Description:
Related to bug#106526 we notice there is very little information related to instrumentation of group replication "topology changes". This makes it hard to identify issues as referenced in bug#106526 and thus to more easily identify the cause (and hopefully resolve it).
How to repeat:
Try to identify information about the group and see that a lot of useful information may be missing.
Suggested fix:
Provide something like the following instrumented values:
- timestamp of group creation
- timestamp of group startup
- timestamp of last group state change
- type of last group change
- time to achieve last group change (latency from begin to end)
- number of changes in the group state (not transaction but "topology changes")
- sum_of_group_change_latency
- min_group_change_latency
- max_group_change_latency
- group_member_timestamp_joined_group
- group_member_timestamp_of_last_member_state_change (I can imagine joined, change between multi-write / single write, or from secondary to primary)
- group_member_version_supported (I think this is already provided somewhere)
- group_member_version_used (I think this is already provided somewhere and probably common to all group members)
This extra information, or something like it would I think be useful for identifying "group topopology/state changes". That information is partly visible now but only via logs so hard(er) to query or monitor externally.