Extreme high load times

I'm experiencing very high load times across devices and browsers. When trying to go to my dashboard https://myurl.prismic.io/documents/working/ it takes almost three minutes to load - all other pages are fast since I am on a 250 MBit/s line. The file in question is main.js which takes 3minutes to download 2MB - something must be off with your caching infra or so...

Chrome latest stable

Hey Jannis, thanks for reaching out and welcome to the Community forum

It is weird that even if you have a fast connection the load time is still very long. Usually this happens because the browser doesn't have the file in the cache yet.

There's nothing unusual happening on out end, perhaps your connection goes unstable at times? If you're ever curious about our connectivity status and the health of the API you can always check out Status page

Hmm it's very odd, for example when I cURL using this formatter I get:

curl -w "@curl-format.txt" -o /dev/null -s "https://eyesbound.prismic.io/...6c549e9/javascripts/main.js"
    time_namelookup:  0.045091s
       time_connect:  0.157528s
    time_appconnect:  0.388975s
   time_pretransfer:  0.389079s
      time_redirect:  0.000000s
 time_starttransfer:  0.507990s
                    ----------
         time_total:  56.735730s

While for example this 5MB test file comes down in no-time.

curl -w "@curl-format.txt" -o /dev/null -s "http://ipv4.download.thinkbroadband.com/5MB.zip"
    time_namelookup:  0.030086s
       time_connect:  0.078942s
    time_appconnect:  0.000000s
   time_pretransfer:  0.079021s
      time_redirect:  0.000000s
 time_starttransfer:  0.129667s
                    ----------
         time_total:  2.442349s

Also the forum here or the prismic homepage and api are blazing fast - just that one page. Turning Adblockers on/off doesn't make a difference :man_shrugging:

Are you using some global cdn or something? I'm in central europe, maybe something reroutes me to some very remote datacenter instead of using a local?

There's a cool tool that I often use to know which technologies a website is using, it's a Google chrome extension called Wappalyzer.

Wappalyzer: Find out what CMS a website is using, as well as any framework, ecommerce platform, JavaScript libraries and many more. I helps you check for more than a thousand technologies in dozens of categories such as programming languages, analytics, marketing tools, payment processors, CRM, CDN and others.

OK thanks - now I can see that you are using Amazon Cloudfront - but I guess it is easier to check your own infrastructure than me analysing it, no? The issue still persists across devices, and networks. I even tried to use a VPN to see if I don't get the error in a different location... Am I really the only one that has this issue? Is there then maybe something off with my specific subdomain under prismic?

You're probably right, I wanted to share this tool with you because I find it useful!
Your case it's probably not the only one, but the reasons may vary depending on each individual environment, internet connection, etc.

We haven't find any particular logs/errors related to your repository.
Sometimes repos load slowly because there are too many tags existing in the repo, this behaviour will change in the future, but it is true that this could be one possible reason of why you're experiencing high load times.

Hi @prismic.io,

I've reopened this for you now.

P.S. nice user name.

@Phil Hah! Thanks, this forum picked it automagically for me based on my email address :smiley:

I'm worried this seems to be with my ISP, as many AWS S3 pages are slow, but when I ssh into my server it's way faster from there.

I used GitHub - reorx/httpstat: curl statistics made simple for formatting

Server:

Connected to 34.204.163.152:443 from 100.64.15.2:39574

HTTP/2 200
date: Mon, 04 Jan 2021 17:21:36 GMT
content-type: application/javascript; charset=utf-8
content-length: 8493643
server: nginx
x-ratelimit-limit: 200
x-ratelimit-remaining: 182
x-ratelimit-reset: 1609780896
cache-control: max-age=315360000
etag: "5463ac5c6602ba002b15faca9ce4356d99cc5d4d"
last-modified: Thu, 26 Nov 2020 09:34:18 GMT

Body stored in: /tmp/tmpCEvn1o

  DNS Lookup   TCP Connection   TLS Handshake   Server Processing   Content Transfer
[     7ms    |      119ms     |     526ms     |       126ms       |      15308ms     ]
             |                |               |                   |                  |
    namelookup:7ms            |               |                   |                  |
                        connect:126ms         |                   |                  |
                                    pretransfer:652ms             |                  |
                                                      starttransfer:778ms            |
                                                                                total:16086ms

Local:

Connected to 34.204.163.152:443 from 192.168.179.2:50714

HTTP/2 200
date: Mon, 04 Jan 2021 17:22:31 GMT
content-type: application/javascript; charset=utf-8
content-length: 8493643
server: nginx
x-ratelimit-limit: 200
x-ratelimit-remaining: 197
x-ratelimit-reset: 1609780952
cache-control: max-age=315360000
etag: "5463ac5c6602ba002b15faca9ce4356d99cc5d4d"
last-modified: Thu, 26 Nov 2020 09:34:18 GMT

Body stored in: /tmp/tmpUEaR0V

  DNS Lookup   TCP Connection   TLS Handshake   Server Processing   Content Transfer
[    30ms    |      108ms     |     232ms     |       216ms       |      550394ms     ]
             |                |               |                   |                   |
    namelookup:30ms           |               |                   |                   |
                        connect:138ms         |                   |                   |
                                    pretransfer:370ms             |                   |
                                                      starttransfer:586ms             |
                                                                                 total:550980ms

Hello Janis, there's another thing that sometimes causes repositories to be slow. The way we do it is: when a repo isn't used for a period of time, we put this repo in sleep mode. When a user tries to access it after a while (it can be one night) we "wake up" the repository and it can take some time, and the timeouts appear after 30 sec.

We will improve this when refactoring our database later this year, but this behaviour will remain the same.

Ahh I see! Is there a way for you to find out on the fly if a repo was put to sleep / freeze or not, so that for the user you can show a loading screen? This would make a better UX and the user knows something to expect to happen after a while.

@prismic.io You're definitely not the only one. We have the same issues for weeks with different accounts, different users and different networks/locations. Most of the time we get a white blank page after loading for several minutes.

Hello @alex2, welcome to the community

Could you please send me the repository URL where you and other users are having this issue to help me debug it?

I see that you have multiple repositories created with the email you use here in the forum. But I can't see the other users (are you sharing your account ?) also from different networks/locations (you are in Europe, is that right ?)

Hi Paulina,

i've sent you the other repository url via direct message and yes, we're in Europe.

1 Like

Hello @prismic.io @alex2 ,

Thanks for the infos you provided. We received some complaints from clients and they are all based in Germany,
After some research it seems that Telekom Germany are having some routing issues with GitHub & AWS,
are you using the same ISP ?

(cf: https://telekomhilft.telekom.de/t5/Telefonie-Internet/Amazon-AWS-S3-Github-downloads-sehr-langsam-nicht-nutzbar/td-p/4910937)

Can you please try to download some releases from Github and check if it's also slow ?

In the meantime i opened a case with AWS support to check if they have more informations.

Yes, I fear that is exactly the case. That's why I ssh-ed into my server and it was way faster from there.

1 Like

This issue has been closed due to inactivity. Flag to reopen.