Bad Worker Performance. Very High TTFB for simple HTML output Workers!

Hi All.

I am since Day 1 a paid PRO plan with ARGO enabled cloudflare customer.

Starting last week of december my WebVital values for my cloudflare website started
going from good to bad without that i changed anything on the Website or messed
with the cloudflare configuration while this time …

See Evidence 1 here:

Performance1

Becouse of this i have executed a cloudflare speed test insde the cloudflare speed overview dash board to see if the speed test for my website is good but got a page load time of 17 Seconds showed …
The stated test location was:
South Carolina, USA
Connection Type: Cable

See Evidence 2:

After this i have done a webpagetest to check what is going on
with my cloudflare worker website and was surprised to see non constant page speed values
for my website.

As a example for the first test run executed
From: California, USA - EC2 - Chrome - Cable
1/12/2022, 5:32:10 AM
the Worker TTFB was the right one like it should be with 69 ms

The Second test run which was executed nearly in parallel
From: California, USA - EC2 - Chrome - Cable
1/12/2022, 5:32:10 AM
the Worker TTFB got very slow and bad 690ms

See Evidence 3:


After searching the Forum here for more info related to bad TTFB
i found several people complaining too starting from december.
See here:

Used the suggested tool to test TTFB for my website worldwide and got surprised too bad ttfb.

See here Evidence 4:

Something is clearly not right with Workers environment.
Such bad TTFB for a simple html ouput worker should not appear.

Thanks in advance for fixing this bad TTFB Problem with Workers.

I make heavy use of Workers that are nothing but static HTML, so I’ve been keeping a close eye on performance. Weekly tests on FastOrSlow for one domain showed degraded performance from Nov 23 for about three or four weekly tests, then got back to normal. In that time, though, TTFB was still <100ms.

Here’s a current test from Sucuri from my b8a.me site. This site uses a Worker for just the home page static HTML, plus Workers for caching the rest of the site.

2 Likes

Ohh Wooww Thank you very much @sdayman for the helpfull and fast reply !
Really strange this Workers performance degradation.

Tested my Website with your tool FastOrSlow and see for some Location very bad TTFB also.
I really dont understand why in specific POPs Location at random time the output of a simple Static HTML Worker goes so bad.
TTFB-Performance1

I just tested your Website performance with the same Tool and see for your website
too bad TTFB for specific POPs.

I guess at specific POPs the cloudflare infrastructure is underpowered.
This would guess explain why around black firday / christmas/ new year
the performance started to get bad.

Have to keep a eye on this cloudflare TTFB slow performance problem.

The drawback to FoS testing (and all the rest, I suppose) is that I don’t know which datacenter they’re routed to.

Actually, with both gtmetrix.com and webpagetest.org your can see a waterfall with each file and the returned headers.

1 Like

Of course I didn’t think of that…and even the screenshot the OP posted shows that very same thing…and I didn’t notice.

@seoworks, which test is shown in Evidence 4?

1 Like

Hi @sdayman

The Evidence 4 test is speedvitals test => Measure TTFB from 25 Locations - SpeedVitals to measure ttfb.

Found it reading this reply here.

1 Like

I definitely see major differences between my free plan and enterprise in where requests are routed.

1 Like

You can tag a test with query string or specific site page to test and then use Cloudflare Cache, Web analytics to log and inspect them. If you use Enterprise plan cloudflare, you can also inspect the tagged requests via Logpush CF edge server logs :slight_smile:

expected as Cloudflare Enterprise allows access to all CF datacenter pops while lower plans wouldn’t as expected.

Have you checked your Worker metrics for exception errors? If you configured your CF Worker to bypass exception errors to hit your origin server rather than give an end user a CF Worker exception error message, then the high TTFB could be due to Worker exception errors which route requests back to the origin and that will have a much higher TTFB than CF Workers if you’re using CF caching with CF Worker.

I came across that issue as I had CF Worker caching configured but also had some HTTP/2 specific fields in use and everytime a HTTP/3 request came in, it would cause a CF Worker exception error and bypass and route to my origin. I had to add some logic in my CF Worker to only set HTTP/2 specific fields if it’s a HTTP/2 request and not HTTP/3 request.

3 Likes

Hi @eva2000

Thanks a lot for your suggestion.
Will check the log in cloudflare workers to see if i can find something.

I just went to my google webmaster console and saw that google reported for
my cloudflare worker website a huge SERVER CONNECTIVITY ERROR Rate which is very strange too.

Never had this problem before.

The google SERVER CONNECTIVITY ERROR crawler error rate was always Zero before and now out of sudden the last days jumped to 7% without that anything changed neither in traffic nor in configuration.

See here evidence from Google Console !

My whole website is proxied behind cloudflare and everything is setup like all the time before.
This kind of slow ttfb and Cloudflare SERVER CONNECTIVITY ERROR errors should not have happened or appeared.

Really strange!

There seems to be outages from time to time while the whole day.
Something is not right again.

Huge Increase in Workers Latency for simple HTML Output Worker.
TTFB is huge.

Content is delivered from the nearest POP so it is not
routing but it takes a lot of time after the Request is send from the Webbrowser
to recieve anything from the nearest Cloudflare POP.

Just witnesed it myself a TTFB of 1500 ms
for the nearest POP for me ZRH a few minutes ago.

Everything feels Slow.

Looks like its not just only Workers but all the Cloudflare Services
that slows down everything.

Could make a Screenshot of a slowed down request to my nearest POP.

Normally TTFB is under 100 ms but now its 4X what is normaly is.

Could now also catch a Connection Problem for a Request.

Looks like some Cloudflare POPs are having problem even establishing Connections
from time to time.

https://developer.chrome.com/docs/devtools/network/reference/#timing-explanation

Looks like a internet wide problem that is going on the whole day.

Just visited the support website of the biggest internet provider in switzerland
and see that nearly all internetweb services have problems at the moment.

So this is more of a Internet wide ISP provider problem it looks like hmmm…

Realtime Internet Disturbances in Switzerland.