It is quaite easy to implement a configurable timeout when connecting to the server: instead of this, in func New(opts):
// try all available URLs in a loop and quit as soon as it
// can successfully establish a connection to one of them
for _, url := range opts.URLs {
if opts.UseTLS {
tlsConfig := &tls.Config{}
if opts.SkipVerifyTLS {
tlsConfig.InsecureSkipVerify = true
}
ac, err = amqp.DialTLS(url, tlsConfig)
} else {
ac, err = amqp.Dial(url)
}
if err == nil {
// yes, we made it!
break
}
}we might do something like this:
config := amqp.Config{
Dial: func(network, addr string) (net.Conn, error) {
conn, err := net.DialTimeout(network, addr, opts.ConnectionTimeout)
if err != nil {
return nil, err
}
// Heartbeating hasn't started yet, don't stall forever on a dead server.
// A deadline is set for TLS and AMQP handshaking. After AMQP is established,
// the deadline is cleared in openComplete.
if err := conn.SetDeadline(time.Now().Add(opts.ConnectionTimeout)); err != nil {
return nil, err
}
return conn, nil
},
}
if opts.UseTLS {
config.TLSClientConfig = &tls.Config{}
if opts.SkipVerifyTLS {
config.TLSClientConfig.InsecureSkipVerify = true
}
}
ac, err = amqp.DialConfig(url, config)This change entains adding ConnectionTimeout time.Duration to the options, declaring a new DefaultConnectionTimeout of 30s, and handling the lack of a sensible value in the options inside func applyDefaults().
Please let me know if you're interested in a pull request; you can check it out in my fork, in the feature/replace-logger-with-slog branch (where I'm also playing with replacing the passed in logger with log/slog - to be proposed in a separate issue).
It is quaite easy to implement a configurable timeout when connecting to the server: instead of this, in
func New(opts):we might do something like this:
This change entains adding
ConnectionTimeout time.Durationto the options, declaring a new DefaultConnectionTimeout of 30s, and handling the lack of a sensible value in the options insidefunc applyDefaults().Please let me know if you're interested in a pull request; you can check it out in my fork, in the
feature/replace-logger-with-slogbranch (where I'm also playing with replacing the passed in logger with log/slog - to be proposed in a separate issue).