Random DNS problems during Docker build

[lidel]

While looking into #370, I noticed that alpine images are still non-deterministic when it comes to dns lookups (commands executed on the same box, one after another):

> docker run -it --rm node:9.5.0-alpine ping -c 1 nodejs.org
PING nodejs.org (104.20.23.46): 56 data bytes
64 bytes from 104.20.23.46: seq=0 ttl=59 time=21.822 ms

--- nodejs.org ping statistics ---
1 packets transmitted, 1 packets received, 0% packet loss
round-trip min/avg/max = 21.822/21.822/21.822 ms

$ docker run -it --rm node:9.5.0-alpine ping -c 1 nodejs.org
ping: bad address 'nodejs.org'

$ docker run -it --rm node:9.5.0-alpine ping -c 1 nodejs.org
PING nodejs.org (104.20.22.46): 56 data bytes
64 bytes from 104.20.22.46: seq=0 ttl=59 time=22.924 ms

--- nodejs.org ping statistics ---
1 packets transmitted, 1 packets received, 0% packet loss
round-trip min/avg/max = 22.924/22.924/22.924 ms

$ docker run -it --rm node:9.5.0-alpine ping -c 1 nodejs.org
ping: bad address 'nodejs.org'

what the..

If i force ipv4 lookups (ping -4) then it always succeeds.
Seems that the problem surfaces only on boxes that default to ipv6 (like my fiber).

ps. this might be underlying source of related issues:

• generated /etc/resolv.conf in containers not compatible with alpine:3.4 rancher/rancher#5041 (comment)
• Not resolving using search domain gliderlabs/docker-alpine#8 (comment)
• Doesn't work well on a flaky connection yarnpkg/yarn#746

TL;DR

Why is this a problem? We want reproducible builds to work not only on Jenkins, but also on local machines (eg. on laptop of AMO reviewer).

A fix is to find workaround for alpine, or switch to glibc-based image.

[ghost]

A glibc-based image will allow you to run binaries from dist.ipfs.io (go-ipfs, etc.)

[lidel]

Closed by #372