Merge lp:~laurynas-biveinis/percona-xtrabackup/bug1098498-1132763-2.0 into lp:percona-xtrabackup/2.0

[Laurynas Biveinis]

Take 3:
Addressed the review comments.
http://jenkins.percona.com/job/percona-xtrabackup-2.0-param/363/
BT 28340.

Take 2:
http://jenkins.percona.com/job/percona-xtrabackup-2.0-param/347/

Alexey's suggestion of killing always is implemented. But xb_export.sh still needs a shutdown due to what seems to be a table import crash-non-safety. Also, not all the testcases have the server datadirs removed before the restore. Bug reports pending for both of these issues.

Take 1:
No BT or ST but is one of the prerequisites for BT 28340.
http://jenkins.percona.com/job/percona-xtrabackup-2.0-param/342/

Fix bug 1098498 (Sporadic xb_incremental_compressed_* failures in
Jenkins) and bug 1132763 (No need to shutdown servers cleanly before
wiping their datadirs in the testsuite).

Most tests shutdown the server cleanly before copying back the
restored datadir. These shutdowns may:
1) timeout sporadically (bug 1098498);
2) are not required in the first place (bug 1132763), because the
server data is about to be wiped, thus kill -9 would work just as
fine.

Fixed by
1) introducing new function kill_server() in test/inc/common.sh that
kill -9s, removes the PID file, releases the mysqld port and testsuite
server variables.
2) removing the shutdown timeout and kill-if-not-shutdown logic from
stop_server_with_id(). For those remaining cases where the server is
required to stop cleanly, kill is a wrong substitute, and the shutdown
timeouts can be taken care of by the overall testcase timeouts.
3) reviewing the testcases and replacing stop_server calls with
kill_server where safe.
4) fixing a typo in testrun.sh and using kill_server() in
get_version_info() too.
5) removing an unused function stop_all_servers() in common.sh.

[George Ormond Lorch III (gl-az)]

Nice work. I had started on something similar but like your solution better.

review: Approve (g2)

[Alexey Kopytov (akopytov)]

I went through the remaining stop_server occurrences and I see no reason to keep them. But then, if we don't really need stop_server, why not just make it do "kill -9" and avoid mass changes in test files?

*If* we ever need to stop the server and keep its data in a test case, then we can add either a new function, or a flag to stop_server.

review: Needs Fixing

[Laurynas Biveinis (laurynas-biveinis)]

Alexey -

We do need to keep the current stop_server in its remaining uses. These are the partial backups, where the old datadir is not wiped, and a corrupted ibd backup test - where a server kill precludes a checkpoint and apparently crash recovery fixes the corruption attempt.

If we renamed kill_server to stop_server and stop_server to shutdown_server, then the diff would be a bit smaller. What do you think?

[Alexey Kopytov (akopytov)]

The old datadir is wiped in partial backup tests, as you cannot restore a part of the dataset, you have to overwrite it.

In the corrupted ibd backup test, xtrabackup is started right after the server shutdown, so no crash recovery occurs.

[Laurynas Biveinis (laurynas-biveinis)]

Alexey -

In the partial backup testcases I reviewed I found the datadir wipe to be partial too. E.g. ib_part_tf_innodb.sh, after stopping the server, calls ib_part_restore, which takes care to wipe the datadir selectively only:

 rm -rf $mysql_datadir/test/*
 rm -rf $topdir/ext/*

It is possible that there are bugs too. Here is an example of xb_partial.sh failure when the server is killed instead of stopped: http://jenkins.percona.com/job/percona-xtrabackup-2.0-param/341/BUILD_TYPE=release,Host=debian6-32,xtrabackuptarget=innodb51_builtin/testReport/junit/%28root%29/t_xb_partial/sh/

Likewise for bug722638.sh: http://jenkins.percona.com/job/percona-xtrabackup-2.0-param/333/BUILD_TYPE=release,Host=ubuntu-precise-32bit,xtrabackuptarget=galera55/testReport/junit/%28root%29/t_bug722638/sh/

I cannot find the bug766033.sh failure I have mentioned, but I believe that something is wrong there, I did see that xtrabackup passed with a corrupted ibd. Not the crash recovery though.

If what I'm saying is correct, then we have to keep stop_server() for the partial-restore testcases, and we have to examine in more detail the rest. But the latter may result in bug reports, and, until they are fixed, they should keep on using stop_server().

[Alexey Kopytov (akopytov)]

The partial datadir wipe in in_part is wrong, it should remove it fully and restore from the backup directory. You can fix it in the same MP.

The second link for bug722638.sh failure is Galera-specific, and has nothing to do with the server shutdown. Bug #1098501.

The point is, until XtraBackup supports offline backups, it should never rely on the datadir state after stopping the server. If that's not true, those are test suite bugs and should be fixed together with this change.

[Laurynas Biveinis (laurynas-biveinis)]

Ack. Sorry for the Galera, I failed to noticed that somehow.

For the record, there are server restarts to force a checkpoint: bug766607, tar4ibd_symlink, xb_export. These can be replaced with kill + start for the same effect.

I will replace all shutdowns with kills then.

[Laurynas Biveinis (laurynas-biveinis)]

Actual BT 28340.

[George Ormond Lorch III (gl-az)]

review: Approve (g2)

[Alexey Kopytov (akopytov)]

- I don't understand the reason to call free_reserved_port and reset_server_variables only if the PID file exists in stop_server / shutdown_server. Don't we still want to cleanup in case the PID file disappeared for whatever reasons?

- Was the following change really necessary?

128 -dd if=/dev/zero of=$mysql_datadir/sakila/rental.ibd seek=1000 count=1
129 +printf '\xAA\xAA\xAA\xAA' | dd of=$mysql_datadir/sakila/rental.ibd seek=16384 count=4

- Was the following change really necessary?

215 # also test xtrabackup --stats work with --tables-file
216 -COUNT=`xtrabackup --stats --tables='test.test$' --datadir=$topdir/backup \
217 +COUNT=`xtrabackup --stats --tables='^(mysql.*|performance_schema.*|test.test)$' --datadir=$topdir/backup \

review: Needs Fixing

[Laurynas Biveinis (laurynas-biveinis)]

> - I don't understand the reason to call free_reserved_port and
> reset_server_variables only if the PID file exists in stop_server /
> shutdown_server. Don't we still want to cleanup in case the PID file
> disappeared for whatever reasons?

I'll revert it as it would make for a smaller diff. But I am not sure what would be the whatever reasons for PID to disappear. If this is a "can't happen" situation, then it's probably the case that the testsuite run wouldn't be able to recover to continue running for trusted results anyway?

> - Was the following change really necessary?
>
> 128 -dd if=/dev/zero of=$mysql_datadir/sakila/rental.ibd seek=1000 count=1
> 129 +printf '\xAA\xAA\xAA\xAA' | dd of=$mysql_datadir/sakila/rental.ibd
> seek=16384 count=4

I described it in the commit message:

"2) bug766033: replace the database corruption of poking one zero byte
to the first page with overwriting the start of the second page with
0xAAAAAAAA to reduce the risk of the previous corruption attempt not
corrupting anything due to the zero byte already existing at the
location."

Replacing the shutdown with the crash and using the original corruption attempt makes that testcase pass reliably. Poking a single zero byte to a first tablespace page doesn't look like a very reliable corruption attempt to me due to the chances that that byte will be zero already, which is presumably what happens with the crashed server.

> - Was the following change really necessary?
>
> 215 # also test xtrabackup --stats work with --tables-file
> 216 -COUNT=`xtrabackup --stats --tables='test.test$'
> --datadir=$topdir/backup \
> 217 +COUNT=`xtrabackup --stats
> --tables='^(mysql.*|performance_schema.*|test.test)$' --datadir=$topdir/backup
> \

No, thank you.

[Laurynas Biveinis (laurynas-biveinis)]

> Replacing the shutdown with the crash and using the original corruption
> attempt makes that testcase pass reliably.

Err, "makes xtrabackup pass and the testcase fail reliably".

[Alexey Kopytov (akopytov)]

Hi Laurynas,

On Wed, 06 Mar 2013 12:45:33 -0000, Laurynas Biveinis wrote:
>> - I don't understand the reason to call free_reserved_port and
>> reset_server_variables only if the PID file exists in stop_server /
>> shutdown_server. Don't we still want to cleanup in case the PID file
>> disappeared for whatever reasons?
>
> I'll revert it as it would make for a smaller diff. But I am not sure what would be the whatever reasons for PID to disappear. If this is a "can't happen" situation, then it's probably the case that the testsuite run wouldn't be able to recover to continue running for trusted results anyway?
>

I don't know. A test case removing the PID file or its directory for
some evil server/backup torturing? Or shutting down the server on its
own with mysqladmin shutdown for the same reason?

Just don't see any sense in avoiding those calls in case of a missing
PID file. It looks less reliable this way.

>> - Was the following change really necessary?
>>
>> 128 -dd if=/dev/zero of=$mysql_datadir/sakila/rental.ibd seek=1000 count=1
>> 129 +printf '\xAA\xAA\xAA\xAA' | dd of=$mysql_datadir/sakila/rental.ibd
>> seek=16384 count=4
>
> I described it in the commit message:
>
> "2) bug766033: replace the database corruption of poking one zero byte
> to the first page with overwriting the start of the second page with
> 0xAAAAAAAA to reduce the risk of the previous corruption attempt not
> corrupting anything due to the zero byte already existing at the
> location."
>
> Replacing the shutdown with the crash and using the original corruption attempt makes that testcase pass reliably. Poking a single zero byte to a first tablespace page doesn't look like a very reliable corruption attempt to me due to the chances that that byte will be zero already, which is presumably what happens with the crashed server.
>

OK. The reason I asked was that the code is not portable (dd semantics
is different on OSX/BSD). So the test case was corrupting another page
and worked mostly by accident on anything but Linux.

But I have checked it on my laptop, and the new code still works by
accident on OSX/BSD.

>> - Was the following change really necessary?
>>
>> 215 # also test xtrabackup --stats work with --tables-file
>> 216 -COUNT=`xtrabackup --stats --tables='test.test$'
>> --datadir=$topdir/backup \
>> 217 +COUNT=`xtrabackup --stats
>> --tables='^(mysql.*|performance_schema.*|test.test)$' --datadir=$topdir/backup
>> \
>
> No, thank you.
>

[Alexey Kopytov (akopytov)]

review: Approve

[Laurynas Biveinis (laurynas-biveinis)]

All the issues described in the MP and its reviews have their bug reports now:
1) testcases that don't wipe the server dirs before restore: https://bugs.launchpad.net/percona-xtrabackup/+bug/1155076;
2) tablespace import not crash safe: https://bugs.launchpad.net/percona-server/+bug/1000224;
3) dd nonportable: https://bugs.launchpad.net/percona-xtrabackup/+bug/1155079

=== modified file 'test/inc/common.sh'
--- test/inc/common.sh	2013-01-15 01:39:48 +0000
+++ test/inc/common.sh	2013-03-07 19:24:26 +0000
@@ -283,28 +283,22 @@
 }
 
 ########################################################################
-# Stop server with the id specified as the first argument and additional
-# command line arguments (if specified)
+# Stop server with the id specified as the first argument.  The server 
+# is stopped in the fastest possible way.
 ########################################################################
 function stop_server_with_id()
 {
     local id=$1
     switch_server $id
 
-    vlog "Stopping server with id=$id..."
+    vlog "Killing server with id=$id..."
 
     if [ -f "${MYSQLD_PIDFILE}" ]
     then
-	${MYSQLADMIN} ${MYSQL_ARGS} --shutdown-timeout=60 shutdown
-	if [ -f "${MYSQLD_PIDFILE}" ]
-	then
-	    vlog "Could not stop the server with id=$id, using kill -9"
-	    kill -9 `cat ${MYSQLD_PIDFILE}`
-	    rm -f ${MYSQLD_PIDFILE}
-	fi
-	vlog "Server with id=$id has been stopped"
+        kill -9 `cat ${MYSQLD_PIDFILE}`
+        rm -f ${MYSQLD_PIDFILE}
     else
-	vlog "Server pid file '${MYSQLD_PIDFILE}' doesn't exist!"
+        vlog "Server PID file '${MYSQLD_PIDFILE}' doesn't exist!"
     fi
 
     # unlock the port number
@@ -331,15 +325,25 @@
 }
 
 ########################################################################
-# Stop all running servers
+# Shutdown server with id=1 cleanly
 ########################################################################
-function stop_all_servers()
+function shutdown_server()
 {
-    local id
-    for id in ${SRV_MYSQLD_IDS[*]}
-    do
-	stop_server_with_id ${SRV_MYSQLD_IDS[$id]}
-    done
+    switch_server 1
+
+    vlog "Shutting down server with id=1..."
+
+    if [ -f "${MYSQLD_PIDFILE}" ]
+    then
+        ${MYSQLADMIN} ${MYSQL_ARGS} shutdown
+    else
+        vlog "Server PID file '${MYSQLD_PIDFILE}' doesn't exist!"
+    fi
+
+    # unlock the port number
+    free_reserved_port $MYSQLD_PORT
+
+    reset_server_variables 1
 }
 
 ########################################################################

=== modified file 'test/inc/ib_part.sh'
--- test/inc/ib_part.sh	2013-01-02 15:52:13 +0000
+++ test/inc/ib_part.sh	2013-03-07 19:24:26 +0000
@@ -56,18 +56,32 @@
 	ib_part_data $topdir $engine | run_cmd $MYSQL $MYSQL_ARGS test
 }
 
+function ib_part_add_mandatory_tables()
+{
+	local mysql_datadir=$1
+	local tables_file=$2
+	for table in $mysql_datadir/mysql/*.frm
+	do
+	        echo mysql.`basename $table` >> $tables_file
+	done
+	for table in $mysql_datadir/performance_schema/*.frm
+	do
+	    echo performance_schema.`basename $table` >> $tables_file
+	done
+}
+
 function ib_part_restore()
 {
 	topdir=$1
 	mysql_datadir=$2
 
 	# Remove database
-	rm -rf $mysql_datadir/test/*
+	rm -rf $mysql_datadir/*
 	rm -rf $topdir/ext/*
 	vlog "Original database removed"
 
 	# Restore database from backup
-	cp -rv $topdir/backup/test/* $mysql_datadir/test
+	cp -rv $topdir/backup/* $mysql_datadir
 	vlog "database restored from backup"
 
 }

=== modified file 'test/t/bug766033.sh'
--- test/t/bug766033.sh	2012-06-05 12:35:33 +0000
+++ test/t/bug766033.sh	2013-03-07 19:24:26 +0000
@@ -13,7 +13,7 @@
 vlog "Starting backup"
 
 # corrupt database
-dd if=/dev/zero of=$mysql_datadir/sakila/rental.ibd seek=1000 count=1
+printf '\xAA\xAA\xAA\xAA' | dd of=$mysql_datadir/sakila/rental.ibd seek=16384 count=4
 
 # we want xtrabackup to be failed on rental.ibd
 run_cmd_expect_failure $XB_BIN $XB_ARGS  --backup --datadir=$mysql_datadir \

=== modified file 'test/t/bug766607.sh'
--- test/t/bug766607.sh	2012-06-05 12:35:33 +0000
+++ test/t/bug766607.sh	2013-03-07 19:24:26 +0000
@@ -3,12 +3,10 @@
 start_server --innodb_file_per_table
 load_dbase_schema incremental_sample
 
-# Full backup dir
-mkdir -p $topdir/data/full
-# Incremental backup dir
-mkdir -p $topdir/data/delta
+# Backup dir
+mkdir -p $topdir/backup
 
-xtrabackup --datadir=$mysql_datadir --backup --target-dir=$topdir/data/full
+innobackupex --no-timestamp $topdir/backup/full
 
 vlog "Creating and filling a new table"
 run_cmd $MYSQL $MYSQL_ARGS test <<EOF
@@ -21,26 +19,25 @@
 start_server --innodb_file_per_table
 
 vlog "Making incremental backup"
-xtrabackup --datadir=$mysql_datadir --backup --target-dir=$topdir/data/delta --incremental-basedir=$topdir/data/full
+innobackupex --incremental --no-timestamp --incremental-basedir=$topdir/backup/full $topdir/backup/delta
 
 vlog "Preparing full backup"
-xtrabackup --datadir=$mysql_datadir --prepare --apply-log-only \
-    --target-dir=$topdir/data/full
+innobackupex --apply-log --redo-only $topdir/backup/full
 
 # The following would fail before the bugfix
 vlog "Applying incremental delta"
-xtrabackup --datadir=$mysql_datadir --prepare --apply-log-only \
-    --target-dir=$topdir/data/full --incremental-dir=$topdir/data/delta
+innobackupex --apply-log --redo-only --incremental-dir=$topdir/backup/delta $topdir/backup/full
 
 vlog "Preparing full backup"
-xtrabackup --datadir=$mysql_datadir --prepare --target-dir=$topdir/data/full
+innobackupex --apply-log $topdir/backup/full
 vlog "Data prepared for restore"
 
 stop_server
+rm -r $mysql_datadir/*
 
 vlog "Copying files"
 
-cd $topdir/data/full/
+cd $topdir/backup/full/
 cp -r * $mysql_datadir
 cd $topdir
 

=== modified file 'test/t/ib_part_databases.sh'
--- test/t/ib_part_databases.sh	2013-01-02 15:52:13 +0000
+++ test/t/ib_part_databases.sh	2013-03-07 19:24:26 +0000
@@ -22,6 +22,8 @@
 
 # Take a backup
 cat >$topdir/databases_file <<EOF
+mysql
+performance_schema
 test.test
 EOF
 innobackupex --no-timestamp --databases=$topdir/databases_file $topdir/backup

=== modified file 'test/t/ib_part_include.sh'
--- test/t/ib_part_include.sh	2013-01-02 15:52:13 +0000
+++ test/t/ib_part_include.sh	2013-03-07 19:24:26 +0000
@@ -21,10 +21,7 @@
 
 # Take a backup
 # Only backup of test.test table will be taken
-cat >$topdir/tables <<EOF
-test.test
-EOF
-innobackupex --no-timestamp --include='test.test$' $topdir/backup
+innobackupex --no-timestamp --include='^(mysql.*|performance_schema.*|test.test)$' $topdir/backup
 innobackupex --apply-log $topdir/backup
 vlog "Backup taken"
 

=== modified file 'test/t/ib_part_include_stream.sh'
--- test/t/ib_part_include_stream.sh	2013-01-02 15:52:13 +0000
+++ test/t/ib_part_include_stream.sh	2013-03-07 19:24:26 +0000
@@ -21,7 +21,7 @@
 
 # Take a backup
 mkdir -p $topdir/backup
-innobackupex --stream=tar --include='test.test$' $topdir/backup > $topdir/backup/backup.tar
+innobackupex --stream=tar --include='^(mysql.*|performance_schema.*|test.test)$' $topdir/backup > $topdir/backup/backup.tar
 $TAR ixvf $topdir/backup/backup.tar -C $topdir/backup 
 $TAR cvhf $topdir/backup/backup11.tar $mysql_datadir/test/*
 

=== modified file 'test/t/ib_part_tf_innodb.sh'
--- test/t/ib_part_tf_innodb.sh	2013-01-02 15:52:13 +0000
+++ test/t/ib_part_tf_innodb.sh	2013-03-07 19:24:26 +0000
@@ -24,6 +24,7 @@
 cat >$topdir/tables <<EOF
 test.test
 EOF
+ib_part_add_mandatory_tables $mysql_datadir $topdir/tables
 innobackupex --no-timestamp --tables-file=$topdir/tables $topdir/backup
 innobackupex --apply-log $topdir/backup
 vlog "Backup taken"

=== modified file 'test/t/ib_part_tf_innodb_stream.sh'
--- test/t/ib_part_tf_innodb_stream.sh	2013-01-02 15:52:13 +0000
+++ test/t/ib_part_tf_innodb_stream.sh	2013-03-07 19:24:26 +0000
@@ -24,6 +24,7 @@
 cat >$topdir/tables <<EOF
 test.test
 EOF
+ib_part_add_mandatory_tables $mysql_datadir $topdir/tables
 mkdir -p $topdir/backup
 innobackupex --stream=tar --no-timestamp --tables-file=$topdir/tables $topdir/backup > $topdir/backup/backup.tar
 $TAR ixvf $topdir/backup/backup.tar -C $topdir/backup 

=== modified file 'test/t/ib_part_tf_myisam.sh'
--- test/t/ib_part_tf_myisam.sh	2013-01-02 15:52:13 +0000
+++ test/t/ib_part_tf_myisam.sh	2013-03-07 19:24:26 +0000
@@ -25,6 +25,7 @@
 cat >$topdir/tables <<EOF
 test.test
 EOF
+ib_part_add_mandatory_tables $mysql_datadir $topdir/tables
 innobackupex --no-timestamp --tables-file=$topdir/tables $topdir/backup
 innobackupex --apply-log $topdir/backup
 vlog "Backup taken"

=== modified file 'test/t/xb_export.sh'
--- test/t/xb_export.sh	2012-10-15 16:14:59 +0000
+++ test/t/xb_export.sh	2013-03-07 19:24:26 +0000
@@ -88,7 +88,7 @@
 # consistent backup results. Otherwise we risk ending up with no test.ibd
 # in the backup in case importing has not finished before taking backup
 
-stop_server
+shutdown_server
 start_server $mysql_extra_args
 
 # Some testing queries

=== removed file 'test/t/xb_incremental.sh'
--- test/t/xb_incremental.sh	2012-10-15 16:14:59 +0000
+++ test/t/xb_incremental.sh	1970-01-01 00:00:00 +0000
@@ -1,106 +0,0 @@
-. inc/common.sh
-
-start_server --innodb_file_per_table
-
-load_dbase_schema incremental_sample
-
-# Adding 10k rows
-
-vlog "Adding initial rows to database..."
-
-numrow=100
-count=0
-while [ "$numrow" -gt "$count" ]
-do
-	${MYSQL} ${MYSQL_ARGS} -e "insert into test values ($count, $numrow);" incremental_sample
-	let "count=count+1"
-done
-
-
-vlog "Initial rows added"
-
-# Full backup
-
-# Full backup folder
-mkdir -p $topdir/data/full
-# Incremental data
-mkdir -p $topdir/data/delta
-
-vlog "Starting backup"
-
-xtrabackup --datadir=$mysql_datadir --backup --target-dir=$topdir/data/full
-
-vlog "Full backup done"
-
-# Changing data in sakila
-
-vlog "Making changes to database"
-
-${MYSQL} ${MYSQL_ARGS} -e "create table t2 (a int(11) default null, number int(11) default null) engine=innodb" incremental_sample
-let "count=numrow+1"
-let "numrow=1000"
-while [ "$numrow" -gt "$count" ]
-do
-	${MYSQL} ${MYSQL_ARGS} -e "insert into test values ($count, $numrow);" incremental_sample
-        ${MYSQL} ${MYSQL_ARGS} -e "insert into t2 values ($count, $numrow);" incremental_sample
-	let "count=count+1"
-done
-
-vlog "Changes done"
-
-# Saving the checksum of original table
-checksum_test_a=`checksum_table incremental_sample test`
-checksum_t2_a=`checksum_table incremental_sample t2`
-
-vlog "Table 'test' checksum is $checksum_test_a"
-vlog "Table 't2' checksum is $checksum_t2_a"
-vlog "Making incremental backup"
-
-# Incremental backup
-xtrabackup --datadir=$mysql_datadir --backup --target-dir=$topdir/data/delta --incremental-basedir=$topdir/data/full
-
-vlog "Incremental backup done"
-vlog "Preparing backup"
-
-# Prepare backup
-xtrabackup --datadir=$mysql_datadir --prepare --apply-log-only \
-    --target-dir=$topdir/data/full
-vlog "Log applied to backup"
-xtrabackup --datadir=$mysql_datadir --prepare --apply-log-only \
-    --target-dir=$topdir/data/full --incremental-dir=$topdir/data/delta
-vlog "Delta applied to backup"
-xtrabackup --datadir=$mysql_datadir --prepare --target-dir=$topdir/data/full
-vlog "Data prepared for restore"
-
-# removing rows
-run_cmd ${MYSQL} ${MYSQL_ARGS} -e "delete from test;" incremental_sample
-run_cmd ${MYSQL} ${MYSQL_ARGS} -e "delete from t2;" incremental_sample
-vlog "Tables cleared"
-
-# Restore backup
-stop_server
-vlog "Copying files"
-cd $topdir/data/full/
-cp -r * $mysql_datadir
-cd $topdir
-
-vlog "Data restored"
-start_server --innodb_file_per_table
-
-vlog "Checking checksums"
-checksum_test_b=`checksum_table incremental_sample test`
-checksum_t2_b=`checksum_table incremental_sample t2`
-
-if [ "$checksum_test_a" != "$checksum_test_b"  ]
-then 
-	vlog "Checksums of table 'test' are not equal"
-	exit -1
-fi
-
-if [ "$checksum_t2_a" != "$checksum_t2_b"  ]
-then 
-	vlog "Checksums of table 't2' are not equal"
-	exit -1
-fi
-
-vlog "Checksums are OK"

=== removed file 'test/t/xb_partial.sh'
--- test/t/xb_partial.sh	2012-10-15 16:14:59 +0000
+++ test/t/xb_partial.sh	1970-01-01 00:00:00 +0000
@@ -1,53 +0,0 @@
-. inc/common.sh
-
-start_server --innodb_file_per_table
-
-load_dbase_schema incremental_sample
-
-# Adding 10k rows
-vlog "Adding initial rows to database..."
-numrow=1000
-count=0
-while [ "$numrow" -gt "$count" ]
-do
-	${MYSQL} ${MYSQL_ARGS} -e "insert into test values ($count, $numrow);" incremental_sample
-	let "count=count+1"
-done
-vlog "Initial rows added"
-checksum_a=`checksum_table incremental_sample test`
-vlog "Table checksum is $checksum_a"
-
-# Backup directory
-mkdir -p $topdir/data/parted
-
-vlog "Starting backup"
-xtrabackup --datadir=$mysql_datadir --backup --target-dir=$topdir/data/parted \
-    --tables="^incremental_sample[.]test"
-vlog "Partial backup done"
-
-# Prepare backup
-xtrabackup --datadir=$mysql_datadir --prepare --target-dir=$topdir/data/parted
-vlog "Data prepared for restore"
-
-# removing rows
-run_cmd ${MYSQL} ${MYSQL_ARGS} -e "delete from test;" incremental_sample
-vlog "Table cleared"
-
-# Restore backup
-stop_server
-vlog "Copying files"
-cd $topdir/data/parted/
-cp -r * $mysql_datadir
-cd $topdir
-vlog "Data restored"
-start_server --innodb_file_per_table
-vlog "Checking checksums"
-checksum_b=`checksum_table incremental_sample test`
-
-if [ "$checksum_a" != "$checksum_b"  ]
-then 
-	vlog "Checksums are not equal"
-	exit -1
-fi
-
-vlog "Checksums are OK"

=== modified file 'test/testrun.sh'
--- test/testrun.sh	2013-01-08 06:23:22 +0000
+++ test/testrun.sh	2013-03-07 19:24:26 +0000
@@ -171,8 +171,7 @@
 	    else
 		XB_BIN="xtrabackup"    # InnoDB 5.1 plugin or Percona Server 5.1
 	    fi
-	elif [ "${MYSQL_VERSION:0:3}" = "5.2" -o
-	       "${MYSQL_VERSION:0:3}" = "5.3"]
+	elif [ "${MYSQL_VERSION:0:3}" = "5.2" -o "${MYSQL_VERSION:0:3}" = "5.3" ]
 	then
 	    XB_BIN="xtrabackup"
 	elif [ "${MYSQL_VERSION:0:3}" = "5.5" ]