Tabletmanager tests cases migrated from Python

[ajeetj]

Adding/Fixing test cases that were flaky before.
Signed-off-by: Ajeet jain ajeet@planetscale.com

[ajeetj]

@deepthi Need your help to debug why TestIgnoreHealthError is failing here.

The structure of the test is like this:
We will have a NOT_SERVING tablet.
We then execute vtctl cmd IgnoreHealthError zone1-slaveuid .*no slave status.*
Ideally, this should ignore the slave status and show us that the tablet is SERVING
Which is not working for me.

The difference between Python and GO is:
In Go, I have shutdown the mysql forcing tablet to go in NOT_SERVING mode. I have also tried it by shutting down the master.
In Python we are only starting a replica. And without a master the replica will remain in NOT_SERVING mode.

Code pointer: https://github.com/planetscale/vitess/blob/tal_tabletmanager_4/go/test/endtoend/tabletmanager/tablet_health_test.go#L249

[deepthi]

@deepthi Need your help to debug why TestIgnoreHealthError is failing here.

The structure of the test is like this:
We will have a NOT_SERVING tablet.
We then execute vtctl cmd IgnoreHealthError zone1-slaveuid .*no slave status.*
Ideally, this should ignore the slave status and show us that the tablet is SERVING
Which is not working for me.

The difference between Python and GO is:
In Go, I have shutdown the mysql forcing tablet to go in NOT_SERVING mode. I have also tried it by shutting down the master.
In Python we are only starting a replica. And without a master the replica will remain in NOT_SERVING mode.

Code pointer: https://github.com/planetscale/vitess/blob/tal_tabletmanager_4/go/test/endtoend/tabletmanager/tablet_health_test.go#L249

If mysql is down, you will get a different error (probably CRServerGone or CRServerLost), not no slave status.
https://github.com/vitessio/vitess/blob/master/go/vt/mysqlctl/replication.go#L219
https://github.com/vitessio/vitess/blob/master/go/vt/mysqlctl/query.go#L34

[ajeetj]

@deepthi Need your help to debug why TestIgnoreHealthError is failing here.
The structure of the test is like this:
We will have a NOT_SERVING tablet.
We then execute vtctl cmd IgnoreHealthError zone1-slaveuid .*no slave status.*
Ideally, this should ignore the slave status and show us that the tablet is SERVING
Which is not working for me.
The difference between Python and GO is:
In Go, I have shutdown the mysql forcing tablet to go in NOT_SERVING mode. I have also tried it by shutting down the master.
In Python we are only starting a replica. And without a master the replica will remain in NOT_SERVING mode.
Code pointer: https://github.com/planetscale/vitess/blob/tal_tabletmanager_4/go/test/endtoend/tabletmanager/tablet_health_test.go#L249

If mysql is down, you will get a different error (probably CRServerGone or CRServerLost), not no slave status.
https://github.com/vitessio/vitess/blob/master/go/vt/mysqlctl/replication.go#L219
https://github.com/vitessio/vitess/blob/master/go/vt/mysqlctl/query.go#L34

Got it. I have replicated the Python test case scenario and now the tests are passing.

[deepthi]

Does this mean that we are not testing the IgnoreHealthError functionality? Is there a different test case that covers it?

[ajeetj]

That's just github's view. They usually mess up if the same signature method is added up/down.
The TestIgnoreHealthError is still present at line 249.

[deepthi]

Ah ok

[deepthi]

LGTM