pt-online-schema-change: Use of uninitialized value $host in string

[Paul Namuag]

We tried to run on a customer's side identically like the command below:

pt-online-schema-change --execute --alter-foreign-keys-method auto --set-vars "tx_isolation='READ-COMMITTED'" --alter "CHANGE COLUMN employerIdentificationNumber employerIdentificationNumber VARCHAR(50) NULL DEFAULT NULL" --critical-load Threads_running=200 --max-load Threads_running=200 h=10.5.1.1,D=somedb,t=sometable

and we encountered this error:

Use of uninitialized value $host in string eq at /usr/bin/pt-online-schema-change line 4245.

Upon checking with the PTDEBUG=1, it shows the following:

# Cxn:3886 18596 DBI::db=HASH(0x18a04d8) Setting dbh
# Cxn:3891 18596 DBI::db=HASH(0x18a04d8) SELECT @@server_id /*!50038 , @@hostname*/
# Cxn:3893 18596 DBI::db=HASH(0x18a04d8) hostname: xxxxxx.ocal 52
# Cxn:3874 18596 DBI::db=HASH(0x18a04d8) Connected dbh to xxxxx.local h=10.15.11.52
# MasterSlave:4218 18596 Looking for slaves on D=somedb,h=10.15.11.52 using methods processlist hosts
# MasterSlave:4225 18596 Finding slaves with _find_slaves_by_processlist
# MasterSlave:4287 18596 DBI::db=HASH(0x18a04d8) SHOW GRANTS FOR CURRENT_USER()
# MasterSlave:4317 18596 DBI::db=HASH(0x18a04d8) SHOW PROCESSLIST
Use of uninitialized value $host in string eq at /usr/bin/pt-online-schema-change line 4245.
# Cxn:4012 18596 Destroying cxn
# Cxn:4021 18596 DBI::db=HASH(0x17c6a58) Disconnecting dbh on xxxxx h=10.15.11.46
# Cxn:4012 18596 Destroying cxn
# Cxn:4021 18596 DBI::db=HASH(0x18a04d8) Disconnecting dbh on xxxxx h=10.15.11.52
# Cxn:4012 18596 Destroying cxn
# Cxn:4021 18596 DBI::db=HASH(0x188d240) Disconnecting dbh on xxxxx h=10.15.11.23
# Cxn:4012 18596 Destroying cxn
# Cxn:4021 18596 DBI::db=HASH(0x17c94f0) Disconnecting dbh on xxxxx h=10.15.11.46

Upon checking the line, I got this line of code (based on the line from error: Use of uninitialized value $host in string eq at /usr/bin/pt-online-schema-change line 4245.)

$ vim /usr/bin/pt-online-schema-change +4245

 4234 sub _find_slaves_by_processlist {
 4235 my ( $self, $dsn_parser, $dbh, $dsn ) = @_;
 4236
 4237 my @slaves = map {
 4238 my $slave = $dsn_parser->parse("h=$_", $dsn);
 4239 $slave->{source} = 'processlist';
 4240 $slave;
 4241 }
 4242 grep { $_ }
 4243 map {
 4244 my ( $host ) = $_->{host} =~ m/^([^:]+):/;
 4245 if ( $host eq 'localhost' ) {
 4246 $host = '127.0.0.1'; # Replication never uses sockets.
 4247 }
 4248 $host;
 4249 } $self->get_connected_slaves($dbh);
 4250
 4251 return @slaves;
 4252 }

The bug occurs there as it seems that the host it tries to detect is empty or null as it seems that it fails here -> my ( $host ) = $_->{host} =~ m/^([^:]+):/;

but it founds a host that has no "semi-colon", for example "localhost"

Based on the processlist, I see this:

| 1 | event_scheduler | localhost | NULL | Daemon | 40041 | Waiting for next activation | NULL |
| 9 | someuser | 10.15.11.23:50630 | NULL | Binlog Dump | 1031379 | Master has sent all binlog to slave; waiting for binlog to be updated | NULL |
| 17 | someuser | 10.15.11.52:57571 | NULL | Binlog Dump | 1031369 | Master has sent all binlog to slave; waiting for binlog to be updated | NULL |
| 18 | someuser | 10.19.11.31:54500 | NULL | Binlog Dump | 1031369 | Master has sent all binlog to slave; waiting for binlog to be updated | NULL |
| 284044 | someuser | 10.15.11.145:55431 | NULL | Binlog Dump | 843843 | Master has sent all binlog to slave; waiting for binlog to be updated | NULL |

So the bug occurs there.

The work-around on this is to use --recursion-method=dsn="h=yourdsnhost,D=percona,t=dsns".

Version:
 ~ #> pt-online-schema-change --version
pt-online-schema-change 2.2.17

[Shahriyar Rzayev (rzayev-sehriyar)]

Percona now uses JIRA for bug reports so this bug report is migrated to: https://jira.percona.com/browse/PT-1344