Very slow HTTP redirect

Hello,
we use Kubernetes (GKE) on GCP and our main gateway is controlled by Istio.
We’re facing a weird very slow (60 seconds) HTTP redirect only on our domains with Free plan.
The issue doesn’t occur if:

  • domains on Pro plan
  • bypassing Cloudflare proxy
  • using the Cloudflare Page Rule to redirect
  • using HEAD HTTP method
  • using another proxy (nginx) between Cloudflare and our origin servers

You can reproduce the problem simply running: curl -v https://weroad.co.uk
(it should redirect to https://www.weroad.co.uk)

Using Wireshark I see that one of the error is TCP Out-Of-Order. I attached the screenshot. (You can see the response HTTP 301 is 60 seconds after the request)

We noticed on our origin servers that the requests are HTTP/2. The doc [1] states that connections from Cloudflare’s Edge to your origin server(s) only support HTTP/1.1.
We also see that the requests on the domains with the Pro plan are HTTP/1.1

We checked the config on Istio for both domains and they are the same.

We’ve also opened a support ticket on Cloudflare but they are replying very slow and with generic answers.

If you need any further details, I’ll be glad to help.
Thanks

[1] Understanding Cloudflare HTTP/2 and HTTP/3 Support – Cloudflare Help Center

I’m seeing the same behaviour as your cURL command above. I see very different behaviour with curl -v https://weroad.co.uk --http1.1, I get the payload back, but the connection does not close for another 60 seconds. It looks to me like the Origin is not correctly ending the HTTP transaction, and you end up waiting for a 60 second timeout to occur somewhere.

If you run the same commands, but use --connect-to ::<YOUR ORIGIN IP> do you see the same behaviour?

Strangely, I get the opposite behaviour running --http1.1 and --http2 with a HEAD command.

1 Like

Hi,
I don’t see any issue connecting to the origin IP with both HTTP2 and HTTP1.1. The response is back just after a few milliseconds.

Thanks for helping.

The origin behaves normally to me. Perhaps one of the other @MVP have some ideas. The only thing that is different is the content-length header, but I don’t think that is mandatory.

% curl --dump-header - https://weroad.co.uk --http1.1 -w "@curl-format.txt" --connect-to ::192.0.2.1
HTTP/1.1 301 Moved Permanently
location: https://www.weroad.co.uk/
date: Fri, 17 Sep 2021 12:25:26 GMT
server: istio-envoy
content-length: 0

     time_namelookup:  0.015722s
        time_connect:  0.051990s
     time_appconnect:  0.158412s
    time_pretransfer:  0.158471s
       time_redirect:  0.000000s
  time_starttransfer:  0.188962s
                     ----------
          time_total:  0.189039s
% curl --dump-header - https://weroad.co.uk --http2 -w "@curl-format.txt" --connect-to ::192.0.2.1
HTTP/2 301
location: https://www.weroad.co.uk/
date: Fri, 17 Sep 2021 12:25:34 GMT
server: istio-envoy

     time_namelookup:  0.015037s
        time_connect:  0.053507s
     time_appconnect:  0.155554s
    time_pretransfer:  0.155686s
       time_redirect:  0.000000s
  time_starttransfer:  0.204107s
                     ----------
          time_total:  0.204315s

% curl https://weroad.co.uk --http2 -w "@curl-format.txt" --connect-to ::192.0.2.1 --head
HTTP/2 301
location: https://www.weroad.co.uk/
date: Fri, 17 Sep 2021 12:39:47 GMT
server: istio-envoy

     time_namelookup:  0.035334s
        time_connect:  0.074029s
     time_appconnect:  0.226184s
    time_pretransfer:  0.226531s
       time_redirect:  0.000000s
  time_starttransfer:  0.261735s
                     ----------
          time_total:  0.262415s
% curl https://weroad.co.uk --http1.1 -w "@curl-format.txt" --connect-to ::192.0.2.1 --head
HTTP/1.1 301 Moved Permanently
location: https://www.weroad.co.uk/
date: Fri, 17 Sep 2021 12:39:56 GMT
server: istio-envoy
transfer-encoding: chunked

     time_namelookup:  0.070452s
        time_connect:  0.102332s
     time_appconnect:  0.221772s
    time_pretransfer:  0.221937s
       time_redirect:  0.000000s
  time_starttransfer:  0.252150s
                     ----------
          time_total:  0.252512s

Slowness is from TTFB (%{time_starttransfer}) and it still happens for me with HEAD HTTP method. Testing via my curltimes.sh script comparing TLSv1.2 vs TLSv1.3 via compare mode https://github.com/centminmod/curltimes#compare-mode and both have 59-60s TTFB over a 3 sample run average test.

./curltimes.sh compare https://weroad.co.uk
-TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256
+TLS_AES_128_GCM_SHA256
 HTTP/2
 curl 7.78.0
-Connected to weroad.co.uk (2606:4700:3030::ac43:d3b9) port 443 (#0)
+Connected to weroad.co.uk (2606:4700:130:436c:6f75:6466:6c61:7265) port 443 (#0)
 Sample Size: 3

       time_dns
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.006361,0.002398,0.002347,0.014339,0.008369,0.013145,0.014100
tls13: 0.005671,0.002485,0.002393,0.012135,0.007310,0.011170,0.011942
       time_connect
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.014657,0.011011,0.010444,0.022515,0.016763,0.021365,0.022285
tls13: 0.016043,0.013658,0.011259,0.023212,0.018435,0.022257,0.023021
       time_appconnect
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.177748,0.174009,0.173559,0.185675,0.179842,0.184508,0.185442
tls13: 0.177909,0.178660,0.171236,0.183830,0.181245,0.183313,0.183727
       time_pretransfer
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.177847,0.174106,0.173653,0.185781,0.179944,0.184613,0.185548
tls13: 0.178003,0.178769,0.171326,0.183913,0.181341,0.183399,0.183810
       time_ttfb
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 20.375496,0.424580,0.421083,60.280826,30.352703,54.295201,59.083701
tls13: 40.333579,60.288287,0.418282,60.294169,60.291228,60.293581,60.294051
       time_total
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 20.375536,0.424616,0.421126,60.280866,30.352741,54.295241,59.083741
tls13: 40.333624,60.288334,0.418320,60.294217,60.291275,60.293629,60.294099

Notice

  • the median and minimum TTFB for TLSv1.2 run was 0.424580s and 0.421083s respectively, so not all runs were slow just some of the 3 sample runs.
  • the average TTFB for TLSv1.2 was 20.375s and TLSv1.3 was 40.333579s with 75% percentile for TLSv1.2 at 30.352703s and TLSv1.3 at 60.381228s which suggests there’s a wide variance between runs for TLSv1.2 and TLSv1.3 and the 2x CF IPv6 edge IP servers.

So retest to each IPv6 to see if it’s a specific CF edge server issue

Testing 2606:4700:3030::ac43:d3b9 shows very little variance in TTFB for TLSv1.3 for average, min, max and percentile metrics. But there’s variance in TLSv1.3 run for average and min TTFB

./curltimes.sh compare https://weroad.co.uk 2606:4700:3030::ac43:d3b9
-TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256
+TLS_AES_128_GCM_SHA256
 HTTP/2
 curl 7.78.0
 Connected to weroad.co.uk (2606:4700:3030::ac43:d3b9) port 443 (#0)
 Sample Size: 3
 

       time_dns
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.000027,0.000027,0.000027,0.000028,0.000027,0.000028,0.000028
tls13: 0.000026,0.000027,0.000023,0.000028,0.000027,0.000028,0.000028
       time_connect
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.008576,0.008507,0.008270,0.008952,0.008729,0.008907,0.008943
tls13: 0.008267,0.008197,0.008147,0.008457,0.008327,0.008431,0.008452
       time_appconnect
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.174458,0.172791,0.170233,0.180349,0.176570,0.179593,0.180198
tls13: 0.167212,0.163148,0.160755,0.177732,0.170440,0.176274,0.177440
       time_pretransfer
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.174555,0.172891,0.170329,0.180444,0.176667,0.179689,0.180293
tls13: 0.167301,0.163248,0.160846,0.177808,0.170528,0.176352,0.177517
       time_ttfb
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 60.271457,60.268942,60.268722,60.276706,60.272824,60.275930,60.276551
tls13: 40.367353,60.258555,0.426687,60.416817,60.337686,60.400991,60.413652
       time_total
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 60.271505,60.268995,60.268768,60.276752,60.272874,60.275976,60.276597
tls13: 40.367396,60.258598,0.426727,60.416863,60.337730,60.401037,60.413698

Testing 2606:4700:130:436c:6f75:6466:6c61:7265 again shows very little variance in TTFB for TLSv1.2 or TLSv1.3 for average, min, max and percentile metrics.

./curltimes.sh compare https://weroad.co.uk 2606:4700:130:436c:6f75:6466:6c61:7265
-TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256
+TLS_AES_128_GCM_SHA256
 HTTP/2
 curl 7.78.0
 Connected to weroad.co.uk (2606:4700:130:436c:6f75:6466:6c61:7265) port 443 (#0)
 Sample Size: 3
 

       time_dns
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.000028,0.000028,0.000028,0.000028,0.000028,0.000028,0.000028
tls13: 0.000027,0.000027,0.000027,0.000028,0.000027,0.000028,0.000028
       time_connect
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.014295,0.015836,0.011187,0.015862,0.015849,0.015859,0.015861
tls13: 0.011227,0.011214,0.011098,0.011370,0.011292,0.011354,0.011367
       time_appconnect
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.190015,0.191945,0.184906,0.193194,0.192570,0.193069,0.193169
tls13: 0.173744,0.173472,0.168771,0.178988,0.176230,0.178436,0.178878
       time_pretransfer
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.190108,0.192040,0.185000,0.193285,0.192663,0.193161,0.193260
tls13: 0.173838,0.173563,0.168871,0.179080,0.176321,0.178528,0.178970
       time_ttfb
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 60.307710,60.306213,60.298459,60.318459,60.312336,60.317234,60.318214
tls13: 60.287613,60.290566,60.280275,60.291997,60.291281,60.291854,60.291968
       time_total
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 60.307751,60.306253,60.298500,60.318499,60.312376,60.317274,60.318254
tls13: 60.287659,60.290613,60.280321,60.292043,60.291328,60.291900,60.292014

Test IPv4 IPs

dig +short A weroad.co.uk
104.21.93.157
172.67.211.185

Testing 104.21.93.157 shows there is variance in TTFB for TLSv1.2 or TLSv1.3 for average and min i.e. TLSv1.2 at 40.325739s and 0.438052s respectively but max and percentile metrics are in 60s

./curltimes.sh compare https://weroad.co.uk 104.21.93.157
-TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256
+TLS_AES_128_GCM_SHA256
 HTTP/2
 curl 7.78.0
 Connected to weroad.co.uk (104.21.93.157) port 443 (#0)
 Sample Size: 3
 

       time_dns
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.000028,0.000028,0.000027,0.000029,0.000029,0.000029,0.000029
tls13: 0.000027,0.000027,0.000027,0.000027,0.000027,0.000027,0.000027
       time_connect
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.008453,0.008211,0.008107,0.009040,0.008626,0.008957,0.009023
tls13: 0.008260,0.008181,0.008111,0.008489,0.008335,0.008458,0.008483
       time_appconnect
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.173376,0.173049,0.170823,0.176257,0.174653,0.175936,0.176193
tls13: 0.167114,0.167519,0.163852,0.169971,0.168745,0.169726,0.169922
       time_pretransfer
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.173477,0.173150,0.170922,0.176360,0.174755,0.176039,0.176296
tls13: 0.167208,0.167612,0.163948,0.170063,0.168838,0.169818,0.170014
       time_ttfb
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 40.325739,60.265718,0.438052,60.273447,60.269582,60.272674,60.273292
tls13: 40.324704,60.261087,0.444903,60.268121,60.264604,60.267418,60.267980
       time_total
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 40.325781,60.265760,0.438095,60.273487,60.269623,60.272714,60.273332
tls13: 40.324749,60.261133,0.444945,60.268168,60.264651,60.267465,60.268027

Testing 172.67.211.185 shows there is variance in TTFB for TLSv1.2 or TLSv1.3 for average and min i.e. TLSv1.2 at 40.382045s and 0.438585s respectively but max and percentile metrics are in 60s

./curltimes.sh compare https://weroad.co.uk 172.67.211.185
-TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256
+TLS_AES_128_GCM_SHA256
 HTTP/2
 curl 7.78.0
 Connected to weroad.co.uk (172.67.211.185) port 443 (#0)
 Sample Size: 3
 

       time_dns
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.000027,0.000027,0.000026,0.000029,0.000028,0.000029,0.000029
tls13: 0.000028,0.000028,0.000027,0.000028,0.000028,0.000028,0.000028
       time_connect
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.008504,0.008422,0.008408,0.008681,0.008551,0.008655,0.008676
tls13: 0.008256,0.008156,0.008142,0.008471,0.008313,0.008439,0.008465
       time_appconnect
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.178511,0.178111,0.174456,0.182967,0.180539,0.182481,0.182870
tls13: 0.165941,0.163887,0.162735,0.171200,0.167544,0.170469,0.171054
       time_pretransfer
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.178614,0.178221,0.174559,0.183062,0.180642,0.182578,0.182965
tls13: 0.166032,0.163977,0.162839,0.171279,0.167628,0.170549,0.171133
       time_ttfb
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 40.382045,60.276327,0.438585,60.431222,60.353774,60.415733,60.428124
tls13: 40.318993,60.263514,0.413441,60.280024,60.271769,60.278373,60.279694
       time_total
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 40.382086,60.276368,0.438623,60.431268,60.353818,60.415778,60.428170
tls13: 40.319036,60.263562,0.413478,60.280068,60.271815,60.278417,60.279738

This suggests it isn’t always slowing down but enough times to make it an issue that Cloudflare tech support should look into it if that HTTP redirect is being done on CF server. Or you’d need to look at it if the HTTP redirect is on your origin server end due to variance in response time back to CF edge servers from your origin.

The redirect is being done on our origin servers.
I tried your script using our public Origin IP:

➜  curltimes git:(master) ✗ ./curltimes.sh compare https://weroad.co.uk x.x.x.x
 curl 7.79.0
-TLSv1.2 ECDHE-RSA-CHACHA20-POLY1305
+TLSv1.3 TLS_AES_256_GCM_SHA384
 HTTP/2
 Connected to weroad.co.uk (x.x.x.x) port 443 (#0)
 Sample Size: 3
 

       time_dns
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.000035,0.000036,0.000033,0.000037,0.000036,0.000037,0.000037
tls13: 0.000036,0.000036,0.000035,0.000038,0.000037,0.000038,0.000038
       time_connect
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.021321,0.021136,0.020933,0.021894,0.021515,0.021818,0.021879
tls13: 0.021811,0.022010,0.021200,0.022224,0.022117,0.022203,0.022220
       time_appconnect
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.066370,0.065927,0.065487,0.067696,0.066811,0.067519,0.067661
tls13: 0.045951,0.045739,0.045670,0.046443,0.046091,0.046373,0.046429
       time_pretransfer
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.066823,0.066410,0.065780,0.068280,0.067345,0.068093,0.068243
tls13: 0.046150,0.045922,0.045859,0.046669,0.046296,0.046594,0.046654
       time_ttfb
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.087502,0.086975,0.086791,0.088740,0.087858,0.088564,0.088705
tls13: 0.067059,0.066755,0.066542,0.067881,0.067318,0.067768,0.067858
       time_total
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.087610,0.087098,0.086872,0.088859,0.087979,0.088683,0.088824
tls13: 0.067142,0.066817,0.066647,0.067961,0.067389,0.067847,0.067938

As you can see the responses are quite fast compared to 60 seconds.
I’ve tried to run the script multiple times and I always get similar result.

Thanks for helping

In that case probably best to wait till CF tech support can investigate the CF edge to origin connection side of things. Though the easiest solution is to move the HTTP redirect to CF edge side via page rule I suppose for now.

edit: are you running curl direct IP tests on same server as your origin? What if from a different server not within GCP network ?

Out of curiosity, I raised the sample runs from 3 to 9 to see if more variance exists and it does for TTFB

export curlruns=9; ./curltimes.sh compare https://weroad.co.uk 104.21.93.157
-TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256
+TLS_AES_128_GCM_SHA256
 HTTP/2
 curl 7.79.0
 Connected to weroad.co.uk (104.21.93.157) port 443 (#0)
 Sample Size: 9
 

       time_dns
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.000027,0.000026,0.000025,0.000030,0.000027,0.000029,0.000030
tls13: 0.000026,0.000026,0.000025,0.000027,0.000026,0.000027,0.000027
       time_connect
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.008143,0.008132,0.008104,0.008208,0.008155,0.008196,0.008206
tls13: 0.008455,0.008442,0.008137,0.009085,0.008477,0.009047,0.009077
       time_appconnect
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.171076,0.171257,0.168629,0.173749,0.171997,0.173066,0.173612
tls13: 0.163319,0.163226,0.159941,0.169896,0.163755,0.167990,0.169515
       time_pretransfer
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 0.171180,0.171357,0.168731,0.173843,0.172101,0.173168,0.173708
tls13: 0.163416,0.163319,0.160036,0.169994,0.163853,0.168088,0.169613
       time_ttfb
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 26.986086,0.428044,0.260266,60.268744,60.268090,60.268552,60.268706
tls13: 32.863600,53.215361,0.251801,60.270971,60.258487,60.268792,60.270535
       time_total
       avg:,median:,min:,max:,75%:,95%:,99%:
tls12: 26.986134,0.428082,0.260304,60.268793,60.268137,60.268600,60.268754
tls13: 32.863645,53.215412,0.251840,60.271021,60.258537,60.268841,60.270585

and individual test runs from sample run = 5 for IPv4 IP only where 1st, 2nd and 5th runs have TTFB in 60s

export curlruns=5; ./curltimes.sh csv-sum https://weroad.co.uk 104.21.93.157
TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256
HTTP/2
curl 7.79.0
Connected to weroad.co.uk (104.21.93.157) port 443 (#0)
Sample Size: 5

2.6e-05,0.008485,0.172112,0.172213,60.420048,60.420098
2.6e-05,0.008464,0.17078,0.17089,60.265623,60.265675
2.8e-05,0.008155,0.17147,0.171574,0.268667,0.268704
2.6e-05,0.008147,0.168822,0.168923,0.425061,0.4251
2.6e-05,0.008154,0.169423,0.169538,60.264577,60.264625

time_dns 
  avg:    0.000026 
  median: 0.000026 
  min:    0.000026 
  max:    0.000028 
  75%:    0.000026 
  95%:    0.000028 
  99%:    0.000028
time_connect 
  avg:    0.008281 
  median: 0.008155 
  min:    0.008147 
  max:    0.008485 
  75%:    0.008464 
  95%:    0.008481 
  99%:    0.008484
time_appconnect 
  avg:    0.170521 
  median: 0.170780 
  min:    0.168822 
  max:    0.172112 
  75%:    0.171470 
  95%:    0.171984 
  99%:    0.172086
time_pretransfer 
  avg:    0.170628 
  median: 0.170890 
  min:    0.168923 
  max:    0.172213 
  75%:    0.171574 
  95%:    0.172085 
  99%:    0.172187
time_ttfb 
  avg:    36.328795 
  median: 60.264577 
  min:    0.268667 
  max:    60.420048 
  75%:    60.265623 
  95%:    60.389163 
  99%:    60.413871
time_total 
  avg:    36.328840 
  median: 60.264625 
  min:    0.268704 
  max:    60.420098 
  75%:    60.265675 
  95%:    60.389213 
  99%:    60.413921

time_dns
avg:,median:,min:,max:,75%:,95%:,99%:
0.000026,0.000026,0.000026,0.000028,0.000026,0.000028,0.000028
time_connect
avg:,median:,min:,max:,75%:,95%:,99%:
0.008281,0.008155,0.008147,0.008485,0.008464,0.008481,0.008484
time_appconnect
avg:,median:,min:,max:,75%:,95%:,99%:
0.170521,0.170780,0.168822,0.172112,0.171470,0.171984,0.172086
time_pretransfer
avg:,median:,min:,max:,75%:,95%:,99%:
0.170628,0.170890,0.168923,0.172213,0.171574,0.172085,0.172187
time_ttfb
avg:,median:,min:,max:,75%:,95%:,99%:
36.328795,60.264577,0.268667,60.420048,60.265623,60.389163,60.413871
time_total
avg:,median:,min:,max:,75%:,95%:,99%:
36.328840,60.264625,0.268704,60.420098,60.265675,60.389213,60.413921

Cloudflare recently fixed 520 errors due to origins not handling keepalives properly @simon might have some ideas if the fix is related to this ? @miroberto tried enable or disabling Keepalives on origin side or playing with related settings?

In all tests, I always run curl from my computer which is not in the same network of our servers.
I’ve also tried from different network (4G network from my phone) and I get similar result.

@eva2000 I’ve tried again with curltimes script and I’ve got very similar result to yours, more variance when raising the sample. (but I do NOT notice variance if I use the Origin IP, it’s always fast)

We noticed the problem the first time on August 23, 2021, the same day we opened the support ticket to Cloudflare. Originally the ticket was for the domain weroad.es (which share the same infrastructure of weroad.co.uk), but as soon as we updated the subscription from Free to Pro plan the website started to respond fast.

Unfortunately we still have not a proper solution since we have several redirects and using the Page Rules is not feasible.

Thanks

Looks like a CF end issue. Actually retested again now and no slow TTFB found all 9 runs are fast so maybe they fixed it on their end?

export connect_from_display='y'; export curlruns=9; ./curltimes.sh csv-sum https://weroad.co.uk 172.67.211.185
TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256
HTTP/2
curl 7.79.0
Connected to weroad.co.uk (172.67.211.185) port 443 (#0)
Cloudflare proxied https://weroad.co.uk
Sample Size: 9

2.7e-05,0.008769,0.17108,0.171182,0.419895,0.419934
2.7e-05,0.00817,0.168058,0.168173,0.433296,0.433334
2.6e-05,0.008655,0.185276,0.185373,0.287577,0.287615
2.7e-05,0.008152,0.172258,0.172367,0.266759,0.266797
2.6e-05,0.008665,0.173615,0.173715,0.275577,0.275615
2.6e-05,0.00814,0.182337,0.182437,0.286607,0.286649
3.3e-05,0.008081,0.174596,0.174706,0.274566,0.274607
2.7e-05,0.008162,0.173617,0.173724,0.272547,0.272587
2.6e-05,0.008468,0.170815,0.170921,0.272304,0.272347

time_dns 
  avg:    0.000027 
  median: 0.000027 
  min:    0.000026 
  max:    0.000033 
  75%:    0.000027 
  95%:    0.000031 
  99%:    0.000033
time_connect 
  avg:    0.008362 
  median: 0.008170 
  min:    0.008081 
  max:    0.008769 
  75%:    0.008655 
  95%:    0.008727 
  99%:    0.008761
time_appconnect 
  avg:    0.174628 
  median: 0.173615 
  min:    0.168058 
  max:    0.185276 
  75%:    0.174596 
  95%:    0.184100 
  99%:    0.185041
time_pretransfer 
  avg:    0.174733 
  median: 0.173715 
  min:    0.168173 
  max:    0.185373 
  75%:    0.174706 
  95%:    0.184199 
  99%:    0.185138
time_ttfb 
  avg:    0.309903 
  median: 0.275577 
  min:    0.266759 
  max:    0.433296 
  75%:    0.287577 
  95%:    0.427936 
  99%:    0.432224
time_total 
  avg:    0.309943 
  median: 0.275615 
  min:    0.266797 
  max:    0.433334 
  75%:    0.287615 
  95%:    0.427974 
  99%:    0.432262

time_dns
avg:,median:,min:,max:,75%:,95%:,99%:
0.000027,0.000027,0.000026,0.000033,0.000027,0.000031,0.000033
time_connect
avg:,median:,min:,max:,75%:,95%:,99%:
0.008362,0.008170,0.008081,0.008769,0.008655,0.008727,0.008761
time_appconnect
avg:,median:,min:,max:,75%:,95%:,99%:
0.174628,0.173615,0.168058,0.185276,0.174596,0.184100,0.185041
time_pretransfer
avg:,median:,min:,max:,75%:,95%:,99%:
0.174733,0.173715,0.168173,0.185373,0.174706,0.184199,0.185138
time_ttfb
avg:,median:,min:,max:,75%:,95%:,99%:
0.309903,0.275577,0.266759,0.433296,0.287577,0.427936,0.432224
time_total
avg:,median:,min:,max:,75%:,95%:,99%:
0.309943,0.275615,0.266797,0.433334,0.287615,0.427974,0.432262

Yes, I can confirm that now the problem doesn’t occur anymore.
We haven’t changed any settings on our side. It would be nice to have the confirmation by Cloudflare that there was a problem and their Edge and they fixed it. (I see no mention of that in the status page)

Anyway, thanks all for helping.

1 Like

This topic was automatically closed 3 days after the last reply. New replies are no longer allowed.