Memory leak: http connections keep opened

[arvatoth-vcc]

Describe the bug
We have got a dkron-3.1.10 inside Kubernetes cluster with single deployment (single pod only). We are using only one http executor which started at every 30 seconds. Every check opens a new http connection and all of previous keep established.

Expected behavior
Close previous established tcp connections or re-use them and do not open new to same endpoint.

Screenshots

** Specifications:**

• OS: Kubernetes v1.21.9 under Ubuntu 18.04 Bionic
• Version dkron-3.1.0

Additional context
Lot of established tcp connections:

Running inside the pod:

Thank you!

[vcastellm]

Under my observations the number of open connections with hundreds of http jobs, the connections reach a stable point. How many jobs are you trying to run?

[arvatoth-vcc]

Only one job with one http executor. I think http executor sends Connection: keep-alive HTTP requests, but not reuse that.

[wuxingzhong]

I ran into the same problem where dkron would not release the connection causing server memory to overflow

[vcastellm]

Can you confirm if this still happens in latest version?

[arvatoth-vcc]

I have just refreshed version to 3.2.2, I can refer in a few days.

[arvatoth-vcc]

Unfortunately version upgrade did not solve the memleak problem.

[wuxingzhong]

I set idletimeout on the server side to block this problem for now. I think this is a bug in dkron

[vcastellm]

Closing this because the workaround was merged