Bug #85966 Expose, on slaves, counters for monitoring parallel information quality.
Submitted: 16 Apr 22:20 Modified: 18 Apr 12:14
Reporter: Jean-François Gagné Email Updates:
Status: Verified Impact on me:
None 
Category:MySQL Server: Replication Severity:S4 (Feature request)
Version:5.7 and 8.0 OS:Any
Assigned to:

[16 Apr 22:20] Jean-François Gagné
Description:
Hi,

in a previous feature request (Bug#85965), I asked for adding global statuses on the master for monitoring parallel information quality.  This feature request is the twin sister of that other feature request but for slaves.

Having the information from the master is not always enough to get a good sense of what is happening on slaves.  Some situations popping to mind are:

-delayed slaves,
-catching up slaves (lag decreasing),
-or struggling to keep up slaves (lag increasing).

In those situations, mapping the current position of the slave in the metric (history) of the master is not easy.  Having the value of the corresponding master counters but as currently seen by the SQL thread would be useful.

This is the equivalent of shifting the metric from the master by the actual lag of the slave.  This could be done by good monitoring tools, but it assumes that the counters were previously collected on the master, that we have a powerful monitoring tool, ...  This is unrealistic in the general case, so having those counters on slaves is much better.

Note that those counters cannot be global statuses: as a slave can have many masters (multi-source), those counters should be replication channel-specific.  Exposing those in SHOW SLAVE STATUS or in the corresponding Performance Schema tables would do.

Many thanks,

JFG

How to repeat:
Not a bug but a feature request.

Suggested fix:
Implement new counters in SHOW SLAVE STATUS or in the corresponding Performance Schema tables.
[18 Apr 12:14] Umesh Shastry
Hello Jean,

Thank you for the feature request!

Thanks,
Umesh