db_id = db_id;
- xlrec.tablespace_id = dst_tblspcoid;
- xlrec.src_db_id = db_id;
- xlrec.src_tablespace_id = src_tblspcoid;
-
- XLogBeginInsert();
- XLogRegisterData((char *) &xlrec, sizeof(xl_dbase_create_rec));
-
- (void) XLogInsert(RM_DBASE_ID,
- XLOG_DBASE_CREATE | XLR_SPECIAL_REL_UPDATE);
- }
+ CopyDatabase(db_id, db_id, src_tblspcoid, dst_tblspcoid);
/*
* Update the database's pg_database tuple
@@ -1326,22 +1593,6 @@ movedb(const char *dbname, const char *tblspcname)
systable_endscan(sysscan);
/*
- * Force another checkpoint here. As in CREATE DATABASE, this is to
- * ensure that we don't have to replay a committed XLOG_DBASE_CREATE
- * operation, which would cause us to lose any unlogged operations
- * done in the new DB tablespace before the next checkpoint.
- */
- RequestCheckpoint(CHECKPOINT_IMMEDIATE | CHECKPOINT_FORCE | CHECKPOINT_WAIT);
-
- /*
- * Force synchronous commit, thus minimizing the window between
- * copying the database files and committal of the transaction. If we
- * crash before committing, we'll leave an orphaned set of files on
- * disk, which is not fatal but not good either.
- */
- ForceSyncCommit();
-
- /*
* Close pg_database, but keep lock till commit.
*/
table_close(pgdbrel, NoLock);
@@ -1350,6 +1601,21 @@ movedb(const char *dbname, const char *tblspcname)
PointerGetDatum(&fparms));
/*
+ * Now drop all buffers holding data of the target database for the old
+ * tablespace oid; We have already copied all the data to the new
+ * tablespace so we no longer required the old buffers.
+ *
+ * It might seem that we could just let these buffers age out of shared
+ * buffers naturally, since they should not get referenced anymore. The
+ * problem with that is that if the user later moves the database back to
+ * its original tablespace, any still-surviving buffers would appear to
+ * contain valid data again --- but they'd be missing any changes made in
+ * the database while it was in the new tablespace. In any case, freeing
+ * buffers that should never be used again seems worth the cycles.
+ */
+ DropDatabaseBuffers(db_id, src_tblspcoid);
+
+ /*
* Commit the transaction so that the pg_database update is committed. If
* we crash while removing files, the database won't be corrupt, we'll
* just leave some orphaned files in the old directory.
@@ -2141,39 +2407,11 @@ dbase_redo(XLogReaderState *record)
if (info == XLOG_DBASE_CREATE)
{
xl_dbase_create_rec *xlrec = (xl_dbase_create_rec *) XLogRecGetData(record);
- char *src_path;
- char *dst_path;
- struct stat st;
-
- src_path = GetDatabasePath(xlrec->src_db_id, xlrec->src_tablespace_id);
- dst_path = GetDatabasePath(xlrec->db_id, xlrec->tablespace_id);
-
- /*
- * Our theory for replaying a CREATE is to forcibly drop the target
- * subdirectory if present, then re-copy the source data. This may be
- * more work than needed, but it is simple to implement.
- */
- if (stat(dst_path, &st) == 0 && S_ISDIR(st.st_mode))
- {
- if (!rmtree(dst_path, true))
- /* If this failed, copydir() below is going to error. */
- ereport(WARNING,
- (errmsg("some useless files may be left behind in old database directory \"%s\"",
- dst_path)));
- }
-
- /*
- * Force dirty buffers out to disk, to ensure source database is
- * up-to-date for the copy.
- */
- FlushDatabaseBuffers(xlrec->src_db_id);
+ char *dbpath;
- /*
- * Copy this subdirectory to the new location
- *
- * We don't need to copy subdirectories
- */
- copydir(src_path, dst_path, false);
+ dbpath = GetDatabasePath(xlrec->db_id, xlrec->tablespace_id);
+ CreateDirAndVersionFile(dbpath, xlrec->db_id, xlrec->tablespace_id,
+ true);
}
else if (info == XLOG_DBASE_DROP)
{
@@ -2201,7 +2439,7 @@ dbase_redo(XLogReaderState *record)
ReplicationSlotsDropDBSlots(xlrec->db_id);
/* Drop pages for this database that are in the shared buffer cache */
- DropDatabaseBuffers(xlrec->db_id);
+ DropDatabaseBuffers(xlrec->db_id, InvalidOid);
/* Also, clean out any fsync requests that might be pending in md.c */
ForgetDatabaseSyncRequests(xlrec->db_id);
diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c
index 59ebac7..f71b344 100644
--- a/src/bin/pg_rewind/parsexlog.c
+++ b/src/bin/pg_rewind/parsexlog.c
@@ -23,6 +23,7 @@
#include "fe_utils/archive.h"
#include "filemap.h"
#include "pg_rewind.h"
+#include "utils/relmapper.h"
/*
* RmgrNames is an array of resource manager names, to make error messages
diff --git a/src/include/commands/dbcommands_xlog.h b/src/include/commands/dbcommands_xlog.h
index f5ed762..21dc58e 100644
--- a/src/include/commands/dbcommands_xlog.h
+++ b/src/include/commands/dbcommands_xlog.h
@@ -23,11 +23,8 @@
typedef struct xl_dbase_create_rec
{
- /* Records copying of a single subdirectory incl. contents */
Oid db_id;
Oid tablespace_id;
- Oid src_db_id;
- Oid src_tablespace_id;
} xl_dbase_create_rec;
typedef struct xl_dbase_drop_rec
--
1.8.3.1
