[WIP] Swap 1.1.1.1 and 8.8.8.8.

[kennytm]

r? @ghost

Just to check if this could workaround Travis issue 9696.

[kennytm]

@bors try

[bors]

⌛ Trying commit e585ad2 with merge 063468c8cd00ce5179ba93b1064e905ede0d369d...

[bors]

☀️ Test successful - status-travis
State: approved= try=True

[kennytm]

@bors retry

[bors]

⌛ Trying commit e585ad2 with merge 3b713cc85f786be9dc2bacf787bedf3faf7e922c...

[kennytm]

Note: according to #51762 (comment), a 9696 happened at least between 07:05:04Z to 07:09:17Z.

[bors]

☀️ Test successful - status-travis
State: approved= try=True

[kennytm]

CI	Time	1.1.1.1	1.0.0.1	8.8.8.8	8.8.4.4
PR-CI 1	06:59:30	52.219.28.41	52.219.28.41	52.219.24.13	52.219.24.13
PR-CI 2	06:59:30	52.219.28.41	52.219.28.41	52.219.24.41	52.219.24.97
First try 1	07:00:15	52.219.28.33	52.219.28.33	52.219.28.41	52.219.24.1
First try 2	07:00:16	52.219.28.33	52.219.28.33	52.219.24.1	52.219.24.1
Second try 1	07:18:38	54.231.236.45	54.231.236.45	52.219.20.21	52.219.24.129
Second try 2	07:18:38	54.231.236.45	54.231.236.45	52.219.20.21	52.219.20.1

Observations:

• The answers from 1.1.1.1 and 1.0.0.1 seem to be consistent with the same IP, while that from 8.8.8.8 and 8.8.4.4 can produce totally different answers with a few seconds difference (the TTL of this domain is extremely short though).
• A 9696 happened between the two try builds. With 1.1.1.1 and 1.0.0.1 we see the IP jumped across an entire /10 net, while 8.8.8.8 still returns 52.219.*.*.

I'll try doing a continuous dig and curl for 30 minutes tomorrow, closing for now.

[pietroalbini]

Hmm, should we merge the digs into master so we're sure we catch the results during the 9696?

[kennytm]

@pietroalbini In https://api.travis-ci.org/v3/job/398515215/log.txt, the network access before 07:05:04Z is completely fine (all crates are successfully vendored at 06:08:52Z), so adding a dig at the beginning may not be useful.

[pietroalbini]

@kennytm we can also add it in the script section after the main script is called, that's executed as soon as the main script exits (so as soon as there is a resolve error)

[kennytm]

We could add a dig check in after_failure though the behavior outside and inside docker may be different. For instance the curl -fs --head https://google.com in the log above works fine.