Page MenuHomePhabricator

TTFB is sometimes unrealistic fast for first view using Chrome
Closed, ResolvedPublic

Description

There's an issue today when we batch testing changes: Say that we access a URL 31 (or 51 it doesn't matter) times. Some of the runs gets a super fast TTFB (checkout the original issue on Github).

That makes it impossible for us to pick the faster run (because of the run are always wrong and gives us wrong values) and it happens so often so chosen the median run is also often wrong.

I've only seen this with connectivity cable.

There's a fix rolling out on WebPageTest (they are changing how they are measuring) but that will only help when we switch to H2. However there's also references to a bug in Chrome that's fixed i 49 (WebPageTest updated to 49 earlier this week), so it could be that this is an not an issue anymore, lets spend some lite to check.

Event Timeline

I've tested this now with Chrome 49, it still happens.I got the sense it doesn't happen as often as before, but I didn't spent any time to get the stats for it. Knowing that it still exists (http://wpt.wmftest.org/result/160318_XQ_4Z/48/details/) means we need to wait on H2.

We've now switched from SPDY to HTTP2. Is this still an issue?

Upstream https://github.com/WPO-Foundation/webpagetest/issues/566 is still open but claims to have it fixed.

hmm, let me check that, I've seen we get wrong TTFB metric on the second view, so I've created an issue for that: https://github.com/WPO-Foundation/webpagetest/issues/615

and it's my guess that it still exists for first view but it's best to check.

@Krinkle I couldn't find one using cable and Chrome (as in the original issue) but 3G and TTFB got 0.080 that seems too fast:

http://wpt.wmftest.org/result/160525_T1_3R/?medianRun=fastest&medianMetric=TTFB

Need to do a run of 31 or 51 to make sure it doesn't exist, I can try to do that between the standard runs later today.

Krinkle triaged this task as Medium priority.May 26 2016, 3:13 PM
Peter claimed this task.

I haven't seen this in our tests now for a while so lets close it an re-open if we find it again.