DNS probing performance regression in 0.19.0

[yegle]

I run blackbox-exporter in docker probing a DNS server in another docker container.

With cache on DNS server end, probing latency is usually ~120us at 0.18.0. After updating to 0.19.0 it jumps to ~450us.

Knowing my patch in #724 I updated my config with preferred_ip_protocol: "ip4" since I don't configure ipv6 in my docker network, but it still result in ~200us latency.

Reverting back to 0.18.0 and now I have consistent ~120us latency again.

[karlism]

I'm seeing similar degradation in my environment too, blackbox_exporter was updated on May 19th and since then probe_dns_duration_seconds{phase="resolve"} and probe_http_duration_seconds{phase="resolve"} has increased significantly:


All exporter VMs are running local instance of caching dnsmasq so DNS latency should be insignificant.

[SuperQ]

I wonder if this is related to #758

[yegle]

Presumably #758 only affects HTTP probing? It looks like the issue reported here is on DNS probing only.

[roidelapluie]

The issue is with #724

I think it might be related to the fact that we now compute the hash IN the resolve phase.

[roidelapluie]

Can you please test #796 ?

[roidelapluie]

I will merge #796 soon but would still appreciate some testing.