pt-table-checksum --check-slave-lag docs aren't clear

[Fernando Ipar]

The documentation implies that if --check-slave-lag is used, pt-table-checksum will only check the lag for the specified slave.
However, the tool still tries to recurse down the replication hierarchy.

Is this expected behavior?
I understand --check-slave-lag as meaning "I only care about lag on this one slave, please just monitor that one", is that wrong?

[Baron Schwartz (baron-xaprb)]

When --check-slave-lag is used, pt-table-checksum should only connect to the specified server and shouldn't interact with any others at all. So this is a bug.

[Baron Schwartz (baron-xaprb)]

The same problem occurs in pt-online-schema-change. That is reported in bug 1012681, which I have marked as a duplicate of this one.

[Daniel Nichter (daniel-nichter)]

Baron, I think --check-slave-lag means "only check lag on this slave", but that's separate from which slaves the tool connects to for other things like waiting for chunks to replicate. By default, all slaves' lag is checked, but if there's one that's always lagging or intentionally being lagged, then this option allows one to specify which slave is the "standard lagger". Using --recursion-method=dsn is the way to make the tool only connect to certain slaves.

So I don't think this is a bug; the tool has always worked this way concerning this option.

[Daniel Nichter (daniel-nichter)]

Fernando agrees that this isn't a bug but rather the docs not reading clearly.

[Shahriyar Rzayev (rzayev-sehriyar)]

Percona now uses JIRA for bug reports so this bug report is migrated to: https://jira.percona.com/browse/PT-499