Nameserver keeps 'old' agents which died or got killed

[one4many]

The nameserver currently keeps agents in his database/storage even when they have died or been killed (with SIGTERM for example).

I have the current workaround implemented:

from osbrain import run_nameserver
import time, sys, traceback

NAMESERVER='10.10.10.42:42422'

def main():
    defunct_agents = {}
    ns = run_nameserver(NAMESERVER)
    while True:
        for alias in ns.agents():
            try:
                agent = ns.proxy(alias)
                agent.ping()
            except:
                if alias in defunct_agents:
                    defunct_agents[alias] += 1
                else:
                    defunct_agents[alias] = 1
            for agn in list(defunct_agents):
                if defunct_agents[agn] > 4:
                    try:
                        ns.remove(agn)
                        defunct_agents.pop(agn, None)
                    except:
                        traceback.print_exc(file=sys.stdout)
        time.sleep(5)
    
if __name__ == '__main__':
    main()

A optional argument to ns_runserver (ns_runserver(NAMESERVER, del_defunct_agents_after_seconds=60) for example) would be great.

[Peque]

Thanks for opening this issue.

Although we have that in our road-map, heartbeating and monitoring will probably be left for a post-1.0.0 release. Meanwhile is up to the user to implement any tricks they need for their application. :wink:

[Peque]

Anyway we will probably implement better handling of SIGINT/SIGTERM for 1.0.0, though. The case in which the agent simply "dies" will still need to be handled by the user.