Spurious 30 minutes timeout on Travis CI

[kennytm]

Symptom: The build just stops in the middle without any reason, and then got killed due to 30-minutes timeout.

Previous instances:

Time	PR	Log	Last logged action
2018-11-20T07:24:15Z	#55663 (comment)	dist-mipsel64-linux	Compile stage0-rustc
2018-11-14T07:08:11Z	#55943 (comment)	arm-android	Download docker cache
2018-11-13T12:22:23Z	#55912 (comment)	dist-i686-freebsd	Clone submodules
2018-11-13T11:48:10Z	#55672 (comment)	dist-x86_64-linux	Clone submodules
2018-11-13T07:04:10Z	#55912 (comment)	dist-mips-linux	(not even booted up)
2018-11-13T06:32:05Z	#55136 (comment)	i686-gnu	Test liballoc
2018-11-07T22:06:07Z	#55717 (comment)	dist-armhf-linux	Build cargo
2018-11-07T09:14:09Z	#55717 (comment)	x86_64-apple	Test debuginfo
2018-11-07T07:14:11Z	#55510 (comment)	arm-android	Download docker cache

[alexcrichton]

Another instance:

• Replace the ICEing on const fn loops with an error #56065 (comment)
• https://travis-ci.org/rust-lang/rust/jobs/457885473
• last in test of libcollections

[kennytm]

These errors become more serious recently (probably just being masked by worse errors before).

Time	PR	Log	Duration	Last action
2019-01-03T09:38:05Z	#55986 (comment)	x86_64-gnu-aux	2h18m19s	Test cargo
2019-01-11T06:48:10Z	#57503 (comment)	x86_64-gnu-full-bootstrap	37m40s	Compile stage0-rustc
2019-01-12T02:00:14Z	#51487 (comment)	i686-apple	42m01s	Compile stage0-rustc
2019-01-12T06:26:10Z	#57532 (comment)	x86_64-gnu-full-bootstrap	1h47m14s	Compile stage2-rustc
2019-01-12T07:06:07Z	#57532 (comment)	x86_64-gnu-llvm-6.0	36m21s	Compile stage0-rustc
2019-01-15T04:06:21Z	#56044 (comment)	dist-s390x-linux	30m00s	(not even booted up)
2019-01-22T04:14:10Z	#56221 (comment)	x86_64-apple	35m27s	make prepare
2019-01-29T16:58:10Z	#57808 (comment)	arm-android	32m26s	Download Docker cache
2019-01-31T22:56:08Z	#58024 (comment)	dist-x86_64-apple-alt	39m12s	make prepare
2019-02-05T05:14:11Z	#57851 (comment)	dist-x86_64-apple-alt	39m24s	make prepare
2019-02-05T14:38:15Z	#58162 (comment)	dist-x86_64-linux-alt	31m04s	Download Docker cache
2019-02-08T13:34:07Z	#58294 (comment)	dist-x86_64-apple	39m19s	make prepare
2019-02-08T14:16:11Z	#58161 (comment)	dist-x86_64-apple	39m25s	make prepare
2019-02-12T15:52:09Z	#58341 (comment)	dist-android	41m12s	Download Docker cache
2019-02-15T15:46:09Z	#58416 (comment)	dist-x86_64-apple-alt	39m33s	make prepare
2019-02-15T16:20:08Z	#58470 (comment)	dist-android	31m31s	Download Docker cache
2019-02-15T21:48:17Z	#58495 (comment)	arm-android	31m00s	Download Docker cache
2019-02-15T22:22:07Z	#58416 (comment)	arm-android	31m13s	Download Docker cache
2019-02-18T02:56:08Z	#58543 (comment)	dist-x86_64-apple	39m23s	Clone submodules
2019-02-19T06:44:08Z	#58566 (comment)	x86_64-gnu-llvm-6.0	30m44s	Clone submodules
2019-02-19T07:20:10Z	#56470 (comment)	test-various	32m41s	make prepare
2019-02-19T19:06:40Z	#58537 (comment)	dist-various-2	42m32s	Download Docker cache
2019-02-20T06:34:08Z	#58056 (comment)	arm-android	53m57s	Compile stage0-rustc
2019-02-20T07:10:10Z	#58578 (comment)	x86_64-gnu	30m58s	Clone rust-lang/rust
2019-02-21T10:20:10Z	#58597 (comment)	dist-i586-gnu-i586-i686-musl	1h50m56s	Test liballoc
2019-02-21T11:10:13Z	#58535 (comment)	dist-i686-apple	48m12s	Compile stage0-rustc
2019-02-21T11:44:13Z	#58597 (comment)	dist-android	31m55s	Download Docker cache
2019-02-21T20:52:20Z	#58592 (comment)	dist-android	33m19s	Download Docker cache
2019-02-21T21:28:13Z	#58592 (comment)	dist-x86_64-linux	31m53s	Download Docker cache
2019-02-22T10:08:08Z	#57364 (comment)	dist-x86_64-linux	32m57s	Download Docker cache

[Mark-Simulacrum]

I'm wondering if we could add the "print building - timestamp" script at the top-level. We do it for make (e.g., curl/gcc/openssl builds) but it seems like it at least would let us figure out if this is literally the builder "stopping" or if we're hanging in rustc...

[kennytm]

#57882 (comment)

Adding -y 30 -Y 10 --connect-timeout 30 from #58655 does not completely fix the 30 minute timeout when fetching docker cache 😕

[crlf0710]

Now that we're using Azure Pipelines, i guess this can be closed now? @kennytm @Centril