LP #918051: chunk size error if you use --resume option

Description

**Reported in Launchpad by Miguel Angel Nieto last update 10-04-2012 14:22:45

I get some errors about chunk size when using the resume option but not always. The example database is "employees":

https://launchpad.net/test-db/

If you don't cancel the checksum it works perfect:

root@debian-sandbox:~/sandboxes/rsandbox_5_1_50# pt-table-checksum --databases=employees u=root,p=msandbox,h=127.0.0.1,P=26165
TS ERRORS DIFFS ROWS CHUNKS SKIPPED TIME TABLE
01-18T08:42:44 0 0 9 1 0 0.009 employees.departments
01-18T08:42:45 0 1 331601 5 0 1.579 employees.dept_emp
01-18T08:42:45 0 0 24 1 0 0.011 employees.dept_manager
01-18T08:42:47 0 0 300024 1 0 1.140 employees.employees
01-18T08:43:00 0 1 2844024 12 0 13.380 employees.salaries
01-18T08:43:01 0 0 443306 1 0 1.297 employees.titles

If you cancel it and try to resume later, sometimes I get this error:

root@debian-sandbox:~/sandboxes/rsandbox_5_1_50# pt-table-checksum --databases=employees u=root,p=msandbox,h=127.0.0.1,P=26165
TS ERRORS DIFFS ROWS CHUNKS SKIPPED TIME TABLE
01-18T08:44:09 0 0 9 1 0 0.010 employees.departments
01-18T08:44:10 0 1 331601 5 0 1.533 employees.dept_emp
01-18T08:44:10 0 0 24 1 0 0.015 employees.dept_manager
^C# Caught SIGINT.
01-18T08:44:11 0 0 300024 1 0 0.606 employees.employees

root@debian-sandbox:~/sandboxes/rsandbox_5_1_50# pt-table-checksum --databases=employees --resume u=root,p=msandbox,h=127.0.0.1,P=26165
01-18T08:44:13 Skipping table employees.employees because on the master it would be checksummed in one chunk but on these replicas it has too many rows:
300024 rows on debian-sandbox
300024 rows on debian-sandbox
The current chunk size limit is 2000 rows (chunk size=1000 * chunk size limit=2.0).
TS ERRORS DIFFS ROWS CHUNKS SKIPPED TIME TABLE
01-18T08:44:26 0 1 2844024 14 0 12.807 employees.salaries
01-18T08:44:26 0 0 443306 1 0 0.532 employees.titles

If I stop the checksum while checksumming other tables, for example departments or salaries, I don't get the error.

Also, if you try to resume with nothing to resume it complains again about the chunk size. The error should be "There is nothing to resume" or something similar:

root@debian-sandbox:~/sandboxes/rsandbox_5_1_50# pt-table-checksum --databases=employees --resume u=root,p=msandbox,h=127.0.0.1,P=26165
01-18T08:45:10 Skipping table employees.titles because on the master it would be checksummed in one chunk but on these replicas it has too many rows:
443306 rows on debian-sandbox
443306 rows on debian-sandbox
The current chunk size limit is 2000 rows (chunk size=1000 * chunk size limit=2.0).

Environment

None

Smart Checklist

Activity

Show:

lpjirasync January 24, 2018 at 6:15 PM

**Comment from Launchpad by: Baron Schwartz on: 10-04-2012 14:22:31

This isn't a bug. It's documented and intended behavior. We are currently discussing it on the mailing list; please contribute your suggestions there and we may include a change to the behavior in a future release:
https://groups.google.com/d/topic/percona-discussion/5zI1-ImxhJA/discussion

lpjirasync January 24, 2018 at 6:15 PM

**Comment from Launchpad by: Philip M. Gollucci on: 10-04-2012 04:38:01

No --resume neccessary.

04-10T04:17:51 Skipping table X.Y because on the master it would be checksummed in one chunk but on these replicas it has too many rows:
162019 rows on H
The current chunk size limit is 154192 rows (chunk size=77096 * chunk size limit=2.0).

Not a Bug

Details

Assignee

Reporter

Priority

Smart Checklist

Created January 24, 2018 at 6:14 PM
Updated February 3, 2018 at 10:19 PM
Resolved January 24, 2018 at 6:15 PM