Should replication lag be calculated from heartbeat metrics?

[genuss]

Currently I use mongodb_consistent_backup as a backup tool for my config-server replica set. In this case updates are very rare. The tool calculates lag using last oplog operation time. In my case this often causes false-positive results of too-high replication lag. Will it be more consious to use lastHeartBeat* metrics or am I missing something here?
As an example please see attached metrics from one of my secondary replica:

Note the oplog lag goes up to 10 seconds although heartBeat doesn't exceed 2 seconds (my heartbeatInterval)

[timvaillancourt]

Interesting question @genuss, I never thought of that.

Are you proposing we subtract the time since lastHeartBeat from the seconds of lag? That seems reasonable to me.

[genuss]

Yes, that's indeed what I'm proposing.

[timvaillancourt]

This logic was adopted in 1.0.0. We now consider the "operational lag" of heartbeats in lag calculations. Thanks for you suggestion!

https://github.com/Percona-Lab/mongodb_consistent_backup/releases/tag/1.0.0.

[genuss]

Very nice! Thank you!