DUNE-DAQ / nanorc

2 stars 2 forks source link

kill services in a try/catch block. Test APP_PORT before startng apps #193

Closed plasorak closed 1 year ago

ron003 commented 1 year ago

I tried running with both nanorc plasorak/connectivity-server-fixes and daqconf plasorak/port-connection-svc -- with this, it seems the default connectivity serv port defaults to 15000. I had an old conn serv running on 5000 and there seemed to be a conflict. I ran the nanorc (non-integ test) and the system failed fairly gracefully (no statck trace). I would like to do more testing.

jcfreeman2 commented 1 year ago

Two additional comments 1) Discussed at this morning's release meeting: while port 15000 is an improvement over 5000, we could reduce the likelihood of port conflict on a system if we didn't make the port number a nice round number. Something more random, e.g. 15489, would generally be safer. 2) It might be nice in the nanorc documentation to describe how to fix a port clog (i.e. netstat -tulpn | grep <port number> to get the PID, then kill <PID>. This would at least work in the event that the user had caused the port clog in the first place, which is what happens, e.g., if one tries running on the np04 cluster without first doing source ~np04daq/bin/web_proxy.sh -u, and then hits Ctrl-c to escape the hang)

jcfreeman2 commented 1 year ago

In fact, concerning (2) Kurt pointed out that in the daqconf Wiki people are instructed to add

[http]
  proxy = http://np04-web-proxy.cern.ch:3128
  sslVerify = false

to their .gitconfig file so they don't have to source ~np04daq/bin/web_proxy.sh -u each time they login; could we also update the https://dune-daq-sw.readthedocs.io/en/latest/packages/nanorc/FAQ/#nanorc-cant-start-the-response-listener-what-do-i-do section of the nanorc documentation to mention this?