timeout on connection to registry

[vohmar]

registrars get epp timeouts. Affects any EPP query.

Theory: epp-proxy timeout threshold is too small so if new passenger process is started for a query epp-proxy bails out too soon and returns timeout to the users while in reality the query is run and result gets created.

Timeout seems to be at 1 or 2 seconds

root@reg-l-epp-wheezy:/etc/apache2# fgrep "Error contacting registry" /var/log/registry.log | wc -l 35

http://localhost/epp/command/poll http://localhost/epp/session/login http://localhost/epp/command/poll http://localhost/epp/command/poll http://localhost/epp/command/check http://localhost/epp/session/login http://localhost/epp/session/login http://localhost/epp/command/update http://localhost/epp/session/login http://localhost/epp/session/login http://localhost/epp/command/update http://localhost/epp/command/update http://localhost/epp/session/login http://localhost/epp/command/renew http://localhost/epp/command/check http://localhost/epp/command/poll http://localhost/epp/command/poll http://localhost/epp/command/renew http://localhost/epp/command/poll http://localhost/epp/command/update http://localhost/epp/command/update http://localhost/epp/command/update http://localhost/epp/command/info http://localhost/epp/command/poll http://localhost/epp/command/check http://localhost/epp/command/poll

[vohmar]

tried changing Passengermaxrequest value from 1000 to 5000 and passengerptestart -i is set. As a result number of timeouts is considerably reduced - about 1 in 15000 requests gets the timeout. Prestart setting seems not to affect anything.

PassengerMaxPoolSize 15 PassengerPreStart http://localhost/ PassengerMaxRequests 5000

we should make the epp-proxy timeout configurable and set it to few seconds longer.