Question: Lag exporter reports astronomical lag in seconds

[AlexeyRaga]

Hello,

This is not a bug report, but more a question at this time.

I have a new/reprocessing consumer that starts processing topic from the beginning.
Until it is done, kafka-lag-exporter returns astronomically high values for the lag in seconds.

Here is my Grafana dashboard screenshot.

On the first graph I am trying to manually calculate how many seconds it'll take to consume all the messages in the topic by taking the total lag and dividing by the amount of messages processed in a second.

The second one is kafka_consumergroup_group_lag_seconds from kafka-lag-exporter, and I tried to find a coefficient that would make it look similar to my manually constructed one.

It looks like that the lag time that kafka-lag-exporter returns is ~200000 times bigger than the manually calculated one!
And, according to the graph, the manually calculated one was pretty close to reality.

What am I missing here? Why the time lag reported by kafka-lag-exporter can be such a huge value? And how to fix it?

[AlexeyRaga]

Here is what I see in the bundled dashboard.
Looks like kafka-lag-exporter thinks that it'd take 3+ years to process 45 min of data...

And it doesn't correct itself while burning down the lag.