zzydxm

By the way we have tested socket NIF version ktls dist, it adds 5% more global CPU usage to our prod node compare to the inet_drv version

> > By the way we have tested socket NIF version ktls dist, it adds 5% more global CPU usage to our prod node compare to the inet_drv version >...

Thanks for the comments! I will work on the changes. On our most major service node (which handles client connection, and message/request routing to 25k other hosts), CPU utilization is...

> The only improvement of your `socket` based `inet_tls_dist` module I can see immediately is that `receiver_loop/6` can avoid flattening all data into a binary when parsing the length. Now,...

> @zzydxm: I am working on an update with my suggested changes above - it became many alterations to set_ktls. Will try to push a commit tomorrow (CEST) Oh really...

These changes looks good to me, thanks a lot for the work!

> =NOTICE REPORT==== 11-Aug-2022::10:42:00.556270 === TLS server: In state start at tls_record.erl:564 generated SERVER ALERT: Fatal - Unexpected Message {unsupported_record_type,0} @RaimoNiskanen Do you still see this? I can't reproduce this...

> We are thinking about merging; have you (@zzydxm) got more tuning to do before that? No I didn't do any more tuning on this. Thanks a lot for the...

Is there any updates on this? Thanks!

Yes I have tested, doing gen_statem:stop(Server, {shutdown, closed}, infinity) should solve the problem.