On Tue, Mar 1, 2022 at 5:15 PM Dilip Kumar <dilipbal...@gmail.com> wrote:
> On Tue, Feb 22, 2022 at 8:27 PM Ashutosh Sharma <ashu.coe...@gmail.com>
> wrote:
>
>> I'm not sure about the current status, but found it while playing
>> around with the latest changes a bit, so thought of sharing it here.
>>
>> + <varlistentry>
>> + <term><replaceable class="parameter">strategy</replaceable></term>
>> + <listitem>
>> + <para>
>> + This is used for copying the database directory. Currently, we
>> have
>> + two strategies the <literal>WAL_LOG_BLOCK</literal> and the
>>
>> Isn't it wal_log instead of wal_log_block?
>>
>> I think when users input wrong strategy with createdb command, we
>> should provide a hint message showing allowed values for strategy
>> types along with an error message. This will be helpful for the users.
>>
>
> I will fix these two comments while posting the next version.
>
>
The new version of the patch fixes these 2 comments pointed by Ashutosh and
also splits the GetRelListFromPage() function as suggested by Robert and
uses the latest snapshot for scanning the pg_class instead of active
snapshot as pointed out by Robert. I haven't yet added the test case to
create a database using this new strategy option. So if we are okay with
these two options FILE_COPY and WAL_LOG then I will add test cases for the
same.
--
Regards,
Dilip Kumar
EnterpriseDB: http://www.enterprisedb.com
From ba8e991e9b24ed8986d4504c44ce85f78cc4a598 Mon Sep 17 00:00:00 2001
From: dilipkumar <dilipbal...@gmail.com>
Date: Mon, 4 Oct 2021 13:50:44 +0530
Subject: [PATCH v10 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) And another interface for getting filenode from oid. We already
have RelationMapOidToFilenode for the same purpose but that assumes
we are connected to the database for which we want to get the mapping.
So this new interface will do the same but instead, it will get the
mapping for the input database.
These interfaces are required for next patch, for supporting the
wal logged created database.
---
src/backend/utils/cache/relmapper.c | 123 +++++++++++++++++++++++++++++++-----
src/include/utils/relmapper.h | 6 +-
2 files changed, 113 insertions(+), 16 deletions(-)
diff --git a/src/backend/utils/cache/relmapper.c b/src/backend/utils/cache/relmapper.c
index 56495f0..86a85c8 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 +256,37 @@ RelationMapFilenodeToOid(Oid filenode, bool shared)
}
/*
+ * RelationMapOidToFilenodeForDatabase
+ *
+ * Same as RelationMapOidToFilenode, but instead of reading the mapping from
+ * the database we are connected to it will read the mapping from the input
+ * database.
+ */
+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 +724,43 @@ 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 +863,18 @@ 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 +900,7 @@ 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 +1042,7 @@ 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 +1134,37 @@ 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 9fbb5a7..e5635bd 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 37b67898404982119116cea73eab2bf750472c69 Mon Sep 17 00:00:00 2001
From: Dilip Kumar <dilipkumar@localhost.localdomain>
Date: Wed, 1 Sep 2021 14:06:29 +0530
Subject: [PATCH v10 1/6] Refactor relmap load and relmap write functions
Currently, relmap reading and writing interfaces are tightly
coupled with shared_map and local_map of the database
it is connected to. But as higher level patch set we need
interfaces where we can read relmap into any input memory
and while writing also we should be able to pass the map.
So as part of this patch, we are doing refactoring of the
existing code such that we can expose the read and write
interfaces that are independent of the shared_map and the
local_map, without changing any logic.
XXX For the code simplicity in write_relmap_file we are
updating the permanent memory copy outside the critical
section but we have already done the disk changes and it
is just a memory change so there is no reason for this
to be in the critical section.
---
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 4f6811f..56495f0 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 44b2398e89fdf4dee1bd072d7d19299e4fac45ae Mon Sep 17 00:00:00 2001
From: Dilip Kumar <dilipkumar@localhost.localdomain>
Date: Fri, 24 Sep 2021 18:13:25 +0530
Subject: [PATCH v10 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 | 68 +++++++++++++++++++++++++---------------
src/include/commands/tablecmds.h | 5 +++
src/tools/pgindent/typedefs.list | 1 +
3 files changed, 48 insertions(+), 26 deletions(-)
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index 3e83f37..a57d6b0 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -14580,54 +14580,70 @@ AlterTableMoveAll(AlterTableMoveAllStmt *stmt)
return new_tablespaceoid;
}
-static void
-index_copy_data(Relation rel, RelFileNode newrnode)
+/*
+ * Copy source smgr relation's all fork's data to the destination.
+ *
+ * copy_storage - storage copy function, which is passed by the caller.
+ */
+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.
+ * Create and copy all forks of the relation.
*
* 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);
+
+ /*
+ * Create and copy all forks of the relation, and schedule unlinking of
+ * old physical files.
+ */
+ 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 5d4037f..cd49471 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);
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index d9b83f7..c1400d3 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -3059,6 +3059,7 @@ config_var_value
contain_aggs_of_level_context
convert_testexpr_context
copy_data_source_cb
+copy_relation_storage
core_YYSTYPE
core_yy_extra_type
core_yyscan_t
--
1.8.3.1
From 824cc17de6f9ca027cc6cfae60f73086490a23db Mon Sep 17 00:00:00 2001
From: Dilip Kumar <dilipkumar@localhost.localdomain>
Date: Thu, 10 Feb 2022 15:55:33 +0530
Subject: [PATCH v10 4/6] Extend bufmgr interfaces
Extend ReadBufferWithoutRelcache interface to take relpersistence
as input. At present, this function may only be used on permanent
relations, because we only use it during XLOG replay. But now as
part of the bigger patch set, we will be using this for reading the
buffer from the database which we are not connected so now we might
have temporary and unlogged relations as well.
---
src/backend/access/transam/xlogutils.c | 9 ++++++---
src/backend/storage/buffer/bufmgr.c | 13 +++----------
src/include/storage/bufmgr.h | 3 ++-
3 files changed, 11 insertions(+), 14 deletions(-)
diff --git a/src/backend/access/transam/xlogutils.c b/src/backend/access/transam/xlogutils.c
index 54d5f20..c292794 100644
--- a/src/backend/access/transam/xlogutils.c
+++ b/src/backend/access/transam/xlogutils.c
@@ -484,7 +484,8 @@ XLogReadBufferExtended(RelFileNode rnode, ForkNumber forknum,
{
/* page exists in file */
buffer = ReadBufferWithoutRelcache(rnode, forknum, blkno,
- mode, NULL);
+ mode, NULL,
+ RELPERSISTENCE_PERMANENT);
}
else
{
@@ -509,7 +510,8 @@ XLogReadBufferExtended(RelFileNode rnode, ForkNumber forknum,
ReleaseBuffer(buffer);
}
buffer = ReadBufferWithoutRelcache(rnode, forknum,
- P_NEW, mode, NULL);
+ P_NEW, mode, NULL,
+ RELPERSISTENCE_PERMANENT);
}
while (BufferGetBlockNumber(buffer) < blkno);
/* Handle the corner case that P_NEW returns non-consecutive pages */
@@ -519,7 +521,8 @@ XLogReadBufferExtended(RelFileNode rnode, ForkNumber forknum,
LockBuffer(buffer, BUFFER_LOCK_UNLOCK);
ReleaseBuffer(buffer);
buffer = ReadBufferWithoutRelcache(rnode, forknum, blkno,
- mode, NULL);
+ mode, NULL,
+ RELPERSISTENCE_PERMANENT);
}
}
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index f5459c6..d6d366a 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -771,24 +771,17 @@ 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);
}
@@ -798,7 +791,7 @@ 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)
diff --git a/src/include/storage/bufmgr.h b/src/include/storage/bufmgr.h
index dd01841..7b80f58 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);
--
1.8.3.1
From 93e3ed5a9d0b385f2a62cba5c96023bb20c97845 Mon Sep 17 00:00:00 2001
From: Dilip Kumar <dilipkumar@localhost.localdomain>
Date: Fri, 24 Sep 2021 18:29:17 +0530
Subject: [PATCH v10 5/6] New interface to lock relation id
Currently, we have LockRelationOid which provide a mechanism to
lock the relation oid but we must be connected to the database
from which this relation belong. As part of this patch we are
providing a new interface which can lock the relation even if we
are not connected to the containing 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 5ae52dd..1543da6 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 49edbcc..be1d2c9 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 2632da8b5770ac67633fbadbcdfaffc3ceeab277 Mon Sep 17 00:00:00 2001
From: Dilip Kumar <dilipkumar@localhost.localdomain>
Date: Mon, 14 Feb 2022 17:48:03 +0530
Subject: [PATCH v10 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.
We are also maintaining the old way of creating the database and for that we
are providing an option to choose the strategy for creating the database.
For the new method the user need to give STRATEGY=WAL_LOG and for the
old method they need to give STRATEGY=FILE_COPY. The default strategy will
be WAL_LOG.
---
doc/src/sgml/ref/create_database.sgml | 22 +
src/backend/commands/dbcommands.c | 858 ++++++++++++++++++++++++++++-----
src/include/commands/dbcommands_xlog.h | 8 +
src/tools/pgindent/typedefs.list | 1 +
4 files changed, 757 insertions(+), 132 deletions(-)
diff --git a/doc/src/sgml/ref/create_database.sgml b/doc/src/sgml/ref/create_database.sgml
index f70d0c7..a906cc8 100644
--- a/doc/src/sgml/ref/create_database.sgml
+++ b/doc/src/sgml/ref/create_database.sgml
@@ -34,6 +34,7 @@ CREATE DATABASE <replaceable class="parameter">name</replaceable>
[ CONNECTION LIMIT [=] <replaceable class="parameter">connlimit</replaceable> ]
[ IS_TEMPLATE [=] <replaceable class="parameter">istemplate</replaceable> ]
[ OID [=] <replaceable class="parameter">oid</replaceable> ] ]
+ [ STRATEGY [=] <replaceable class="parameter">strategy</replaceable> ] ]
</synopsis>
</refsynopsisdiv>
@@ -240,6 +241,27 @@ CREATE DATABASE <replaceable class="parameter">name</replaceable>
</listitem>
</varlistentry>
+ <varlistentry>
+ <term><replaceable class="parameter">strategy</replaceable></term>
+ <listitem>
+ <para>
+ This is used for copying the database directory. Currently, we have
+ two strategies the <literal>WAL_LOG</literal> and the
+ <literal>FILE_COPY</literal>. If <literal>WAL_LOG</literal> strategy
+ is used then the database will be copied block by block and it will
+ also WAL log each copied block. Otherwise, if <literal>FILE_COPY
+ </literal> strategy is used then it will do the file system level copy
+ so individual the block is not WAL logged. If the <literal>FILE_COPY
+ </literal> strategy is used then it has to issue a checkpoint before
+ and after performing the copy and if the shared buffers are large and
+ there are a lot of dirty buffers then issuing checkpoint would be
+ costly and it may impact the performance of the whole system. On the
+ other hand, if we WAL log each block then if the source database is
+ large then creating the database may take more time.
+ </para>
+ </listitem>
+ </varlistentry>
+
</variablelist>
<para>
diff --git a/src/backend/commands/dbcommands.c b/src/backend/commands/dbcommands.c
index c37e3c9..255570f 100644
--- a/src/backend/commands/dbcommands.c
+++ b/src/backend/commands/dbcommands.c
@@ -46,6 +46,7 @@
#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"
@@ -63,13 +64,27 @@
#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"
+/*
+ * Create database strategy. The CREATEDB_WAL_LOG will copy the database at
+ * the block level and WAL log each copied block. Whereas the
+ * CREATEDB_FILE_COPY will directly copy the database at the file level and no
+ * individual operations will be WAL logged.
+ */
+typedef enum CreateDBStrategy
+{
+ CREATEDB_WAL_LOG = 0,
+ CREATEDB_FILE_COPY = 1
+} CreateDBStrategy;
+
typedef struct
{
Oid src_dboid; /* source (template) DB */
Oid dest_dboid; /* DB we are trying to create */
+ CreateDBStrategy strategy; /* create db strategy */
} createdb_failure_params;
typedef struct
@@ -78,6 +93,19 @@ 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);
@@ -92,7 +120,607 @@ 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 CreateDBRelInfo *GetRelInfoFromTuple(HeapTupleData *tuple,
+ Oid tbid, Oid dbid, char *srcpath);
+static List *GetRelListFromPage(Page page, Buffer buf, Oid tbid, Oid dbid,
+ char *srcpath, List *rnodelist, Snapshot
+ snapshot);
+static List *GetDatabaseRelationList(Oid srctbid, Oid srcdbid, char *srcpath);
+static void RelationCopyStorageUsingBuffer(SMgrRelation src, SMgrRelation dst,
+ ForkNumber forkNum, char relpersistence);
+static void CopyDatabaseWithWal(Oid src_dboid, Oid dboid, Oid src_tsid,
+ Oid dst_tsid);
+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;
+ char versionfile[MAXPGPATH];
+ char buf[16];
+
+ /* Prepare version data before starting a critical section. */
+ sprintf(buf, "%s\n", PG_MAJORVERSION);
+ nbytes = strlen(PG_MAJORVERSION) + 1;
+
+ /* 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;
+ xlrec.src_db_id = InvalidOid;
+ xlrec.src_tablespace_id = InvalidOid;
+
+ 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 write mode.
+ */
+ snprintf(versionfile, sizeof(versionfile), "%s/%s", dbpath, "PG_VERSION");
+
+ fd = OpenTransientFile(versionfile, O_WRONLY | O_CREAT | O_EXCL | PG_BINARY);
+ if (fd < 0 && errno == EEXIST && isRedo)
+ fd = OpenTransientFile(versionfile, O_WRONLY | O_TRUNC | 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, buf, 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();
+}
+
+/*
+ * GetRelInfoFromTuple - Prepare a CreateDBRelInfo element from the tuple
+ *
+ * Helper function for GetRelListFromPage to prepare a single element from the
+ * pg_class tuple.
+ */
+CreateDBRelInfo *
+GetRelInfoFromTuple(HeapTupleData *tuple, Oid tbid, Oid dbid, char *srcpath)
+{
+ CreateDBRelInfo *relinfo;
+ Form_pg_class classForm;
+ Oid relfilenode = InvalidOid;
+
+ classForm = (Form_pg_class) GETSTRUCT(tuple);
+
+ /* We don't need to copy the shared objects to the target. */
+ if (classForm->reltablespace == GLOBALTABLESPACE_OID)
+ return NULL;
+
+ /*
+ * 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))
+ return NULL;
+
+ /*
+ * If relfilenode is valid then directly use it. Otherwise,
+ * consult the relmapper for the mapped relation.
+ */
+ 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;
+
+ return relinfo;
+}
+
+/*
+ * GetRelListFromPage - Helper function for GetDatabaseRelationList.
+ *
+ * Iterate over each tuple of input pg_class and get a list of all the valid
+ * relfilenodes of the given block and append them to input rnodelist.
+ */
+static List *
+GetRelListFromPage(Page page, Buffer buf, Oid tbid, Oid dbid, char *srcpath,
+ List *rnodelist, Snapshot snapshot)
+{
+ BlockNumber blkno = BufferGetBlockNumber(buf);
+ OffsetNumber offnum;
+ OffsetNumber maxoff;
+ HeapTupleData tuple;
+
+ 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, snapshot, buf))
+ {
+ CreateDBRelInfo *relinfo;
+
+ relinfo = GetRelInfoFromTuple(&tuple, tbid, dbid, srcpath);
+ /* Add it to the list. */
+ if (relinfo != NULL)
+ rnodelist = lappend(rnodelist, relinfo);
+ }
+ }
+
+ return rnodelist;
+}
+
+/*
+ * 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;
+ Buffer buf;
+ Oid relfilenode;
+ Page page;
+ List *rnodelist = NIL;
+ LockRelId relid;
+ Snapshot snapshot;
+ 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);
+
+ /* Get latest snapshot for scanning the pg_class. */
+ snapshot = GetLatestSnapshot();
+
+ /* 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))
+ {
+ UnlockReleaseBuffer(buf);
+ continue;
+ }
+
+ /*
+ * Process pg_class tuple for the current page and add all the valid
+ * relfilenode entries to the rnodelist.
+ */
+ rnodelist = GetRelListFromPage(page, buf, tbid, dbid, srcpath,
+ rnodelist, snapshot);
+
+ /* 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.
+ */
+static 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);
+ }
+}
+
+/*
+ * CopyDatabaseWithWal - Copy source database to the target database with WAL
+ *
+ * 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
+CopyDatabaseWithWal(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 source db's 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 = dst_tsid;
+ /*
+ * In case of ALTER DATABASE SET TABLESPACE we don't need to do
+ * anything for the object which are not in the source db's default
+ * tablespace. The source and destination dboid will be same in
+ * case of ALTER DATABASE SET TABLESPACE.
+ */
+ else if (src_dboid == dst_dboid)
+ continue;
+ else
+ dstrnode.spcNode = srcrnode.spcNode;
+
+ 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);
+}
+
+/*
+ * CopyDatabase - Copy source database to the target database
+ *
+ * Copy source database directory to the destination directory using copydir
+ * operation.
+ */
+static void
+CopyDatabase(Oid src_dboid, Oid dst_dboid, Oid src_tsid, Oid dst_tsid)
+{
+ TableScanDesc scan;
+ Relation rel;
+ HeapTuple tuple;
+
+ /*
+ * 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);
+
+ /*
+ * 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_tsid)
+ dsttablespace = dst_tsid;
+ else
+ dsttablespace = srctablespace;
+
+ dstpath = GetDatabasePath(dst_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 = dst_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);
+}
/*
* CREATE DATABASE
@@ -100,8 +728,6 @@ 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;
@@ -132,6 +758,7 @@ createdb(ParseState *pstate, const CreatedbStmt *stmt)
DefElem *dallowconnections = NULL;
DefElem *dconnlimit = NULL;
DefElem *dcollversion = NULL;
+ DefElem *dstrategy = NULL;
char *dbname = stmt->dbname;
char *dbowner = NULL;
const char *dbtemplate = NULL;
@@ -145,6 +772,7 @@ createdb(ParseState *pstate, const CreatedbStmt *stmt)
char *dbcollversion = NULL;
int notherbackends;
int npreparedxacts;
+ CreateDBStrategy dbstrategy = CREATEDB_WAL_LOG;
createdb_failure_params fparms;
/* Extract options from the statement node tree */
@@ -250,6 +878,12 @@ createdb(ParseState *pstate, const CreatedbStmt *stmt)
(errcode(ERRCODE_INVALID_PARAMETER_VALUE)),
errmsg("OIDs less than %u are reserved for system objects", FirstNormalObjectId));
}
+ else if (strcmp(defel->defname, "strategy") == 0)
+ {
+ if (dstrategy)
+ errorConflictingDefElem(defel, pstate);
+ dstrategy = defel;
+ }
else
ereport(ERROR,
(errcode(ERRCODE_SYNTAX_ERROR),
@@ -374,6 +1008,23 @@ createdb(ParseState *pstate, const CreatedbStmt *stmt)
dbtemplate)));
}
+ /* Validate the database creation strategy. */
+ if (dstrategy && dstrategy->arg)
+ {
+ char *strategy;
+
+ strategy = defGetString(dstrategy);
+ if (strcmp(strategy, "wal_log") == 0)
+ dbstrategy = CREATEDB_WAL_LOG;
+ else if (strcmp(strategy, "file_copy") == 0)
+ dbstrategy = CREATEDB_FILE_COPY;
+ else
+ ereport(ERROR,
+ (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
+ errmsg("invalid create strategy %s", strategy),
+ errhint("Valid strategies are \"wal_log\", and \"file_copy\".")));
+ }
+
/* If encoding or locales are defaulted, use source's setting */
if (encoding < 0)
encoding = src_encoding;
@@ -668,19 +1319,6 @@ 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
@@ -689,114 +1327,47 @@ createdb(ParseState *pstate, const CreatedbStmt *stmt)
*/
fparms.src_dboid = src_dboid;
fparms.dest_dboid = dboid;
+ fparms.strategy = dbstrategy;
+
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.
+ * If the user has asked to create a database with WAL_LOG strategy
+ * then call CopyDatabaseWithWal, which will copy the database at the
+ * block level and it will WAL log each copied block. Otherwise,
+ * call CopyDatabase that will copy the database file by file.
*/
- rel = table_open(TableSpaceRelationId, AccessShareLock);
- scan = table_beginscan_catalog(rel, 0, NULL);
- while ((tuple = heap_getnext(scan, ForwardScanDirection)) != NULL)
+ if (dbstrategy == CREATEDB_WAL_LOG)
{
- 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);
+ CopyDatabaseWithWal(src_dboid, dboid, src_deftablespace,
+ dst_deftablespace);
/*
- * Copy this subdirectory to the new location
- *
- * We don't need to copy subdirectories
+ * Close pg_database, but keep lock till commit.
*/
- 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_close(pg_database_rel, NoLock);
}
- table_endscan(scan);
- table_close(rel, AccessShareLock);
+ else
+ {
+ Assert(dbstrategy == CREATEDB_FILE_COPY);
- /*
- * 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);
+ CopyDatabase(src_dboid, dboid, src_deftablespace,
+ dst_deftablespace);
- /*
- * Close pg_database, but keep lock till commit.
- */
- table_close(pg_database_rel, NoLock);
+ /*
+ * 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();
+ /*
+ * 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();
+ }
}
PG_END_ENSURE_ERROR_CLEANUP(createdb_failure_callback,
PointerGetDatum(&fparms));
@@ -870,6 +1441,21 @@ createdb_failure_callback(int code, Datum arg)
createdb_failure_params *fparms = (createdb_failure_params *) DatumGetPointer(arg);
/*
+ * If we were copying database at block levels then drop pages for the
+ * destination database that are in the shared buffer cache. And tell
+ * checkpointer to forget any pending fsync and unlink requests for
+ * files in the database. The reasoning behind doing this is same as
+ * explained in dropdb function. But unlike dropdb we don't need to call
+ * pgstat_drop_database because this database is still not created so there
+ * should not be any stat for this.
+ */
+ if (fparms->strategy == CREATEDB_WAL_LOG)
+ {
+ DropDatabaseBuffers(fparms->dest_dboid);
+ ForgetDatabaseSyncRequests(fparms->dest_dboid);
+ }
+
+ /*
* Release lock on source database before doing recursive remove. This is
* not essential but it seems desirable to release the lock as soon as
* possible.
@@ -2387,32 +2973,40 @@ dbase_redo(XLogReaderState *record)
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 (!OidIsValid(xlrec->src_db_id))
{
- 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)));
+ CreateDirAndVersionFile(dst_path, xlrec->db_id, xlrec->tablespace_id,
+ true);
}
+ else
+ {
+ /*
+ * 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);
+ /*
+ * Force dirty buffers out to disk, to ensure source database is
+ * up-to-date for the copy.
+ */
+ FlushDatabaseBuffers(xlrec->src_db_id);
- /*
- * Copy this subdirectory to the new location
- *
- * We don't need to copy subdirectories
- */
- copydir(src_path, dst_path, false);
+ /*
+ * Copy this subdirectory to the new location
+ *
+ * We don't need to copy subdirectories
+ */
+ copydir(src_path, dst_path, false);
+ }
}
else if (info == XLOG_DBASE_DROP)
{
diff --git a/src/include/commands/dbcommands_xlog.h b/src/include/commands/dbcommands_xlog.h
index 593a857..8f59870 100644
--- a/src/include/commands/dbcommands_xlog.h
+++ b/src/include/commands/dbcommands_xlog.h
@@ -20,6 +20,7 @@
/* record types */
#define XLOG_DBASE_CREATE 0x00
#define XLOG_DBASE_DROP 0x10
+#define XLOG_DBASE_CREATEDIR 0x20
typedef struct xl_dbase_create_rec
{
@@ -30,6 +31,13 @@ typedef struct xl_dbase_create_rec
Oid src_tablespace_id;
} xl_dbase_create_rec;
+typedef struct xl_dbase_createdir_rec
+{
+ /* Records creating database directory */
+ Oid db_id;
+ Oid tablespace_id;
+} xl_dbase_createdir_rec;
+
typedef struct xl_dbase_drop_rec
{
Oid db_id;
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index c1400d3..317c2f2 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -460,6 +460,7 @@ CoverPos
CreateAmStmt
CreateCastStmt
CreateConversionStmt
+CreateDBRelInfo
CreateDomainStmt
CreateEnumStmt
CreateEventTrigStmt
--
1.8.3.1
