Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Tasks queued up in Travis #9500

Closed
lannka opened this issue May 23, 2017 · 9 comments
Closed

Tasks queued up in Travis #9500

lannka opened this issue May 23, 2017 · 9 comments

Comments

@lannka
Copy link
Contributor

lannka commented May 23, 2017

screenshot from 2017-05-23 11 03 00

anyway to increase the throughput?

@rsimha
Copy link
Contributor

rsimha commented May 23, 2017

There are a few reasons for this:

  • The setting for auto-cancelling multiple (duplicate) test runs for consecutive commits on one branch had been turned off. I've turned it back on, and this should de-dupe test runs from the same PRs.
  • Now that we're sharding unit tests and integration tests, we're building twice, once on each shard. This will be fixed with Clean up unit tests and move all integration tests to integration/ #9404 and Optimizations to speed up Travis PR builds #9626
  • We're unnecessarily doing a dist for the integration_tests shard on PRs. We only need to do so on master. This will be fixed with Don't do a dist in the integration_tests shard for PRs #9530
  • Prior to sharding, a run with a presubmit error would fail early prior to the full build. Now, the unrelated shards will still be run. This is a case of a shorter wait in the average case, but a potentially longer wait in the worst case. I'm investigating a way to cancel all jobs for a run if one job fails. (Edit: It would actually be preferable not to do this, since it'd be beneficial to see results from pre_build_tests even if integration_tests were to fail, for example)

Of course, if it's feasible, adding more VMs would alleviate the worst case and greatly improve the best case.

@dreamofabear
Copy link

dreamofabear commented May 24, 2017

Addendum to second item above, we should skip building the runtime since integration tests are skipped on PR builds anyways.

@taymonbeal
Copy link
Member

Is there any instrumentation or profiling that tells us how long is spent on each step of the CI pipeline?

@rsimha
Copy link
Contributor

rsimha commented May 30, 2017

@taymonbeal, all the build and test tasks kicked off by pr-check.js are instrumented. Open the logs for each shard to see how long they took.

@taymonbeal
Copy link
Member

Do you need sudo: required? Removing it means Travis doesn't have to spin up VMs. And there's no reason in principle why it shouldn't be possible to install Chrome without root, is there?

@rsimha
Copy link
Contributor

rsimha commented May 31, 2017

/cc @erwinmombay to comment on why we needed sudo, and whether we still do.

From http://docs.travis-ci.com/user/trusty-ci-environment/, there is now beta support for trusty without sudo, although, as you point out, it doesn't spin up VMs, and therefore, doesn't offer the same kind of isolation between consecutive test runs since there might be state that persists between builds.

@taymonbeal
Copy link
Member

It still offers full isolation. It's just that isolation is achieved through Docker containers, not through VMs, so it's faster. That's why I'm suggesting that it be used.

@rsimha
Copy link
Contributor

rsimha commented May 31, 2017

While Docker appears to offer full isolation when it comes to persistent state (you're right), I don't believe it reserves resources like VMs do, and this may affect some of our tests. I'll spin this off into its own issue for further discussion, and if it turns out that we can do without full VMs, I'm open to making the change.

@rsimha
Copy link
Contributor

rsimha commented May 31, 2017

Update: The PR build time (including a gulp build and a full unit test run) is now down to around 15 minutes, with a total CPU time consumption of less than 20 minutes. With this, we can close out this issue.

torch2424 pushed a commit to torch2424/amphtml that referenced this issue Jun 6, 2017
Travis builds have become slow during busy periods of the day due to various reasons. This PR does the following:

Reorganizes build tasks across fewer build shards
Reduces the number of VMs used by PR and master builds to 2
No longer does a full gulp dist for unit tests (Possible to do after ampproject#9404)
Reduces the total CPU time used for PR builds to < 20 mins
Fixes ampproject#9500
Fixes ampproject#9387
torch2424 pushed a commit to torch2424/amphtml that referenced this issue Jun 6, 2017
* Bidtellect amp implementation… (ampproject#9518)

* Bidtellect amp implementation…

* removing renderStartImplemented…

* spaces?

* adding example

* Fix test-amp-ad-network-doubleclick-impl (ampproject#9645)

* Optimizations to speed up Travis PR builds (ampproject#9626)

Travis builds have become slow during busy periods of the day due to various reasons. This PR does the following:

Reorganizes build tasks across fewer build shards
Reduces the number of VMs used by PR and master builds to 2
No longer does a full gulp dist for unit tests (Possible to do after ampproject#9404)
Reduces the total CPU time used for PR builds to < 20 mins
Fixes ampproject#9500
Fixes ampproject#9387

* Animations: full on-action API (ampproject#9641)

* Animations: full on-action API

* review fixes

* Fix Validator tests on Travis, which now uses Node v4 (ampproject#9654)

* Amp-imgur : Implement imgur embed (ampproject#9405) (ampproject#9434)

* Amp-imgur : Implement imgur embed (ampproject#9405)

* amp-imgur test code modified

* amp-imgur validation code changed

* fix amp-imgur lint check

* Remove unused Layour variables

* Test code modified

* Ingegration amp-imgur extension

* [amp-imgur] add resize logic

* Remove 3p iframes and add resize message listener

* Fix some code with reviews
* change some syntax
* change validation rules
* add find event source

* [amp-imgur] remove unnecessary code and add object check

* remove spec_name from validation code

* Remove unnecessary code and add owners file

* Change required tag name  to  in validation file

* data-imgurid -> data-imgur-id

* Add whitelist for amp-imgur

* remove unused AmpImgur import in test file

* remove unused preconnect callback

* Update amp-cors-requests.md (ampproject#9636)

Added updates and corrections per Dima's feedback.

* Expose login url building to amp access service (ampproject#9300)

* Animations: support style keyframes (ampproject#9647)

* Animations: support style keyframes

* docs

* review fixes

* lints

* Animations: subtargets format (ampproject#9655)

* Animations: subtargets format

* fix docs

* Fix extern for integration test (ampproject#9657)

* Remove whitelisted link for PR 9434 (ampproject#9656)

* Add callouts for CORS + cleanup (ampproject#9591)

* Add callouts for CORS + cleanup

* Update amp-access.md

* Doubleclick Fast Fetch: Send default safeframe version on ad request (ampproject#9613)

* Send default safeframe version on ad request

* fix test failure

* amp-ima-video: Fixes some undefined variables in compiled extension (ampproject#9617)

* Add experimental input-debounced to the actions and event doc (ampproject#9650)

* Copy and rename compiled script directly for deprecated version (ampproject#9587)

* Remove A4A Dependency on ads/_config.js (ampproject#9462)

* Added doubleclick specific config

* Modified other networks

* Update MockA4AImpl

* Remove unneeded config lines

* Add cid, remove renderStartImplemented

* Fix lint complaint

* Changed from mandatory config to overrrideable method

* Addressed feedback, modified getAdCid

* Addressed additional feedback

* Responded to feedback

* Removed unnneeded guard against null value

* Changed signature of function

* Fixed lint issues

* addressed comments

* Addressed comments, fixed broken test

* Change signature / update tests

* Fix wrong contents and white spaces in amp-imgur.md (ampproject#9658)

* Performance: separate ini-load from first visible ini-load (ampproject#9665)

* amp-bind: Fix more integration tests (ampproject#9598)

* more bind test fixes

* include object.assign polyfill

* move polyfills to separate file

* replace use of map() from bind-validator with ownProperty()

* fix types and test

* Stop using IOS elastic scroll when slidescroll takes over animation (Only for the NON SNAP-POINTS flow) (ampproject#9668)

* removing overflow touch on no-scroll for ios

* Stop using IOS elastic scroll when slidescroll takes over animation (Only for the NON SNAP-POINTS flow)

* make sure css is compiled before entry points (ampproject#9643)

* update npm package version (ampproject#9671)

* A4A: expose visibilityCsi (ampproject#9667)

* A4A: expose visibilityCsi

* tests

* amp-bind: Fix embedding in FIF (ampproject#9541)

* move element service changes from ampproject#9447

* store element service ids in extension struct

* add unit test, fix other tests

* Fix null value binding (ampproject#9674)

* Add amp-form's submit action to the docs (ampproject#9675)

* Write cookie for ad-cid. (ampproject#9594)

* Write cookie for ad-cid.

* fix tests

* amp-access-laterpay fixes (ampproject#9633)

* Add support for LaterPay subscriptions

* Fix canonical link on laterpay example

* Tweak default styling

* Add an example locale message for the header

* Fix indentation of JsDoc (ampproject#9677)

It should be indented at the same level as the method it's documenting.

* Add example to amp iframe docs (ampproject#9660)

* Change image to static link hosted on ampproject.org

* Update amp-iframe doc w example + gen cleanup

* Adds SFG experiment branches to doubleclick-a4a-config.js. (ampproject#9662)

* Added experiment branches corresponding to exp=a4a:(5|6).

* Removed test file.

* Removed changes to yarn.lock + minor fixes.

* Fixed adsense-a4a-config.js.

* Fixed return statement.

* Removed reference to style tag for opacity (ampproject#9634)

As style tag for `opacity` was replaced with `amp-boilerplate (back in this [commit](ampproject@0a056ca), updated the text to reflect those changes.

* De-flake amp-bind integration tests (ampproject#9683)

* split bind fixtures into one file per extension

* fix width/height binding bug, avoid race condition w/ dynamic containers

* actually, just skip the tests affected by race condition for now

* Remove timer calls to prevent future flakiness (ampproject#9478)

* Remove timer calls to prevent future flakiness

* Revert the line order change

* Poll instead of using stub callback constructor

* Add jsdoc and rename variables

* amp-animation: polyfill partial keyframes from CSS (ampproject#9689)

* Animations: whitelist offset distance and regroup condition types (ampproject#9688)

* Use Docker containers in Travis (ampproject#9666)

* clean up web-worker experiment (ampproject#9706)

* Load examiner.js when #development=2 (ampproject#9680)

* Load examiner.js when #development=1

* address comments

* Use development=2

* Fix presubmit

* Implementation of amp-ad-exit (ampproject#9390)

* Implementation of the amp-ad-exit component for managing ad navigation (ampproject#8515).

Changes from the I2I:
- Only RANDOM, CLICK_X, and CLICK_Y variables can be used, alongside custom vars.
- It doesn't work well with <a> tags - the tap action doesn't trigger on middle clicks. Recommendation is to use other elements.
- ClickLocationFilter for border click protection is not yet implemented. It will be added in a future change.

The component does some lightweight validation of the JSON config when it's
built. This may be overkill.

Tested:
gulp test --files 'extensions/amp-ad-exit/0.1/test/*'
gulp check-types

* Update JSDoc for some functions. Fix lint issue (ignore unused parameters in an interface method).

* sinon.spy -> sandbox.spy

* Add amp-ad-exit to the list of allowed extensions in A4A pages and fix validation rules.

Address PR: add experiment flag; use camelCase for config properties

address some comments

Instantiate a new Filter for each spec.

* Instantiate a new Filter for each spec.

* remove duplicate doctype tag in example

* Make ActionEventDef public for type annotations.

* Add filter factory.

* fix event var

* update type annotation for Filter.filter

* address PR comments

* Move filter config validation to the ctor. Remove Filter.buildCallback().

* camelCase in documentation examples

* Introducing "it.yield", a convenient way to test promise code (ampproject#9601)

* Introduce yield.

* Address comments.

* fix done state error handling

* Document amp-access as a special target (ampproject#9568)

`amp-access` is a special target that's not mentioned in this document.
It's special because you can't give an arbitrary ID to it, i.e., you
can't just change the id of the `amp-access` script tag to `amp-access2`
and expect `on="tap:amp-access2.login"` to work. Given that the
"actions" for `amp-access` is a bit dynamic depending on the structure
of the `amp-access` JSON, instead of listing out the actions in a
tabular form, a reference to the `amp-access` documentation is added.

* Run presubmit tests soon after building (ampproject#9716)

* Revert "Use Docker containers in Travis (ampproject#9666)" (ampproject#9717)

This reverts commit 98ecb9b.

It turns out that while using Docker containers instead of VMs on Travis reduces VM startup time, it has significantly increased build and test time, increasing the overall PR check round trip from < 20 mins to > 30 mins.

Compare https://travis-ci.org/ampproject/amphtml/jobs/238464404 (Docker) against https://travis-ci.org/ampproject/amphtml/jobs/238462066 (VM).

Fixes ampproject#9651

* Validator Rollup (ampproject#9719)

* Add validator tests for amp-imgur.

* Disallow amp-embed as child of amp-app-banner.

* Minor cleanup, Make sure our set numbers are consistent in javascript.

* amp-state: Allow both `src` and script child (ampproject#9721)

* adapt code from ampproject#8897

* also_requires_attr should be in trigger

* Remove experiment for input-debounced. Update docs. (ampproject#9724)

* Validator Rollup (ampproject#9727)

* Minor cleanup of amp-ad-exit

* Disallow amp-ad/amp-embed as children of amp ad containers when data-multi-size is present.

* Revision bump

* Enabling dynamic queryparam addition to anchour links  (ampproject#9684)

* Enabling queryparam addition to anchour links

* Code review changes

* Additional cache urls (ampproject#9733)

* Significantly speed up gulp build --css-only (ampproject#9726)

Prior to this PR, gulp build --css-only would take ~ 1 minute to run, and end up building a bunch of js files in addition to compiling css.

After this PR, gulp build --css-only takes ~ 0.5 seconds to run (a 100x speed up), and the only output files in build/ after running it are .css, and .css.js files.

Resulting speeds on my workstation:
gulp build --css-only: ~500ms
gulp build: ~1m 45s
gulp dist --fortesting: ~3m 30s

Fixes ampproject#9640

* add amp-analytics Facebook Pixel support (ampproject#9449)

* add amp-analytics Facebook Pixel support

* add amp-analytics Facebook Pixel support

* fix test on extensions/amp-analytics/0.1/test/vendor-requests.json

* add Facebook Pixel standard events https://developers.facebook.com/docs/ads-for-websites/pixel-events/v2.9#events

* Update vendors.js

fix conflicts

* Update vendors.js

fix conflicts

* Update vendor-requests.json

fix conflicts

* Update vendor-requests.json

fix conflicts

* Update vendor-requests.json

* Update analytics.amp.html

* Update analytics-vendors.amp.html

* Update vendors.js

* Update analytics-vendors.amp.html

* fix lint error

* update

* facebook pixel

* Update vendor-requests.json

* Update analytics.amp.html

* fix lint error

* fix lint error

*  fix alphabetical order

* Update yarn.lock

* AMP-ad supports Seznam Imedia ad network (ampproject#8893)

* AMP-ad supports Seznam Imedia ad network

* Fixed requested changes

* Fixed requested changes

* Render API implemented

* Improved placing ads

* Fixed requested changes

* Render API fixed context
* Renamed IM.cz to Imedia
* Described JSON parameters
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

4 participants