pg_start_backup does not actually allow for consistent, file-level backup

Am 2015-06-08 14:45, schrieb otheus uibk:

The manual and in this mailing list, the claim is made that
consistent, file-level backups may be made by bracketing the
file-copy
operation with the postgresql pg_start_backup and pg_stop_backup
operations.  Many people including myself have found that in some
circumstances, using "tar" to copy these files will result in an
error
if one of the data files changes during the tar operation. The
responses to those queries on this mailing list are unsatisfactory
("everything is fine, trust us").

[...]

I decided to test this claim that these messages are "perfectly
harmless" and "can be ignored":

1. I executed pg_start_backup() on server
2. Ran md5sum recursively through PGs data directories
3. waited a split second
4. Ran md5sum recursively through PGs data directories as in step 2
5. Compared output from #2 and #4

As you can see below, there were non-zero changes made to these
files. 

< a1a571bfd1e4a98b20245edbdfce6d9a
 /var/lib/pgsql/data/base/41514/809275
---

21de5b864019c96c55e81a38fa1c9ccf

 /var/lib/pgsql/data/base/41514/809275
1783c1783
< 8eb4a578ecb56667e1698174f89c462c
 /var/lib/pgsql/data/base/41514/809280
---

b4c7b4ef30dda9543181465f53a85d72

 /var/lib/pgsql/data/base/41514/809280

Such changes occurred EVEN WHEN TAR DID NOT WARN of changed files.
Further, when step 3 involved an actual backup, involving minutes,
not
milliseconds, dozens of differences to files in data/base/... are
reported. To be clear, I excluded from consideration all files in
pg_xlog, pg_clog, pg_subtrans, pg_stat_tmp.

If these files are changing during the pg_start_backup() and
pg_stop_backup, then exactly what is their purpose? Might they be
changing during the tar, as tar thinks? How may an operator be
assured
the snapshot is consistent (unless one stops the databases)?  Will
the redo logs restore the files to a consistent state, no matter when
these files are changed? I find it hard to believe that would be the
case.

This test was performed using Postgresql 9.1.8. A scan of the
CHANGELOG since then indicates that if this is a bug, it has not been
reported as fixed.

Still everything is fine here. You need to understand that in between
pg_start_ and pg_stop_backup Postgres continues to operate aus usual -
so files in $PGDATA directory WILL change. That's why it is necessary to
also keep all the WAL segments that where created during _start and
_stop to actually recover to a consistent state. When you recover from a
full (file based) backup the WAL files file be applied, too (that is why
you need a recovery.conf and a restore_command.

You should possibly re-read
http://www.postgresql.org/docs/9.4/static/continuous-archiving.html#BACKUP-PITR-RECOVERY
especially 24.3.3 and 24.3.4.

hth

Jan

--
Sent via pgsql-general mailing list (pgsql-general@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-general

Le 8 juin 2015 2:48 PM, "otheus uibk" <otheus.uibk@gmail.com> a écrit :

The manual and in this mailing list, the claim is made that consistent,

file-level backups may be made by bracketing the file-copy operation with
the postgresql pg_start_backup and pg_stop_backup operations. Many people
including myself have found that in some circumstances, using "tar" to copy
these files will result in an error if one of the data files changes during
the tar operation. The responses to those queries on this mailing list are
unsatisfactory ("everything is fine, trust us").

For example:

/messages/by-id/AANLkTikg70iP91TPUwVuF9gb0XtS4fukaKM4qXLTwUfP@mail.gmail.com

and

/messages/by-id/201004201350.o3KDoKR02633@momjian.us
(quoted below)

On Fri, Apr 16, 2010 at 11:55 AM, raghavendra t
<raagavendra(dot)rao(at)gmail(dot)com> wrote:

Hi All,

For some setups reason, i started taking Hot backup. In this course I

have first issued pg_start_backup('backup') and went to the data directory
for backing up in OS format using the command "tar -cf backup.tar /data" .
When i issued this command , tar was generating some errors as show below.

bash-3.00# tar -cf 16aprilstandby.tar /db-data/
tar: Removing leading `/' from member names
tar: /db-data/base/24643/151667: file changed as we read it
tar: /db-data/base/24643/151764.2: file changed as we read it
tar: /db-data/base/24643/151766: file changed as we read it
tar: /db-data/base/24643/151768: file changed as we read it
tar: /db-data/base/66412/151969: file changed as we read it

After sometime tar has ended and i also issued pg_stop_backup() and i

continued the process.

My question here is, is this errors generated by tar are to worrisome

or whats happening in the backend.

Is "tar" file is safewell to use. Could you please tell me.

Those are not errors, they are warnings. As long as you use
pg_start_backup() and pg_stop_backup() before and after the tar, they
are perfectly harmless, and can be ignored.

The above scenario is exactly what I saw, albeit with less frequency and

severity.

I decided to test this claim that these messages are "perfectly harmless"

and "can be ignored":

1. I executed pg_start_backup() on server
2. Ran md5sum recursively through PG's data directories
3. waited a split second
4. Ran md5sum recursively through PG's data directories as in step 2
5. Compared output from #2 and #4

As you can see below, there were non-zero changes made to these files.

< a1a571bfd1e4a98b20245edbdfce6d9a /var/lib/pgsql/data/base/41514/809275
---

21de5b864019c96c55e81a38fa1c9ccf /var/lib/pgsql/data/base/41514/809275

1783c1783
< 8eb4a578ecb56667e1698174f89c462c /var/lib/pgsql/data/base/41514/809280
---

b4c7b4ef30dda9543181465f53a85d72 /var/lib/pgsql/data/base/41514/809280

Such changes occurred EVEN WHEN TAR DID NOT WARN of changed files.

Further, when step 3 involved an actual backup, involving minutes, not
milliseconds, dozens of differences to files in data/base/... are reported.
To be clear, I excluded from consideration all files in pg_xlog, pg_clog,
pg_subtrans, pg_stat_tmp.

If these files are changing during the pg_start_backup() and

pg_stop_backup, then exactly what is their purpose?

First, files change during backup. The files backup isn't consistent in
itself. You need the archived wal files to make it consistent.

pg_start_backup(), among other things, creates the backup_label file which
is needed at restore time so that postgres knows at which wal file it needs
to start the replay. pg_stop_backup() gives the last position in the wal
file needs to get a consistent backup.

Might they be changing during the tar, as tar thinks?

Yes.

How may an operator be assured the snapshot is consistent (unless one

stops the databases)? Will the redo logs restore the files to a consistent
state, no matter when these files are changed? I find it hard to believe
that would be the case.

You should because it works great :-)

This test was performed using Postgresql 9.1.8. A scan of the CHANGELOG

since then indicates that if this is a bug, it has not been reported as
fixed.

otheus uibk wrote:

The manual and in this mailing list, the claim is made that consistent, file-level backups may be made
by bracketing the file-copy operation with the postgresql pg_start_backup and pg_stop_backup
operations. Many people including myself have found that in some circumstances, using "tar" to copy
these files will result in an error if one of the data files changes during the tar operation. The
responses to those queries on this mailing list are unsatisfactory ("everything is fine, trust us").

Everything is fine, trust us.

bash-3.00# tar -cf 16aprilstandby.tar /db-data/
tar: Removing leading `/' from member names
tar: /db-data/base/24643/151667: file changed as we read it
tar: /db-data/base/24643/151764.2: file changed as we read it
tar: /db-data/base/24643/151766: file changed as we read it
tar: /db-data/base/24643/151768: file changed as we read it
tar: /db-data/base/66412/151969: file changed as we read it

The above scenario is exactly what I saw, albeit with less frequency and severity.

I decided to test this claim that these messages are "perfectly harmless" and "can be ignored":

[...]

As you can see below, there were non-zero changes made to these files.

[...]

Such changes occurred EVEN WHEN TAR DID NOT WARN of changed files. Further, when step 3 involved an
actual backup, involving minutes, not milliseconds, dozens of differences to files in data/base/...
are reported. To be clear, I excluded from consideration all files in pg_xlog, pg_clog, pg_subtrans,
pg_stat_tmp.

If these files are changing during the pg_start_backup() and pg_stop_backup, then exactly what is
their purpose? Might they be changing during the tar, as tar thinks? How may an operator be assured
the snapshot is consistent (unless one stops the databases)? Will the redo logs restore the files to
a consistent state, no matter when these files are changed? I find it hard to believe that would be
the case.

The files are indeed changing while they are backed up.

The tar archive is not a consistent backup.

Redo using the Write Ahead Log will indeed restore the files to a consistent state,
astonishing as that may be.

Yours,
Laurenz Albe

--
Sent via pgsql-general mailing list (pgsql-general@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-general

On 06/08/15 14:45, otheus uibk wrote:

The manual and in this mailing list, the claim is made that consistent,
file-level backups may be made by bracketing the file-copy operation
with the postgresql pg_start_backup and pg_stop_backup operations. Many
people including myself have found that in some circumstances, using
"tar" to copy these files will result in an error if one of the data
files changes during the tar operation. The responses to those queries
on this mailing list are unsatisfactory ("everything is fine, trust us").

I don't really see what you find unsatisfactory on those responses?
While performing the backup, the database is running and either
modifying or even deleting the files. Tar does not expect that, and thus
emits warnings/errors. That's clearly explained in the docs, and
actually linked in one of the comments you quoted. And pg_start_backup
takes care of that.

If you don't like that, you have multiple options - stop the database
while performing the backup, perform file system level backup (e.g. lvm
snapshot) or use tools like pg_basebackup.

regards

--
Tomas Vondra http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

--
Sent via pgsql-general mailing list (pgsql-general@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-general

Thank you, all. The manual for 9.4 is indeed clearer on this point than
the 9.1 version.

On Mon, Jun 8, 2015 at 3:13 PM, otheus uibk <otheus.uibk@gmail.com> wrote:

Thank you, all. The manual for 9.4 is indeed clearer on this point than
the 9.1 version.

Just to nit-pick, I see nowhere in either version of the manual the
indication that it is normal for postgresql to continue to update files in
its data catalog between pg_start_backup and pg_stop_backup. The closest I
see comes in this paragraph:

** Some file system backup tools emit warnings or errors if the files they
are trying to copy change while the copy proceeds. When taking a base
backup of an active database, this situation is normal and not an error.

Does "this situation" refer to the tools emitting warnings or to the fact
that postgresql is updating the files? It might be the case, for instance,
that timestamps are updated but not the contents of the files (this is what
I had assumed prior to today).

--
Otheus
otheus.uibk@gmail.com
otheus.shelling@uibk.ac.at

On 06/08/2015 05:45 AM, otheus uibk wrote:

The manual and in this mailing list, the claim is made that consistent,
file-level backups may be made by bracketing the file-copy operation
with the postgresql pg_start_backup and pg_stop_backup operations. Many
people including myself have found that in some circumstances, using
"tar" to copy these files will result in an error if one of the data
files changes during the tar operation. The responses to those queries
on this mailing list are unsatisfactory ("everything is fine, trust us").

If these files are changing during the pg_start_backup() and
pg_stop_backup, then exactly what is their purpose? Might they be
changing during the tar, as tar thinks? How may an operator be assured
the snapshot is consistent (unless one stops the databases)? Will the
redo logs restore the files to a consistent state, no matter when these
files are changed? I find it hard to believe that would be the case.

The important part can be found here:

http://www.postgresql.org/docs/9.4/interactive/wal-intro.html

"Briefly, WAL's central concept is that changes to data files (where
tables and indexes reside) must be written only after those changes have
been logged, that is, after log records describing the changes have been
flushed to permanent storage."

So basicall as long as you have the relevant WAL logs, which is what
pg_start_backup and pg_stop_backup log, then you are covered.

This test was performed using Postgresql 9.1.8. A scan of the CHANGELOG
since then indicates that if this is a bug, it has not been reported as
fixed.

--
Otheus
otheus.uibk@gmail.com <mailto:otheus.uibk@gmail.com>
otheus.shelling@uibk.ac.at <mailto:otheus.shelling@uibk.ac.at>

--
Adrian Klaver
adrian.klaver@aklaver.com

--
Sent via pgsql-general mailing list (pgsql-general@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-general

On 06/08/2015 06:26 AM, otheus uibk wrote:

On Mon, Jun 8, 2015 at 3:13 PM, otheus uibk <otheus.uibk@gmail.com
<mailto:otheus.uibk@gmail.com>> wrote:

Thank you, all. The manual for 9.4 is indeed clearer on this point
than the 9.1 version.

Just to nit-pick, I see nowhere in either version of the manual the
indication that it is normal for postgresql to continue to update files
in its data catalog between pg_start_backup and pg_stop_backup. The
closest I see comes in this paragraph:

** Some file system backup tools emit warnings or errors if the files
they are trying to copy change while the copy proceeds. When taking a
base backup of an active database, this situation is normal and not an
error.

Does "this situation" refer to the tools emitting warnings or to the
fact that postgresql is updating the files? It might be the case, for
instance, that timestamps are updated but not the contents of the files
(this is what I had assumed prior to today).

Not sure what you are using these backups for, but have you looked at?:

http://www.postgresql.org/docs/9.4/static/app-pgbasebackup.html

--
Otheus
otheus.uibk@gmail.com <mailto:otheus.uibk@gmail.com>
otheus.shelling@uibk.ac.at <mailto:otheus.shelling@uibk.ac.at>

--
Adrian Klaver
adrian.klaver@aklaver.com

--
Sent via pgsql-general mailing list (pgsql-general@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-general

On Mon, Jun 8, 2015 at 9:26 AM, otheus uibk <otheus.uibk@gmail.com> wrote:

On Mon, Jun 8, 2015 at 3:13 PM, otheus uibk <otheus.uibk@gmail.com> wrote:

Thank you, all. The manual for 9.4 is indeed clearer on this point than
the 9.1 version.

Just to nit-pick, I see nowhere in either version of the manual the
indication that it is normal for postgresql to continue to update files in
its data catalog between pg_start_backup and pg_stop_backup. The closest I
see comes in this paragraph:

It may not be explicit because the author assumes that such writing
occurring is self-evident. If it did not then for any reasonably sized
database the system would I/O starve or simply run out of memory as the
changes continue to pile up in the buffers while the backup is running.

The normal operation of the database operates on the same principle. In
between checkpoints the data files are constantly being written. A crash
means that there is more data in WAL to write out to the data files. The
data files cannot be reverted back to their state at the time of the
checkpoint. Instead, the replay of the WAL simply reapplies every change -
even those that did make it out during the period between the checkpoint
and the crash.

​As far as the backup routine is concerned a checkpoint occurs only when
you call pg_start/end_backup and the system ensures that every WAL file
generated during the intervening period is kept around so that the data
files being copied, and changed, in the intervening period can be made
whole. The first checkpoint ensure that all data in present in those files
and a restoration will replay every WAL file to ensure that the final state
of each data file matches the state of the database as of the time the last
WAL file present was created (typically the one closed out at the
pg_end_backup call).

David J.​

otheus uibk wrote:

Just to nit-pick, I see nowhere in either version of the manual the indication that it is normal for
postgresql to continue to update files in its data catalog between pg_start_backup and pg_stop_backup.
The closest I see comes in this paragraph:

** Some file system backup tools emit warnings or errors if the files they are trying to copy change
while the copy proceeds. When taking a base backup of an active database, this situation is normal and
not an error.

Does "this situation" refer to the tools emitting warnings or to the fact that postgresql is updating
the files? It might be the case, for instance, that timestamps are updated but not the contents of the
files (this is what I had assumed prior to today).

The manual does not contain all the details how backup and recovery works internally,
you'd have to see the source code for that.

It is normal for the files to change while backup is in progress (in fact, the database
continues working normally, but more information is written to the Write Ahead Log).
It is also normal for backup tools to complain if the file they copy changes while they
read it.

Yours,
Laurenz Albe

--
Sent via pgsql-general mailing list (pgsql-general@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-general