bug: http service should not be started before etcd data is pulled

[hizzd]

Issue description

that when accessing immediately after apisix is started, it will return:

{"error_msg ": "404 Route Not Found"}

Is it because apisix did not pull the route data in etcd in time?
maybe the http service of apisix should be started after pulling etcd data.

Environment

• apisix version: [docker] apache/apisix:2.2-alpine
• OS: [docker] apache/apisix:2.2-alpine
• OpenResty / Nginx version: [docker] apache/apisix:2.2-alpine
• etcd enabled auto-tls.
• apisix disable etcd.tls.verify.

[spacewander]

This is a known issue. We plan to start pulling etcd data in the init_worker phase. To achieve this, we need to implement two preconditions first:

1. add Chaos test so that we can cover such a big change
2. allow lua-resty-etcd to use luasocket.

Maybe we can do this in version 2.5.