What is the reason for ejabberd Crash when paired with Redis? - tcp

Question

I have been able to scale to more than a million users using
How to scale ejabberd Server machine on CentOS to handle 200 K connections?
I am now using Redis as a backend for ejabberd.
After a million users, I am getting the following error:
2016-04-15 12:50:24 =ERROR REPORT====
** State machine <0.24986.34> terminating
** Last event in was {xmlstreamelement,{xmlel,<<"iq">>,[{<<"type">>,<<"set">>},{<<"id">>,<<"820919">>}],[{xmlel,<<"bind">>,[{<<"xmlns">>,<<"urn:ietf:params:xml:ns:xmpp-bind">>}],[{xmlel,<<"resource">>,[],[{xmlcdata,<<"tsung">>}]}]}]}}
** When State == wait_for_bind
** Data == {state,{socket_state,gen_tcp,#Port<0.418817>,<0.24984.34>},ejabberd_socket,#Ref<0.0.36.113046>,false,<<"2087913259">>,undefined,c2s,c2s_shaper,false,true,false,false,[verify_none,compression_none,{protocol_options,<<"no_sslv3">>},{certfile,<<"/opt/ejabberd-15.11/conf/ejabberd.pem">>}],true,undefined,<<"mac52944bec562c9c82eae8e818abdea7e4b">>,<<"ejabberd-benchmark">>,<<>>,{{1460,704802,37025},<0.24986.34>},{pres_t,0},{pres_f,0},{pres_a,0},undefined,undefined,{userlist,none,[],false},unknown,ejabberd_auth_external,{{10,245,32,24},29307},[],active,[],inactive,undefined,undefined,1000,undefined,300,300,true,0,0,<<>>}
** Reason for termination =
** {timeout,{gen_server,call,[ejabberd_redis_client,{request,[[<<"*">>,"2",<<"\r\n">>],[[<<"$">>,"7",<<"\r\n">>,<<"HGETALL">>,<<"\r\n">>],[<<"$">>,"67",<<"\r\n">>,<<"ejabberd:sm:mac52944bec562c9c82eae8e818abdea7e4b#ejabberd-benchmark">>,<<"\r\n">>]]]},20000]}}

Answer 1

This error simply means that your system is saturated. It does not mean anything by itself.
You have to analyse to find bottlenecks for your use case, tune the platform and possibly optimize the code.