pt-table-checksum stuck on "Waiting to check replicas for differences"

Stony · December 30, 2020, 4:24pm

mariadb: 10.2
percona-toolkit: 3.0.4
OS: centOS7.0

I’m sure replication is working well. also I restart master and slave.

The user checksums has enough privileges on both master and slave.

The tables schema are same exactly; actually only 1 test table.

command: pt-table-checksum --user=checksums --password=123456 --no-check-binlog-format --no-check-replication-filters --databases test

but only got,

Waiting to check replicas for differences: 0% 00:00 remain

Waiting to check replicas for differences: 0% 00:00 remain
…
…

seems the issue happened after I clean the table “checksums”; but not reasonable.

Any suggestion? Thanks very much in advance!

carlos.salguero · December 30, 2020, 4:34pm

Could you please include debug infortmation?
Run the command again, adding PTDEBUG=1 in front of the command.

Thanks

Stony · December 30, 2020, 4:34pm

Thanks… I will.

In the slave, I saw the loop,

and I checked the log on slave, it also loop,
171129 8:34:10 88 Query SELECT MAX(chunk) FROM percona.checksums WHERE db=‘test’ AND tbl=’_logs’ AND master_crc IS NOT NULL

171129 8:34:11 88 Query SELECT MAX(chunk) FROM percona.checksums WHERE db=‘test’ AND tbl=’_logs’ AND master_crc IS NOT NULL

171129 8:34:12 88 Query SELECT MAX(chunk) FROM percona.checksums WHERE db=‘test’ AND tbl=’_logs’ AND master_crc IS NOT NULL

171129 8:34:13 88 Query SELECT MAX(chunk) FROM percona.checksums WHERE db=‘test’ AND tbl=’_logs’ AND master_crc IS NOT NULL

171129 8:34:15 88 Query SELECT MAX(chunk) FROM percona.checksums WHERE db=‘test’ AND tbl=’_logs’ AND master_crc IS NOT NULL

171129 8:34:16 88 Query SELECT MAX(chunk) FROM percona.checksums WHERE db=‘test’ AND tbl=’_logs’ AND master_crc IS NOT NULL

…

and the table percona.checksums has nothing.

Stony · December 30, 2020, 4:34pm

The reason is checksums is empty in slave. how to fill checksums in slave?
should I replicant from master? after I did that, pt-table-checksum can work, but never find any difference!!

pls tell why slave doesn’t fill the table checksums…

carlos.salguero · December 30, 2020, 4:34pm

Could you try the Toolkit 3.0.5?
What’s the binlog_format?

Stony · December 30, 2020, 4:34pm

Thanks carlos.salguero,

binlog_format is ROW.

I have a question, does the database perconaneed be added into Replicate_Do_DB to sync table checksums?

just like blew in slave,

Slave_IO_Running: Yes

Slave_SQL_Running: Yes

Replicate_Do_DB: servicesdb,percona

And I saw error, not sure if that is the reason,

OptionParser:1915 16346 pt-table-checksum 3.0.4

OptionParser:2255 16346 Cannot open /etc/percona-toolkit/percona-toolkit.conf: No such file or directory

OptionParser:2255 16346 Cannot open /etc/percona-toolkit/pt-table-checksum.conf: No such file or directory

OptionParser:2255 16346 Cannot open /home/stonyz/.percona-toolkit.conf: No such file or directory

OptionParser:2255 16346 Cannot open /home/stonyz/.pt-table-checksum.conf: No such file or directory

carlos.salguero · December 30, 2020, 4:34pm

That’s just a warning. Nothing to care about.
All toolkit programs try to read default config from several locations just in case you have set defaults to avoid typing all parameters every time.

More info: https://www.percona.com/doc/percona-toolkit/LATEST/configuration_files.html