Group has been broken when switchover duration is over several second

[yoku0825]

As of 0.1.2 switchover behaves as following.

1. Turn on read_only and offline_mode at master
2. Find candidate slave.
3. Waiting for synchronizing master and slave using WAIT_UNTIL_SQL_THREAD_AFTER_GTIDS
4. RESET SLAVE and turn off read_only at candidate slave.
5. Reconfigure replication of other slaves (including old-master)

After 0.0.7, mikasafabric detects FAULTY when server is offline_mode. Because of this, (old)master would be set FAULTY when duration between 1. and 5. is greater than FailureDectator's setting.

After 0.0.14, mikasafabric detects replication-breaking. Because of this behavior, other slave status would be set SPARE when duration between 1. and 5. is greater than FailureDectator's setting.

Add turning off FailureDectator before step 1. and add turning FailureDectator to previous state after 5.

[yoku0825]

• Effects:
  • old-master state would be FAULTY after switchover. Should be set it SPARE and SECONDARY manually (After 0.0.7)
  • Breaking status over the whole one group. (After 0.0.14)
• Workaround:
  • Executing mysqlfabric group deactivate before switchover and executing mysqlfabric group activate after switchover.

[yoku0825]

Fixed by 0.2.0