Details
-
New Feature
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Fixed
-
24.02.0
-
None
-
MXS-SPRINT-210, MXS-SPRINT-211, MXS-SPRINT-212, MXS-SPRINT-213, MXS-SPRINT-214
Description
Currently the monitor does not detect failures where server is connectable and responds to read queries but writes hang (often due to storage-related errors). These could be detected by periodically doing a small write (need not be binlogged, avoids advancing gtid) and checking if the write goes through and result is readable. If the write fails or hangs, the server should be considered down and failovered. Will require several settings, possibly as follows:
failover_if_master_fails_write_test: Is the feature on? The other settings can have defaults. This could be number indicating how many tests must fail.
write_test_interval: How often is the test made? Test can be omitted if master gtid is clearly advancing from client writes.
write_test_table: Where should MaxScale try to write?