Memory usage tripled after upgrade from 0.14.12 to 0.14.13

[janfreymann]

Hi,
we use KPL in production on ECS to write to Kinesis. Since we upgraded from 0.14.12 to 0.14.13, memory consumption under load tripled.

This is ECS max memory usage in Grafana, the upgrade was deployed on 9/14:

Did other people experience this behavior, too? We tested the upgrade in isolation without any other changes.

Any help or bugfixes highly appreciated :) Until then, we downgrade to 0.14.12.

Thanks!

[dennisTS]

Observed the same situation with a couple of services as well. So far our only solution was to downgrade - I am surprised it is not yet fixed

[eankit]

Hey, @janfreymann have you ever faced an issue where memory is retained by the kinesis_producer daemon process even after processing all the events?

[SHession]

Just adding to this issue, we are seeing memory usage problems after upgrading from 0.14.12 to 0.15.5. This caused our instances to encounter out of memory exceptions over the weekend. We are downgrading back to 0.14.12 until this is resolved.