The log line has a well defined structure that we can process by simple
string manipulation, instead of relying on regex.
name old time/op new time/op delta
_handleLogs-12 17.3µs ±23% 1.1µs ±11% -93.81% (p=0.000 n=10+10)
name old alloc/op new alloc/op delta
_handleLogs-12 20.1kB ± 0% 4.1kB ± 0% -79.59% (p=0.002 n=8+10)
name old allocs/op new allocs/op delta
_handleLogs-12 141 ± 0% 1 ± 0% -99.29% (p=0.000 n=10+10)
handleLogs uses bufio scanner to process log output from envoy. When in
debug mode, envoy produces very long log line, causing the scanner
fails, handleLogs stop processing log. But envoy continue writing to its
stdout, which is now not consumed by any process, envoy hangs there
forever.
Fixing this by switching to use bufio.Reader instead. This is also the
real fix for failed integration test, which is interpreted wrongly by
me in #910.
- Update envoy bootstrap config to protobufs
- Reorganize tracing config to avoid cyclic import
- Push down zipkin config to Envoy
- Update tracing options to provide sample rate
* envoy: support autocert
* envoy: fallback to http host routing if sni fails to match
* update comment
* envoy: renew certs when necessary
* fix tests
