pg_upgrade
Tom asked about pg_upgrade as part of our initdb for timezone.
I have made some improvements to pg_upgrade in CVS and have successfully
migrated a regression database from a 7.2 to another 7.2 database using
it. (At least the tables show some data; very light testing.)
pg_upgrade is still disabled in CVS, it doesn't install, and there is no
manual page so it is still an unused command. I have made the commit so
people can review where I have gone and make comments.
To test it, you have to find the line that says 7.2 and remove the '#'
comment. This is for testing purposes only, so far.
--
Bruce Momjian | http://candle.pha.pa.us
pgman@candle.pha.pa.us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026Bruce Momjian wrote:
Tom asked about pg_upgrade as part of our initdb for timezone.
I have made some improvements to pg_upgrade in CVS and have successfully
migrated a regression database from a 7.2 to another 7.2 database using
it. (At least the tables show some data; very light testing.)
Here is a patch I need to /contrib/pg_resetxlog to support a new "-x
XID" option to set the XID in pg_control. Patch attached. This is the
last feature I needed for a functioning pg_upgrade for 7.1->7.2 and
7.2->7.2 databases.
Many commercial distributions like this script, and with our
newly-needed initdb to fix our timezonetz problem, it seemed like a good
time. :-) It certainly reduces upgrade time.
(BTW, where are we on that timezonetz patch anyway? Tom posted it two
days ago and I haven't seen any comments.)
pg_upgrade is still disabled.
--
Bruce Momjian | http://candle.pha.pa.us
pgman@candle.pha.pa.us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026Attachments:
/bjm/difftext/plainDownload
Index: contrib/pg_resetxlog/README.pg_resetxlog
===================================================================
RCS file: /cvsroot/pgsql/contrib/pg_resetxlog/README.pg_resetxlog,v
retrieving revision 1.1
diff -c -r1.1 README.pg_resetxlog
*** contrib/pg_resetxlog/README.pg_resetxlog 2001/03/14 00:57:43 1.1
--- contrib/pg_resetxlog/README.pg_resetxlog 2002/01/10 06:09:04
***************
*** 21,26 ****
--- 21,29 ----
Then run pg_resetxlog, and finally install and start the new version of
the database software.
+ A tertiary purpose it its use by pg_upgrade to set the next transaction
+ id in pg_control.
+
To run the program, make sure your postmaster is not running, then
(as the Postgres admin user) do
Index: contrib/pg_resetxlog/pg_resetxlog.c
===================================================================
RCS file: /cvsroot/pgsql/contrib/pg_resetxlog/pg_resetxlog.c,v
retrieving revision 1.10
diff -c -r1.10 pg_resetxlog.c
*** contrib/pg_resetxlog/pg_resetxlog.c 2001/11/05 17:46:23 1.10
--- contrib/pg_resetxlog/pg_resetxlog.c 2002/01/10 06:09:05
***************
*** 709,742 ****
* Write out the new pg_control file.
*/
static void
! RewriteControlFile(void)
{
int fd;
char buffer[BLCKSZ]; /* need not be aligned */
! /*
! * Adjust fields as needed to force an empty XLOG starting at the next
! * available segment.
! */
! newXlogId = ControlFile.logId;
! newXlogSeg = ControlFile.logSeg;
! /* be sure we wrap around correctly at end of a logfile */
! NextLogSeg(newXlogId, newXlogSeg);
!
! ControlFile.checkPointCopy.redo.xlogid = newXlogId;
! ControlFile.checkPointCopy.redo.xrecoff =
! newXlogSeg * XLogSegSize + SizeOfXLogPHD;
! ControlFile.checkPointCopy.undo = ControlFile.checkPointCopy.redo;
! ControlFile.checkPointCopy.time = time(NULL);
!
! ControlFile.state = DB_SHUTDOWNED;
! ControlFile.time = time(NULL);
! ControlFile.logId = newXlogId;
! ControlFile.logSeg = newXlogSeg + 1;
! ControlFile.checkPoint = ControlFile.checkPointCopy.redo;
! ControlFile.prevCheckPoint.xlogid = 0;
! ControlFile.prevCheckPoint.xrecoff = 0;
!
/* Contents are protected with a CRC */
INIT_CRC64(ControlFile.crc);
COMP_CRC64(ControlFile.crc,
--- 709,747 ----
* Write out the new pg_control file.
*/
static void
! RewriteControlFile(TransactionId set_xid)
{
int fd;
char buffer[BLCKSZ]; /* need not be aligned */
! if (set_xid == 0)
! {
! /*
! * Adjust fields as needed to force an empty XLOG starting at the next
! * available segment.
! */
! newXlogId = ControlFile.logId;
! newXlogSeg = ControlFile.logSeg;
! /* be sure we wrap around correctly at end of a logfile */
! NextLogSeg(newXlogId, newXlogSeg);
!
! ControlFile.checkPointCopy.redo.xlogid = newXlogId;
! ControlFile.checkPointCopy.redo.xrecoff =
! newXlogSeg * XLogSegSize + SizeOfXLogPHD;
! ControlFile.checkPointCopy.undo = ControlFile.checkPointCopy.redo;
! ControlFile.checkPointCopy.time = time(NULL);
!
! ControlFile.state = DB_SHUTDOWNED;
! ControlFile.time = time(NULL);
! ControlFile.logId = newXlogId;
! ControlFile.logSeg = newXlogSeg + 1;
! ControlFile.checkPoint = ControlFile.checkPointCopy.redo;
! ControlFile.prevCheckPoint.xlogid = 0;
! ControlFile.prevCheckPoint.xrecoff = 0;
! }
! else
! ControlFile.checkPointCopy.nextXid = set_xid;
!
/* Contents are protected with a CRC */
INIT_CRC64(ControlFile.crc);
COMP_CRC64(ControlFile.crc,
***************
*** 926,934 ****
static void
usage(void)
{
! fprintf(stderr, "Usage: pg_resetxlog [-f] [-n] PGDataDirectory\n\n"
! " -f\tforce update to be done\n"
! " -n\tno update, just show extracted pg_control values (for testing)\n");
exit(1);
}
--- 931,940 ----
static void
usage(void)
{
! fprintf(stderr, "Usage: pg_resetxlog [-f] [-n] [-x xid] PGDataDirectory\n"
! " -f\tforce update to be done\n"
! " -n\tno update, just show extracted pg_control values (for testing)\n"
! " -x XID\tset XID in pg_control\n");
exit(1);
}
***************
*** 939,944 ****
--- 945,951 ----
int argn;
bool force = false;
bool noupdate = false;
+ TransactionId set_xid = 0;
int fd;
char path[MAXPGPATH];
***************
*** 950,955 ****
--- 957,974 ----
force = true;
else if (strcmp(argv[argn], "-n") == 0)
noupdate = true;
+ else if (strcmp(argv[argn], "-x") == 0)
+ {
+ argn++;
+ if (argn == argc)
+ usage();
+ set_xid = strtoul(argv[argn], NULL, 0);
+ if (set_xid == 0)
+ {
+ fprintf(stderr, "XID can not be 0.");
+ exit(1);
+ }
+ }
else
usage();
}
***************
*** 993,998 ****
--- 1012,1031 ----
GuessControlValues();
/*
+ * Set XID in pg_control and exit
+ */
+ if (set_xid)
+ {
+ if (guessed)
+ {
+ printf("\npg_control appears corrupt. Can not update XID.\n");
+ exit(1);
+ }
+ RewriteControlFile(set_xid);
+ exit(0);
+ }
+
+ /*
* If we had to guess anything, and -f was not given, just print the
* guessed values and exit. Also print if -n is given.
*/
***************
*** 1018,1024 ****
/*
* Else, do the dirty deed.
*/
! RewriteControlFile();
KillExistingXLOG();
WriteEmptyXLOG();
--- 1051,1057 ----
/*
* Else, do the dirty deed.
*/
! RewriteControlFile(0);
KillExistingXLOG();
WriteEmptyXLOG();
Bruce Momjian <pgman@candle.pha.pa.us> writes:
Here is a patch I need to /contrib/pg_resetxlog to support a new "-x
XID" option to set the XID in pg_control.
I don't like this patch. It seems weird to add -x as an independent
function rather than just have pg_resetxlog do its normal thing and
allow -x to override the xid value. -x defined that way makes sense
in the context of pg_resetxlog's original mission (in particular, one
should be able to use it in the situation where the old pg_control is
unrecoverable). Also, there's no good reason for pg_upgrade not to
reset the xlog --- certainly we would not want the records therein to
be replayed against the pg_upgraded database!
There is a more serious problem, also. Pages transferred over from the
old database will contain LSN values pointing into the old xlog. If
these are past the end of the new database's xlog (very probable) then
you have a strong risk of "XLogFlush: request past end of xlog" errors,
which per Vadim's insistence we treat as a system-wide fatal condition.
Probably the cleanest way to deal with that is to tweak pg_resetxlog
further to have an optional switch with a minimum xlog position.
It already knows how to set up its cleared xlog with a position >=
end of the removed log, so you could have an additional option switch
that forces the new position to be >= switch value. To issue the
switch, pg_upgrade would have to look at the old xlog files to determine
the endpoint of the old xlog. Seems messy but not impossible.
regards, tom lane
Tom Lane wrote:
Bruce Momjian <pgman@candle.pha.pa.us> writes:
Here is a patch I need to /contrib/pg_resetxlog to support a new "-x
XID" option to set the XID in pg_control.I don't like this patch. It seems weird to add -x as an independent
function rather than just have pg_resetxlog do its normal thing and
allow -x to override the xid value. -x defined that way makes sense
in the context of pg_resetxlog's original mission (in particular, one
should be able to use it in the situation where the old pg_control is
unrecoverable). Also, there's no good reason for pg_upgrade not to
reset the xlog --- certainly we would not want the records therein to
be replayed against the pg_upgraded database!
OK, if we want to reset WAL at the same time, which does make sense as
you say, here is the patch. This is even easier for me. It just
optionally sets the XID as part of the normal operation. (I am going to
commit this patch because it is better for you and smaller than the one
I just committed from last night.)
There is a more serious problem, also. Pages transferred over from the
old database will contain LSN values pointing into the old xlog. If
these are past the end of the new database's xlog (very probable) then
you have a strong risk of "XLogFlush: request past end of xlog" errors,
which per Vadim's insistence we treat as a system-wide fatal condition.Probably the cleanest way to deal with that is to tweak pg_resetxlog
further to have an optional switch with a minimum xlog position.
It already knows how to set up its cleared xlog with a position >=
end of the removed log, so you could have an additional option switch
that forces the new position to be >= switch value. To issue the
switch, pg_upgrade would have to look at the old xlog files to determine
the endpoint of the old xlog. Seems messy but not impossible.
Wow, that sounds hard. Can you give me some hints which pg_control
field that is in?
--
Bruce Momjian | http://candle.pha.pa.us
pgman@candle.pha.pa.us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026Attachments:
/bjm/difftext/plainDownload
Index: pg_resetxlog.c
===================================================================
RCS file: /cvsroot/pgsql/contrib/pg_resetxlog/pg_resetxlog.c,v
retrieving revision 1.10
diff -c -r1.10 pg_resetxlog.c
*** pg_resetxlog.c 2001/11/05 17:46:23 1.10
--- pg_resetxlog.c 2002/01/10 18:01:51
***************
*** 23,29 ****
* Portions Copyright (c) 1996-2001, PostgreSQL Global Development Group
* Portions Copyright (c) 1994, Regents of the University of California
*
! * $Header: /cvsroot/pgsql/contrib/pg_resetxlog/pg_resetxlog.c,v 1.10 2001/11/05 17:46:23 momjian Exp $
*
*-------------------------------------------------------------------------
*/
--- 23,29 ----
* Portions Copyright (c) 1996-2001, PostgreSQL Global Development Group
* Portions Copyright (c) 1994, Regents of the University of California
*
! * $Header: /cvsroot/pgsql/contrib/pg_resetxlog/pg_resetxlog.c,v 1.11 2002/01/10 17:51:52 momjian Exp $
*
*-------------------------------------------------------------------------
*/
***************
*** 709,715 ****
* Write out the new pg_control file.
*/
static void
! RewriteControlFile(void)
{
int fd;
char buffer[BLCKSZ]; /* need not be aligned */
--- 709,715 ----
* Write out the new pg_control file.
*/
static void
! RewriteControlFile(TransactionId set_xid)
{
int fd;
char buffer[BLCKSZ]; /* need not be aligned */
***************
*** 737,742 ****
--- 737,745 ----
ControlFile.prevCheckPoint.xlogid = 0;
ControlFile.prevCheckPoint.xrecoff = 0;
+ if (set_xid != 0)
+ ControlFile.checkPointCopy.nextXid = set_xid;
+
/* Contents are protected with a CRC */
INIT_CRC64(ControlFile.crc);
COMP_CRC64(ControlFile.crc,
***************
*** 926,934 ****
static void
usage(void)
{
! fprintf(stderr, "Usage: pg_resetxlog [-f] [-n] PGDataDirectory\n\n"
! " -f\tforce update to be done\n"
! " -n\tno update, just show extracted pg_control values (for testing)\n");
exit(1);
}
--- 929,938 ----
static void
usage(void)
{
! fprintf(stderr, "Usage: pg_resetxlog [-f] [-n] [-x xid] PGDataDirectory\n"
! " -f\tforce update to be done\n"
! " -n\tno update, just show extracted pg_control values (for testing)\n"
! " -x XID\tset XID in pg_control\n");
exit(1);
}
***************
*** 939,944 ****
--- 943,949 ----
int argn;
bool force = false;
bool noupdate = false;
+ TransactionId set_xid = 0;
int fd;
char path[MAXPGPATH];
***************
*** 950,955 ****
--- 955,972 ----
force = true;
else if (strcmp(argv[argn], "-n") == 0)
noupdate = true;
+ else if (strcmp(argv[argn], "-x") == 0)
+ {
+ argn++;
+ if (argn == argc)
+ usage();
+ set_xid = strtoul(argv[argn], NULL, 0);
+ if (set_xid == 0)
+ {
+ fprintf(stderr, "XID can not be 0.");
+ exit(1);
+ }
+ }
else
usage();
}
***************
*** 1018,1024 ****
/*
* Else, do the dirty deed.
*/
! RewriteControlFile();
KillExistingXLOG();
WriteEmptyXLOG();
--- 1035,1041 ----
/*
* Else, do the dirty deed.
*/
! RewriteControlFile(set_xid);
KillExistingXLOG();
WriteEmptyXLOG();
Probably the cleanest way to deal with that is to tweak pg_resetxlog
further to have an optional switch with a minimum xlog position.
It already knows how to set up its cleared xlog with a position >=
end of the removed log, so you could have an additional option switch
that forces the new position to be >= switch value. To issue the
switch, pg_upgrade would have to look at the old xlog files to determine
the endpoint of the old xlog. Seems messy but not impossible.
Also, how do I find the current xlog segment position. Is it one of
these fields shown by pg_controldata?
pg_controldata
pg_control version number: 71
Catalog version number: 200110251
Database state: SHUTDOWNED
pg_control last modified: 01/10/02 13:00:04
Current log file id: 0
Next log file segment: 7
Latest checkpoint location: 0/6000010
Prior checkpoint location: 0/0
Latest checkpoint's REDO location: 0/6000010
Latest checkpoint's UNDO location: 0/6000010
Latest checkpoint's StartUpID: 8
Latest checkpoint's NextXID: 105
Latest checkpoint's NextOID: 16557
Time of latest checkpoint: 01/10/02 13:00:04
Database block size: 8192
Blocks per segment of large relation: 131072
LC_COLLATE: C
LC_CTYPE: C
--
Bruce Momjian | http://candle.pha.pa.us
pgman@candle.pha.pa.us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026Bruce Momjian <pgman@candle.pha.pa.us> writes:
Also, how do I find the current xlog segment position. Is it one of
these fields shown by pg_controldata?
"latest checkpoint location" should do.
BTW, if your script is relying on pg_resetxlog to be available, best to
ensure that it's there before you do anything irreversible ...
regards, tom lane
Tom Lane wrote:
Bruce Momjian <pgman@candle.pha.pa.us> writes:
Also, how do I find the current xlog segment position. Is it one of
these fields shown by pg_controldata?"latest checkpoint location" should do.
OK. I will add a -l flag to specify that location.
BTW, if your script is relying on pg_resetxlog to be available, best to
ensure that it's there before you do anything irreversible ...
Oh yes, I will make sure it is available _and_ has the flags from the
new version.
--
Bruce Momjian | http://candle.pha.pa.us
pgman@candle.pha.pa.us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026Tom Lane wrote:
Bruce Momjian <pgman@candle.pha.pa.us> writes:
Also, how do I find the current xlog segment position. Is it one of
these fields shown by pg_controldata?"latest checkpoint location" should do.
BTW, if your script is relying on pg_resetxlog to be available, best to
ensure that it's there before you do anything irreversible ...
Do we want to remove the 7.1beta WAL format code from
/contrib/pg_resetxlog? Remember, this utility was originally written to
allow for a WAL format change during 7.1beta testing. Seems like dead
code now.
--
Bruce Momjian | http://candle.pha.pa.us
pgman@candle.pha.pa.us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026Bruce Momjian <pgman@candle.pha.pa.us> writes:
Do we want to remove the 7.1beta WAL format code from
/contrib/pg_resetxlog?
Sure, I don't think there's any strong need for it anymore. Anyone who
did need that version could get it out of the 7.1 release, anyway.
regards, tom lane
Tom Lane wrote:
Bruce Momjian <pgman@candle.pha.pa.us> writes:
Do we want to remove the 7.1beta WAL format code from
/contrib/pg_resetxlog?Sure, I don't think there's any strong need for it anymore. Anyone who
did need that version could get it out of the 7.1 release, anyway.
OK, I will do that as a separate patch later.
--
Bruce Momjian | http://candle.pha.pa.us
pgman@candle.pha.pa.us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026Tom Lane wrote:
Bruce Momjian <pgman@candle.pha.pa.us> writes:
Also, how do I find the current xlog segment position. Is it one of
these fields shown by pg_controldata?"latest checkpoint location" should do.
BTW, if your script is relying on pg_resetxlog to be available, best to
ensure that it's there before you do anything irreversible ...
OK, here is code to set the checkpoint log id and offset using a new -l
flag. It also now displays the checkpoint location with -n.
--
Bruce Momjian | http://candle.pha.pa.us
pgman@candle.pha.pa.us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026Attachments:
/bjm/difftext/plainDownload
Index: contrib/pg_resetxlog/README.pg_resetxlog
===================================================================
RCS file: /cvsroot/pgsql/contrib/pg_resetxlog/README.pg_resetxlog,v
retrieving revision 1.2
diff -c -r1.2 README.pg_resetxlog
*** contrib/pg_resetxlog/README.pg_resetxlog 2002/01/10 17:51:52 1.2
--- contrib/pg_resetxlog/README.pg_resetxlog 2002/01/10 19:49:23
***************
*** 22,28 ****
the database software.
A tertiary purpose it its use by pg_upgrade to set the next transaction
! id in pg_control.
To run the program, make sure your postmaster is not running, then
(as the Postgres admin user) do
--- 22,28 ----
the database software.
A tertiary purpose it its use by pg_upgrade to set the next transaction
! id and checkpoint location in pg_control.
To run the program, make sure your postmaster is not running, then
(as the Postgres admin user) do
Index: contrib/pg_resetxlog/pg_resetxlog.c
===================================================================
RCS file: /cvsroot/pgsql/contrib/pg_resetxlog/pg_resetxlog.c,v
retrieving revision 1.12
diff -c -r1.12 pg_resetxlog.c
*** contrib/pg_resetxlog/pg_resetxlog.c 2002/01/10 18:08:29 1.12
--- contrib/pg_resetxlog/pg_resetxlog.c 2002/01/10 19:49:24
***************
*** 683,688 ****
--- 683,689 ----
"Catalog version number: %u\n"
"Current log file id: %u\n"
"Next log file segment: %u\n"
+ "Latest checkpoint location: %X/%X\n"
"Latest checkpoint's StartUpID: %u\n"
"Latest checkpoint's NextXID: %u\n"
"Latest checkpoint's NextOID: %u\n"
***************
*** 695,700 ****
--- 696,703 ----
ControlFile.catalog_version_no,
ControlFile.logId,
ControlFile.logSeg,
+ ControlFile.checkPoint.xlogid,
+ ControlFile.checkPoint.xrecoff,
ControlFile.checkPointCopy.ThisStartUpID,
ControlFile.checkPointCopy.nextXid,
ControlFile.checkPointCopy.nextOid,
***************
*** 709,715 ****
* Write out the new pg_control file.
*/
static void
! RewriteControlFile(TransactionId set_xid)
{
int fd;
char buffer[BLCKSZ]; /* need not be aligned */
--- 712,718 ----
* Write out the new pg_control file.
*/
static void
! RewriteControlFile(TransactionId set_xid, XLogRecPtr set_checkpoint)
{
int fd;
char buffer[BLCKSZ]; /* need not be aligned */
***************
*** 733,745 ****
ControlFile.time = time(NULL);
ControlFile.logId = newXlogId;
ControlFile.logSeg = newXlogSeg + 1;
- ControlFile.checkPoint = ControlFile.checkPointCopy.redo;
ControlFile.prevCheckPoint.xlogid = 0;
ControlFile.prevCheckPoint.xrecoff = 0;
if (set_xid != 0)
ControlFile.checkPointCopy.nextXid = set_xid;
!
/* Contents are protected with a CRC */
INIT_CRC64(ControlFile.crc);
COMP_CRC64(ControlFile.crc,
--- 736,753 ----
ControlFile.time = time(NULL);
ControlFile.logId = newXlogId;
ControlFile.logSeg = newXlogSeg + 1;
ControlFile.prevCheckPoint.xlogid = 0;
ControlFile.prevCheckPoint.xrecoff = 0;
if (set_xid != 0)
ControlFile.checkPointCopy.nextXid = set_xid;
!
! if (set_checkpoint.xlogid == 0 &&
! set_checkpoint.xrecoff == 0)
! ControlFile.checkPoint = ControlFile.checkPointCopy.redo;
! else
! ControlFile.checkPoint = set_checkpoint;
!
/* Contents are protected with a CRC */
INIT_CRC64(ControlFile.crc);
COMP_CRC64(ControlFile.crc,
***************
*** 929,938 ****
static void
usage(void)
{
! fprintf(stderr, "Usage: pg_resetxlog [-f] [-n] [-x xid] PGDataDirectory\n"
! " -f\tforce update to be done\n"
! " -n\tno update, just show extracted pg_control values (for testing)\n"
! " -x XID\tset XID in pg_control\n");
exit(1);
}
--- 937,947 ----
static void
usage(void)
{
! fprintf(stderr, "Usage: pg_resetxlog [-f] [-n] [-x xid] [ -l log_id offset ] PGDataDirectory\n"
! " -f\t force update to be done\n"
! " -n\t no update, just show extracted pg_control values (for testing)\n"
! " -x XID set XID in pg_control\n"
! " -l log_id offset set checkpoint location in pg_control\n");
exit(1);
}
***************
*** 944,949 ****
--- 953,959 ----
bool force = false;
bool noupdate = false;
TransactionId set_xid = 0;
+ XLogRecPtr set_checkpoint = {0,0};
int fd;
char path[MAXPGPATH];
***************
*** 967,972 ****
--- 977,999 ----
exit(1);
}
}
+ else if (strcmp(argv[argn], "-l") == 0)
+ {
+ argn++;
+ if (argn == argc)
+ usage();
+ set_checkpoint.xlogid = strtoul(argv[argn], NULL, 0);
+ argn++;
+ if (argn == argc)
+ usage();
+ set_checkpoint.xrecoff = strtoul(argv[argn], NULL, 0);
+ if (set_checkpoint.xlogid == 0 &&
+ set_checkpoint.xrecoff == 0)
+ {
+ fprintf(stderr, "Checkpoint can not be '0 0'.");
+ exit(1);
+ }
+ }
else
usage();
}
***************
*** 1035,1041 ****
/*
* Else, do the dirty deed.
*/
! RewriteControlFile(set_xid);
KillExistingXLOG();
WriteEmptyXLOG();
--- 1062,1068 ----
/*
* Else, do the dirty deed.
*/
! RewriteControlFile(set_xid, set_checkpoint);
KillExistingXLOG();
WriteEmptyXLOG();
Tom Lane wrote:
Bruce Momjian <pgman@candle.pha.pa.us> writes:
Do we want to remove the 7.1beta WAL format code from
/contrib/pg_resetxlog?Sure, I don't think there's any strong need for it anymore. Anyone who
did need that version could get it out of the 7.1 release, anyway.
The following patch removes the V0 WAL handling that was there only for
7.1beta.
--
Bruce Momjian | http://candle.pha.pa.us
pgman@candle.pha.pa.us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026Attachments:
/bjm/difftext/plainDownload
Index: contrib/pg_resetxlog/pg_resetxlog.c
===================================================================
RCS file: /cvsroot/pgsql/contrib/pg_resetxlog/pg_resetxlog.c,v
retrieving revision 1.13
diff -c -r1.13 pg_resetxlog.c
*** contrib/pg_resetxlog/pg_resetxlog.c 2002/01/10 20:09:06 1.13
--- contrib/pg_resetxlog/pg_resetxlog.c 2002/01/10 22:37:33
***************
*** 59,106 ****
(logSeg)++; \
} while (0)
- /*
- * Compute ID and segment from an XLogRecPtr.
- *
- * For XLByteToSeg, do the computation at face value. For XLByteToPrevSeg,
- * a boundary byte is taken to be in the previous segment. This is suitable
- * for deciding which segment to write given a pointer to a record end,
- * for example.
- */
- #define XLByteToSeg(xlrp, logId, logSeg) \
- ( logId = (xlrp).xlogid, \
- logSeg = (xlrp).xrecoff / XLogSegSize \
- )
- #define XLByteToPrevSeg(xlrp, logId, logSeg) \
- ( logId = (xlrp).xlogid, \
- logSeg = ((xlrp).xrecoff - 1) / XLogSegSize \
- )
-
- /*
- * Is an XLogRecPtr within a particular XLOG segment?
- *
- * For XLByteInSeg, do the computation at face value. For XLByteInPrevSeg,
- * a boundary byte is taken to be in the previous segment.
- */
- #define XLByteInSeg(xlrp, logId, logSeg) \
- ((xlrp).xlogid == (logId) && \
- (xlrp).xrecoff / XLogSegSize == (logSeg))
-
- #define XLByteInPrevSeg(xlrp, logId, logSeg) \
- ((xlrp).xlogid == (logId) && \
- ((xlrp).xrecoff - 1) / XLogSegSize == (logSeg))
-
-
#define XLogFileName(path, log, seg) \
snprintf(path, MAXPGPATH, "%s/%08X%08X", \
XLogDir, log, seg)
- /*
- * _INTL_MAXLOGRECSZ: max space needed for a record including header and
- * any backup-block data.
- */
- #define _INTL_MAXLOGRECSZ (SizeOfXLogRecord + MAXLOGRECSZ + \
- XLR_MAX_BKP_BLOCKS * (sizeof(BkpBlock) + BLCKSZ))
/******************** end of stuff copied from xlog.c ********************/
--- 59,68 ----
***************
*** 115,136 ****
static bool guessed = false; /* T if we had to guess at any values */
- static bool CheckControlVersion0(char *buffer, int len);
-
-
- static int
- XLogFileOpen(uint32 log, uint32 seg)
- {
- char path[MAXPGPATH];
- int fd;
-
- XLogFileName(path, log, seg);
-
- fd = open(path, O_RDWR | PG_BINARY, S_IRUSR | S_IWUSR);
- return (fd);
- }
-
-
/*
* Try to read the existing pg_control file.
*
--- 77,82 ----
***************
*** 174,180 ****
if (len >= sizeof(ControlFileData) &&
((ControlFileData *) buffer)->pg_control_version == PG_CONTROL_VERSION)
{
! /* Seems to be current version --- check the CRC. */
INIT_CRC64(crc);
COMP_CRC64(crc,
buffer + sizeof(crc64),
--- 120,126 ----
if (len >= sizeof(ControlFileData) &&
((ControlFileData *) buffer)->pg_control_version == PG_CONTROL_VERSION)
{
! /* Check the CRC. */
INIT_CRC64(crc);
COMP_CRC64(crc,
buffer + sizeof(crc64),
***************
*** 188,612 ****
return true;
}
! fprintf(stderr, "pg_control exists but has invalid CRC; proceed with caution.\n");
/* We will use the data anyway, but treat it as guessed. */
memcpy(&ControlFile, buffer, sizeof(ControlFile));
guessed = true;
return true;
}
- /*
- * Maybe it's a 7.1beta pg_control.
- */
- if (CheckControlVersion0(buffer, len))
- return true;
-
/* Looks like it's a mess. */
fprintf(stderr, "pg_control exists but is broken or unknown version; ignoring it.\n");
return false;
}
- /******************* routines for old XLOG format *******************/
-
-
- /*
- * This format was in use in 7.1 beta releases through 7.1beta5. The
- * pg_control layout was different, and so were the XLOG page headers.
- * The XLOG record header format was physically the same as 7.1 release,
- * but interpretation of the xl_len field was not.
- */
-
- typedef struct crc64V0
- {
- uint32 crc1;
- uint32 crc2;
- } crc64V0;
-
- static uint32 crc_tableV0[] = {
- 0x00000000, 0x77073096, 0xee0e612c, 0x990951ba, 0x076dc419, 0x706af48f,
- 0xe963a535, 0x9e6495a3, 0x0edb8832, 0x79dcb8a4, 0xe0d5e91e, 0x97d2d988,
- 0x09b64c2b, 0x7eb17cbd, 0xe7b82d07, 0x90bf1d91, 0x1db71064, 0x6ab020f2,
- 0xf3b97148, 0x84be41de, 0x1adad47d, 0x6ddde4eb, 0xf4d4b551, 0x83d385c7,
- 0x136c9856, 0x646ba8c0, 0xfd62f97a, 0x8a65c9ec, 0x14015c4f, 0x63066cd9,
- 0xfa0f3d63, 0x8d080df5, 0x3b6e20c8, 0x4c69105e, 0xd56041e4, 0xa2677172,
- 0x3c03e4d1, 0x4b04d447, 0xd20d85fd, 0xa50ab56b, 0x35b5a8fa, 0x42b2986c,
- 0xdbbbc9d6, 0xacbcf940, 0x32d86ce3, 0x45df5c75, 0xdcd60dcf, 0xabd13d59,
- 0x26d930ac, 0x51de003a, 0xc8d75180, 0xbfd06116, 0x21b4f4b5, 0x56b3c423,
- 0xcfba9599, 0xb8bda50f, 0x2802b89e, 0x5f058808, 0xc60cd9b2, 0xb10be924,
- 0x2f6f7c87, 0x58684c11, 0xc1611dab, 0xb6662d3d, 0x76dc4190, 0x01db7106,
- 0x98d220bc, 0xefd5102a, 0x71b18589, 0x06b6b51f, 0x9fbfe4a5, 0xe8b8d433,
- 0x7807c9a2, 0x0f00f934, 0x9609a88e, 0xe10e9818, 0x7f6a0dbb, 0x086d3d2d,
- 0x91646c97, 0xe6635c01, 0x6b6b51f4, 0x1c6c6162, 0x856530d8, 0xf262004e,
- 0x6c0695ed, 0x1b01a57b, 0x8208f4c1, 0xf50fc457, 0x65b0d9c6, 0x12b7e950,
- 0x8bbeb8ea, 0xfcb9887c, 0x62dd1ddf, 0x15da2d49, 0x8cd37cf3, 0xfbd44c65,
- 0x4db26158, 0x3ab551ce, 0xa3bc0074, 0xd4bb30e2, 0x4adfa541, 0x3dd895d7,
- 0xa4d1c46d, 0xd3d6f4fb, 0x4369e96a, 0x346ed9fc, 0xad678846, 0xda60b8d0,
- 0x44042d73, 0x33031de5, 0xaa0a4c5f, 0xdd0d7cc9, 0x5005713c, 0x270241aa,
- 0xbe0b1010, 0xc90c2086, 0x5768b525, 0x206f85b3, 0xb966d409, 0xce61e49f,
- 0x5edef90e, 0x29d9c998, 0xb0d09822, 0xc7d7a8b4, 0x59b33d17, 0x2eb40d81,
- 0xb7bd5c3b, 0xc0ba6cad, 0xedb88320, 0x9abfb3b6, 0x03b6e20c, 0x74b1d29a,
- 0xead54739, 0x9dd277af, 0x04db2615, 0x73dc1683, 0xe3630b12, 0x94643b84,
- 0x0d6d6a3e, 0x7a6a5aa8, 0xe40ecf0b, 0x9309ff9d, 0x0a00ae27, 0x7d079eb1,
- 0xf00f9344, 0x8708a3d2, 0x1e01f268, 0x6906c2fe, 0xf762575d, 0x806567cb,
- 0x196c3671, 0x6e6b06e7, 0xfed41b76, 0x89d32be0, 0x10da7a5a, 0x67dd4acc,
- 0xf9b9df6f, 0x8ebeeff9, 0x17b7be43, 0x60b08ed5, 0xd6d6a3e8, 0xa1d1937e,
- 0x38d8c2c4, 0x4fdff252, 0xd1bb67f1, 0xa6bc5767, 0x3fb506dd, 0x48b2364b,
- 0xd80d2bda, 0xaf0a1b4c, 0x36034af6, 0x41047a60, 0xdf60efc3, 0xa867df55,
- 0x316e8eef, 0x4669be79, 0xcb61b38c, 0xbc66831a, 0x256fd2a0, 0x5268e236,
- 0xcc0c7795, 0xbb0b4703, 0x220216b9, 0x5505262f, 0xc5ba3bbe, 0xb2bd0b28,
- 0x2bb45a92, 0x5cb36a04, 0xc2d7ffa7, 0xb5d0cf31, 0x2cd99e8b, 0x5bdeae1d,
- 0x9b64c2b0, 0xec63f226, 0x756aa39c, 0x026d930a, 0x9c0906a9, 0xeb0e363f,
- 0x72076785, 0x05005713, 0x95bf4a82, 0xe2b87a14, 0x7bb12bae, 0x0cb61b38,
- 0x92d28e9b, 0xe5d5be0d, 0x7cdcefb7, 0x0bdbdf21, 0x86d3d2d4, 0xf1d4e242,
- 0x68ddb3f8, 0x1fda836e, 0x81be16cd, 0xf6b9265b, 0x6fb077e1, 0x18b74777,
- 0x88085ae6, 0xff0f6a70, 0x66063bca, 0x11010b5c, 0x8f659eff, 0xf862ae69,
- 0x616bffd3, 0x166ccf45, 0xa00ae278, 0xd70dd2ee, 0x4e048354, 0x3903b3c2,
- 0xa7672661, 0xd06016f7, 0x4969474d, 0x3e6e77db, 0xaed16a4a, 0xd9d65adc,
- 0x40df0b66, 0x37d83bf0, 0xa9bcae53, 0xdebb9ec5, 0x47b2cf7f, 0x30b5ffe9,
- 0xbdbdf21c, 0xcabac28a, 0x53b39330, 0x24b4a3a6, 0xbad03605, 0xcdd70693,
- 0x54de5729, 0x23d967bf, 0xb3667a2e, 0xc4614ab8, 0x5d681b02, 0x2a6f2b94,
- 0xb40bbe37, 0xc30c8ea1, 0x5a05df1b, 0x2d02ef8d
- };
-
- #define INIT_CRC64V0(crc) ((crc).crc1 = 0xffffffff, (crc).crc2 = 0xffffffff)
- #define FIN_CRC64V0(crc) ((crc).crc1 ^= 0xffffffff, (crc).crc2 ^= 0xffffffff)
- #define COMP_CRC64V0(crc, data, len) \
- do {\
- uint32 __c1 = (crc).crc1;\
- uint32 __c2 = (crc).crc2;\
- char *__data = (char *) (data);\
- uint32 __len = (len);\
- \
- while (__len >= 2)\
- {\
- __c1 = crc_tableV0[(__c1 ^ *__data++) & 0xff] ^ (__c1 >> 8);\
- __c2 = crc_tableV0[(__c2 ^ *__data++) & 0xff] ^ (__c2 >> 8);\
- __len -= 2;\
- }\
- if (__len > 0)\
- __c1 = crc_tableV0[(__c1 ^ *__data++) & 0xff] ^ (__c1 >> 8);\
- (crc).crc1 = __c1;\
- (crc).crc2 = __c2;\
- } while (0)
-
- #define EQ_CRC64V0(c1,c2) ((c1).crc1 == (c2).crc1 && (c1).crc2 == (c2).crc2)
-
-
- #define LOCALE_NAME_BUFLEN_V0 128
-
- typedef struct ControlFileDataV0
- {
- crc64V0 crc;
- uint32 logId; /* current log file id */
- uint32 logSeg; /* current log file segment (1-based) */
- XLogRecPtr checkPoint; /* last check point record ptr */
- time_t time; /* time stamp of last modification */
- DBState state; /* see enum above */
- uint32 blcksz; /* block size for this DB */
- uint32 relseg_size; /* blocks per segment of large relation */
- uint32 catalog_version_no; /* internal version number */
- char lc_collate[LOCALE_NAME_BUFLEN_V0];
- char lc_ctype[LOCALE_NAME_BUFLEN_V0];
- char archdir[MAXPGPATH]; /* where to move offline log files */
- } ControlFileDataV0;
-
- typedef struct CheckPointV0
- {
- XLogRecPtr redo; /* next RecPtr available when we */
- /* began to create CheckPoint */
- /* (i.e. REDO start point) */
- XLogRecPtr undo; /* first record of oldest in-progress */
- /* transaction when we started */
- /* (i.e. UNDO end point) */
- StartUpID ThisStartUpID;
- TransactionId nextXid;
- Oid nextOid;
- bool Shutdown;
- } CheckPointV0;
-
- typedef struct XLogRecordV0
- {
- crc64V0 xl_crc;
- XLogRecPtr xl_prev; /* ptr to previous record in log */
- XLogRecPtr xl_xact_prev; /* ptr to previous record of this xact */
- TransactionId xl_xid; /* xact id */
- uint16 xl_len; /* total len of record *data* */
- uint8 xl_info;
- RmgrId xl_rmid; /* resource manager inserted this record */
- } XLogRecordV0;
-
- #define SizeOfXLogRecordV0 DOUBLEALIGN(sizeof(XLogRecordV0))
-
- typedef struct XLogContRecordV0
- {
- uint16 xl_len; /* len of data left */
- } XLogContRecordV0;
-
- #define SizeOfXLogContRecordV0 DOUBLEALIGN(sizeof(XLogContRecordV0))
-
- #define XLOG_PAGE_MAGIC_V0 0x17345168
-
- typedef struct XLogPageHeaderDataV0
- {
- uint32 xlp_magic;
- uint16 xlp_info;
- } XLogPageHeaderDataV0;
-
- #define SizeOfXLogPHDV0 DOUBLEALIGN(sizeof(XLogPageHeaderDataV0))
-
- typedef XLogPageHeaderDataV0 *XLogPageHeaderV0;
-
-
- static bool RecordIsValidV0(XLogRecordV0 * record);
- static XLogRecordV0 *ReadRecordV0(XLogRecPtr *RecPtr, char *buffer);
- static bool ValidXLOGHeaderV0(XLogPageHeaderV0 hdr);
-
-
- /*
- * Try to interpret pg_control contents as "version 0" format.
- */
- static bool
- CheckControlVersion0(char *buffer, int len)
- {
- crc64V0 crc;
- ControlFileDataV0 *oldfile;
- XLogRecordV0 *record;
- CheckPointV0 *oldchkpt;
-
- if (len < sizeof(ControlFileDataV0))
- return false;
- /* Check CRC the version-0 way. */
- INIT_CRC64V0(crc);
- COMP_CRC64V0(crc,
- buffer + sizeof(crc64V0),
- sizeof(ControlFileDataV0) - sizeof(crc64V0));
- FIN_CRC64V0(crc);
-
- if (!EQ_CRC64V0(crc, ((ControlFileDataV0 *) buffer)->crc))
- return false;
-
- /* Valid data, convert useful fields to new-style pg_control format */
- oldfile = (ControlFileDataV0 *) buffer;
-
- memset(&ControlFile, 0, sizeof(ControlFile));
-
- ControlFile.pg_control_version = PG_CONTROL_VERSION;
- ControlFile.catalog_version_no = oldfile->catalog_version_no;
-
- ControlFile.state = oldfile->state;
- ControlFile.logId = oldfile->logId;
- ControlFile.logSeg = oldfile->logSeg;
-
- ControlFile.blcksz = oldfile->blcksz;
- ControlFile.relseg_size = oldfile->relseg_size;
- strcpy(ControlFile.lc_collate, oldfile->lc_collate);
- strcpy(ControlFile.lc_ctype, oldfile->lc_ctype);
-
- /*
- * Since this format did not include a copy of the latest checkpoint
- * record, we have to go rooting in the old XLOG to get that.
- */
- record = ReadRecordV0(&oldfile->checkPoint,
- (char *) malloc(_INTL_MAXLOGRECSZ));
- if (record == NULL)
- {
- /*
- * We have to guess at the checkpoint contents.
- */
- guessed = true;
- ControlFile.checkPointCopy.ThisStartUpID = 0;
- ControlFile.checkPointCopy.nextXid = (TransactionId) 514; /* XXX */
- ControlFile.checkPointCopy.nextOid = BootstrapObjectIdData;
- return true;
- }
- oldchkpt = (CheckPointV0 *) XLogRecGetData(record);
-
- ControlFile.checkPointCopy.ThisStartUpID = oldchkpt->ThisStartUpID;
- ControlFile.checkPointCopy.nextXid = oldchkpt->nextXid;
- ControlFile.checkPointCopy.nextOid = oldchkpt->nextOid;
-
- return true;
- }
-
- /*
- * CRC-check an XLOG V0 record. We do not believe the contents of an XLOG
- * record (other than to the minimal extent of computing the amount of
- * data to read in) until we've checked the CRCs.
- *
- * We assume all of the record has been read into memory at *record.
- */
- static bool
- RecordIsValidV0(XLogRecordV0 * record)
- {
- crc64V0 crc;
- uint32 len = record->xl_len;
-
- /*
- * NB: this code is not right for V0 records containing backup blocks,
- * but for now it's only going to be applied to checkpoint records, so
- * I'm not going to worry about it...
- */
- INIT_CRC64V0(crc);
- COMP_CRC64V0(crc, XLogRecGetData(record), len);
- COMP_CRC64V0(crc, (char *) record + sizeof(crc64V0),
- SizeOfXLogRecordV0 - sizeof(crc64V0));
- FIN_CRC64V0(crc);
-
- if (!EQ_CRC64V0(record->xl_crc, crc))
- return false;
-
- return (true);
- }
-
- /*
- * Attempt to read an XLOG V0 record at recptr.
- *
- * If no valid record is available, returns NULL.
- *
- * buffer is a workspace at least _INTL_MAXLOGRECSZ bytes long. It is needed
- * to reassemble a record that crosses block boundaries. Note that on
- * successful return, the returned record pointer always points at buffer.
- */
- static XLogRecordV0 *
- ReadRecordV0(XLogRecPtr *RecPtr, char *buffer)
- {
- static int readFile = -1;
- static uint32 readId = 0;
- static uint32 readSeg = 0;
- static uint32 readOff = 0;
- static char *readBuf = NULL;
-
- XLogRecordV0 *record;
- uint32 len,
- total_len;
- uint32 targetPageOff;
-
- if (readBuf == NULL)
- readBuf = (char *) malloc(BLCKSZ);
-
- XLByteToSeg(*RecPtr, readId, readSeg);
- if (readFile < 0)
- {
- readFile = XLogFileOpen(readId, readSeg);
- if (readFile < 0)
- goto next_record_is_invalid;
- readOff = (uint32) (-1); /* force read to occur below */
- }
-
- targetPageOff = ((RecPtr->xrecoff % XLogSegSize) / BLCKSZ) * BLCKSZ;
- if (readOff != targetPageOff)
- {
- readOff = targetPageOff;
- if (lseek(readFile, (off_t) readOff, SEEK_SET) < 0)
- goto next_record_is_invalid;
- if (read(readFile, readBuf, BLCKSZ) != BLCKSZ)
- goto next_record_is_invalid;
- if (!ValidXLOGHeaderV0((XLogPageHeaderV0) readBuf))
- goto next_record_is_invalid;
- }
- if ((((XLogPageHeaderV0) readBuf)->xlp_info & XLP_FIRST_IS_CONTRECORD) &&
- RecPtr->xrecoff % BLCKSZ == SizeOfXLogPHDV0)
- goto next_record_is_invalid;
- record = (XLogRecordV0 *) ((char *) readBuf + RecPtr->xrecoff % BLCKSZ);
-
- if (record->xl_len == 0)
- goto next_record_is_invalid;
-
- /*
- * Compute total length of record including any appended backup
- * blocks.
- */
- total_len = SizeOfXLogRecordV0 + record->xl_len;
-
- /*
- * Make sure it will fit in buffer (currently, it is mechanically
- * impossible for this test to fail, but it seems like a good idea
- * anyway).
- */
- if (total_len > _INTL_MAXLOGRECSZ)
- goto next_record_is_invalid;
- len = BLCKSZ - RecPtr->xrecoff % BLCKSZ;
- if (total_len > len)
- {
- /* Need to reassemble record */
- XLogContRecordV0 *contrecord;
- uint32 gotlen = len;
-
- memcpy(buffer, record, len);
- record = (XLogRecordV0 *) buffer;
- buffer += len;
- for (;;)
- {
- readOff += BLCKSZ;
- if (readOff >= XLogSegSize)
- {
- close(readFile);
- readFile = -1;
- NextLogSeg(readId, readSeg);
- readFile = XLogFileOpen(readId, readSeg);
- if (readFile < 0)
- goto next_record_is_invalid;
- readOff = 0;
- }
- if (read(readFile, readBuf, BLCKSZ) != BLCKSZ)
- goto next_record_is_invalid;
- if (!ValidXLOGHeaderV0((XLogPageHeaderV0) readBuf))
- goto next_record_is_invalid;
- if (!(((XLogPageHeaderV0) readBuf)->xlp_info & XLP_FIRST_IS_CONTRECORD))
- goto next_record_is_invalid;
- contrecord = (XLogContRecordV0 *) ((char *) readBuf + SizeOfXLogPHDV0);
- if (contrecord->xl_len == 0 ||
- total_len != (contrecord->xl_len + gotlen))
- goto next_record_is_invalid;
- len = BLCKSZ - SizeOfXLogPHDV0 - SizeOfXLogContRecordV0;
- if (contrecord->xl_len > len)
- {
- memcpy(buffer, (char *) contrecord + SizeOfXLogContRecordV0, len);
- gotlen += len;
- buffer += len;
- continue;
- }
- memcpy(buffer, (char *) contrecord + SizeOfXLogContRecordV0,
- contrecord->xl_len);
- break;
- }
- if (!RecordIsValidV0(record))
- goto next_record_is_invalid;
- return record;
- }
-
- /* Record does not cross a page boundary */
- if (!RecordIsValidV0(record))
- goto next_record_is_invalid;
- memcpy(buffer, record, total_len);
- return (XLogRecordV0 *) buffer;
-
- next_record_is_invalid:;
- close(readFile);
- readFile = -1;
- return NULL;
- }
-
- /*
- * Check whether the xlog header of a page just read in looks valid.
- *
- * This is just a convenience subroutine to avoid duplicated code in
- * ReadRecord. It's not intended for use from anywhere else.
- */
- static bool
- ValidXLOGHeaderV0(XLogPageHeaderV0 hdr)
- {
- if (hdr->xlp_magic != XLOG_PAGE_MAGIC_V0)
- return false;
- if ((hdr->xlp_info & ~XLP_ALL_FLAGS) != 0)
- return false;
- return true;
- }
-
- /******************* end of routines for old XLOG format *******************/
-
-
/*
* Guess at pg_control values when we can't read the old ones.
*/
--- 134,152 ----
return true;
}
! fprintf(stderr, "pg_control exists but has invalid CRC; proceeding with caution.\n");
/* We will use the data anyway, but treat it as guessed. */
memcpy(&ControlFile, buffer, sizeof(ControlFile));
guessed = true;
return true;
}
/* Looks like it's a mess. */
fprintf(stderr, "pg_control exists but is broken or unknown version; ignoring it.\n");
return false;
}
/*
* Guess at pg_control values when we can't read the old ones.
*/
***************
*** 676,684 ****
* reset by RewriteControlFile().
*/
static void
! PrintControlValues(void)
{
! printf("Guessed-at pg_control values:\n\n"
"pg_control version number: %u\n"
"Catalog version number: %u\n"
"Current log file id: %u\n"
--- 216,224 ----
* reset by RewriteControlFile().
*/
static void
! PrintControlValues(bool guessed)
{
! printf("\n%spg_control values:\n\n"
"pg_control version number: %u\n"
"Catalog version number: %u\n"
"Current log file id: %u\n"
***************
*** 692,697 ****
--- 232,238 ----
"LC_COLLATE: %s\n"
"LC_CTYPE: %s\n",
+ (guessed ? "Guessed-at " : ""),
ControlFile.pg_control_version,
ControlFile.catalog_version_no,
ControlFile.logId,
***************
*** 1042,1048 ****
*/
if ((guessed && !force) || noupdate)
{
! PrintControlValues();
if (!noupdate)
printf("\nIf these values seem acceptable, use -f to force reset.\n");
exit(1);
--- 583,589 ----
*/
if ((guessed && !force) || noupdate)
{
! PrintControlValues(guessed);
if (!noupdate)
printf("\nIf these values seem acceptable, use -f to force reset.\n");
exit(1);
Bruce Momjian wrote:
Tom asked about pg_upgrade as part of our initdb for timezone.
I have made some improvements to pg_upgrade in CVS and have successfully
migrated a regression database from a 7.2 to another 7.2 database using
it. (At least the tables show some data; very light testing.)pg_upgrade is still disabled in CVS, it doesn't install, and there is no
manual page so it is still an unused command. I have made the commit so
people can review where I have gone and make comments.To test it, you have to find the line that says 7.2 and remove the '#'
comment. This is for testing purposes only, so far.
Status report: I have completed all the steps necessary for pg_upgrade
to work for 7.1->7.2 and for 7.2->7.2 databases. I will run tests
tomorrow, and once I am sure it works, I will ask others to test.
I will not enable it until everyone agrees. Are there people interested
in this tool being in 7.2, or who are against this tool being in 7.2?
--
Bruce Momjian | http://candle.pha.pa.us
pgman@candle.pha.pa.us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026
Bruce Momjian wrote:
Status report: I have completed all the steps necessary for pg_upgrade
to work for 7.1->7.2 and for 7.2->7.2 databases. I will run tests
tomorrow, and once I am sure it works, I will ask others to test.I will not enable it until everyone agrees. Are there people interested
in this tool being in 7.2, or who are against this tool being in 7.2?
I have a good sized (~11 GB) database on 7.2b3. I'd like to try
pg_upgrade to move it to CVS tip and/or 7.2RC1, so I guess I can be one
of your testers.
-- Joe
Bruce Momjian writes:
I will not enable it until everyone agrees. Are there people interested
in this tool being in 7.2, or who are against this tool being in 7.2?
You're not going to like my opinion, but I'm going to put it forth anyway.
We've been working on this release for half a year, and there have been
far too many last-minute bright ideas that should have been postponed.
The fact that someone is going to want to upgrade their installation from
a previous release didn't just occur to us yesterday, so while this
development effort is commendable, this is just not the time. I'm not
even going to list any technical reasons here, you can make up your own
list because you're looking at the code. I'm just looking at the emails
and it gives me the creeps already.
--
Peter Eisentraut peter_e@gmx.net
Peter Eisentraut wrote:
Bruce Momjian writes:
I will not enable it until everyone agrees. Are there people interested
in this tool being in 7.2, or who are against this tool being in 7.2?You're not going to like my opinion, but I'm going to put it forth anyway.
We've been working on this release for half a year, and there have been
far too many last-minute bright ideas that should have been postponed.
The fact that someone is going to want to upgrade their installation from
a previous release didn't just occur to us yesterday, so while this
development effort is commendable, this is just not the time. I'm not
even going to list any technical reasons here, you can make up your own
list because you're looking at the code. I'm just looking at the emails
and it gives me the creeps already.
Gives me the creeps too. :-)
I am working on it only because there isn't other stuff to do and it
isn't delaying anything because it is disabled anyway; we can keep it
for 7.3 if we wish. There was also the problem of a system catalog
change, and that got the fire moving. Also, certain commerical
distributors bug me about this from time to time so even if we don't
officially use it my guess is that some of them may enable it anyway for
their distributions.
--
Bruce Momjian | http://candle.pha.pa.us
pgman@candle.pha.pa.us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026
Bruce Momjian said:
I will not enable it until everyone agrees. Are there people interested
in this tool being in 7.2, or who are against this tool being in 7.2?
I have never trusted pg_upgrade on production databases. This is where it is
actually targeted to (to minimize downtime). I have always done full
pg_dumpall and then full recreation of the database system, which
unfortunately in my cases take few hours. With so many little changes in the
structures pg_upgrade it is probably not very safe method anyway?
Just my opinion. :-)
Daniel
Daniel Kalchev wrote:
Bruce Momjian said:
I will not enable it until everyone agrees. Are there people interested
in this tool being in 7.2, or who are against this tool being in 7.2?I have never trusted pg_upgrade on production databases. This is where it is
actually targeted to (to minimize downtime). I have always done full
pg_dumpall and then full recreation of the database system, which
unfortunately in my cases take few hours. With so many little changes in the
structures pg_upgrade it is probably not very safe method anyway?Just my opinion. :-)
I agree. There are just too many people who complain to me about a lack
of pg_upgrade that I have to do my best on it and let people decide if
they want to use it.
--
Bruce Momjian | http://candle.pha.pa.us
pgman@candle.pha.pa.us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026Peter Eisentraut <peter_e@gmx.net> writes:
... I'm just looking at the emails
and it gives me the creeps already.
FWIW, I would *never* trust a production database to pg_upgrade in its
current state; it's untested and can't possibly get enough testing
before release to be trustable. But if Bruce wants to work on it,
where's the harm? The discussions I've had with him over the past
couple days are more than valuable enough for development of a future
bulletproof pg_upgrade, whether or not the current script ever helps
anyone.
The only mistake we could make here is to advertise pg_upgrade as
reliable. Which we will not do.
regards, tom lane
Tom Lane wrote:
Peter Eisentraut <peter_e@gmx.net> writes:
... I'm just looking at the emails
and it gives me the creeps already.FWIW, I would *never* trust a production database to pg_upgrade in its
current state; it's untested and can't possibly get enough testing
before release to be trustable. But if Bruce wants to work on it,
where's the harm? The discussions I've had with him over the past
couple days are more than valuable enough for development of a future
bulletproof pg_upgrade, whether or not the current script ever helps
anyone.The only mistake we could make here is to advertise pg_upgrade as
reliable. Which we will not do.
Some people have large, non-critical databases they want to upgrade to
7.2. I can imagine some people using pg_upgrade for those cases.
--
Bruce Momjian | http://candle.pha.pa.us
pgman@candle.pha.pa.us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026Tom Lane writes:
FWIW, I would *never* trust a production database to pg_upgrade in its
current state; it's untested and can't possibly get enough testing
before release to be trustable. But if Bruce wants to work on it,
where's the harm?
There isn't any harm working on it, but the question was whether we want
to enable it in the 7.2 release. Given that you would "never" trust it in
its current state, and me just having seen the actual code, I think that
it's barely worth being put into contrib. Where in fact it should
probably go.
The only mistake we could make here is to advertise pg_upgrade as
reliable. Which we will not do.
Or ship pg_upgrade in a default installation and undermine the reliability
reputation for people who don't read advertisements.
--
Peter Eisentraut peter_e@gmx.net