Unstable if flow_meter uses with many plugins in one command line about nemea-modules HOT 3 CLOSED

Hello, thanks for the comment.

1. Use many plugins in one command line (eg: basic,http,dns)
The default timeout on output interfaces is WAIT. This will cause module to stop and wait when trying to send message to output interface with no connected clients. flow_meter -i t:11111,t:11112,t:11113 -p basic,http,dns -I eth1 was waiting for clients to connect to 11111 or 11112 port while you expected flows on 11113 port.

However WAIT as default timeout is not the best choice for exporter. We will change the default timeout to HALF_WAIT to prevent this behavior. You can always set timeout manually.
For more information about timeouts see logger -h trap or flow_meter -h trap

2. Can not auto-kill logger if use unix socket process
I will need more info. What was the output of logger when you killed flow_meter with TCP socket?

I killed flow_meter with TCP socket and logger returned an trap_recv() error if you mean this by auto-kill.

from nemea-modules.

Hi Havraji6,

• I just tested with timeout=HALF_WAIT, it works so well
• I just reconsider my old idea with auto-kill, with old commits, if used command pkill -9 meter and logger also be killed. Today with the same command (pkill -9 meter) and logger is still alive. It is my expected
  Thanks you!

from nemea-modules.

@whitetiger21022014 generally, it should be better to kill NEMEA modules using SIGINT - this usually breaks the main loop in modules...

That auto-kill "feature" you mentioned is something caused by "ending message". E.g. when we use logreplay, it sends this message by default in order to stop connected modules when the CSV file is read till the end.
However, I wish to get so called "signalling" feature finished - it will improve internal mechanism of e.g. this ending message use-case.

Thank you very much for all feedback and notes you have already provided.

from nemea-modules.