Enable http3 for a subset of domains

Wed Jun 26 15:56:20 UTC 2024

Hi,

> On 26 Jun 2024, at 7:21 PM, Riccardo Brunetti Host <riccardo.brunetti at host.it> wrote:
> 
> Hello, thanks for the answer.
> 
>> On 26 Jun 2024, at 16:45, Roman Arutyunyan <arut at nginx.com <mailto:arut at nginx.com>> wrote:
>> 
>> Hi,
>> 
>>> On 26 Jun 2024, at 6:15 PM, Riccardo Brunetti Host <riccardo.brunetti at host.it <mailto:riccardo.brunetti at host.it>> wrote:
>>> 
>>> Hello.
>>> 
>>> I have a Nginx server which acts as a reverse proxy for a given number of different domains and I’m trying to enable http3 for a subset of them.
>>> 
>>> I placed the directives:
>>> 
>>>>     http3 on;
>>>>     listen xx.xx.xx.xx:443 default_server quic reuseport;
>>>>     quic_retry on;
>>> 
>>> in the default.conf file and then, only for those domains which I want http3 enabled I also add in the corresponding virtual host:
>>> 
>>>>         http3 on;
>>>>         listen xx.xx.xx.xx:443 quic;
>>>>         quic_retry on;
>>> 
>>> In the server section and 
>>> 
>>>> add_header Alt-Svc 'h3=":8443"; ma=86400';
>>> 
>>> Inside the location.
>>> 
>>> Now, what happens is that if I try to open the sites *with* http3 enabled, everything works as expected. If I try to open a site *without* http3 enabled, sometimes I have an SSL error because the server returns the “default virtual host” certificate.
>>> It seems related to the type of browser.
>>> 
>>> Is it an expected behaviour or I am doing something wrong with the server configuration?
>> 
>> If you use http/3 to access a virtual server that does not support http/3, the default http/3 server will be used.
>> 
>> Make sure the above add_header does not affect the servers which do not support http/3.
> 
> I add the header above only on the virtual hosts which I want to support http/3, thus it should not interfere with the other non http/3 domains.
> Nevertheless, if I try with curl:
> 
> 1) Site http/3 enabled:
> 
>> # curl --http3 -I -v https://<fqdn>
>> * Host <fqdn>t:443 was resolved.
>> * IPv6: (none)
>> * IPv4: <ip>
>> * WARNING: no socket in pollset, transfer may stall!
>> *   Trying <ip>:443...
>> * Server certificate:
>> *  subject: CN=<fqdn>
>> *  start date: May 14 07:39:37 2024 GMT
>> *  expire date: Aug 12 07:39:36 2024 GMT
>> *  subjectAltName: host “<fqdn>" matched cert's “<fqdn>"
>> *  issuer: C=US; O=Let's Encrypt; CN=R3
>> *  SSL certificate verify ok.
>> * Connected to <fqdn> (<ip>) port 443
>> * using HTTP/3
>> * [HTTP/3] [0] OPENED stream for https://<fqdn>/
> …
> …
> 
> So it gets the correct certificate and it goes on using http/3
> 
> 2) Site non http/3 enabled:
> 
>> # curl --http3 -I -v https://<fqdn>
> 
>> * Host <fqdn>:443 was resolved.
>> * IPv6: (none)
>> * IPv4: <ip>
>> * WARNING: no socket in pollset, transfer may stall!
>> *   Trying <ip>:443...
>> * SSL certificate problem: unable to get local issuer certificate
>> * connect to <ip> port 443 failed: SSL peer certificate or SSH remote key was not OK
>> * Failed to connect to<fqdn> port 443 after 33 ms: SSL peer certificate or SSH remote key was not OK
>> *   Trying <ip>:443...
>> * Connected to<fqdn> (<ip>) port 443
>> * ALPN: curl offers h2,http/1.1
>> * TLSv1.2 (OUT), TLS handshake, Client hello (1):
>> * TLSv1.2 (IN), TLS handshake, Server hello (2):
>> * TLSv1.3 (OUT), TLS change cipher, Change cipher spec (1):
>> * TLSv1.3 (IN), TLS handshake, Encrypted Extensions (8):
>> * TLSv1.3 (IN), TLS handshake, Certificate (11):
>> * TLSv1.3 (IN), TLS handshake, CERT verify (15):
>> * TLSv1.3 (IN), TLS handshake, Finished (20):
>> * TLSv1.3 (OUT), TLS handshake, Finished (20):
>> * SSL connection using TLSv1.3 / TLS_AES_256_GCM_SHA384 / [blank] / UNDEF
>> * ALPN: server accepted h2
>> * Server certificate:
>> *  subject: CN=<fqdn>
>> *  start date: Apr 16 08:36:25 2024 GMT
>> *  expire date: Jul 15 08:36:24 2024 GMT
>> *  subjectAltName: host "<fqdn>" matched cert's "<fqdn>"
>> *  issuer: C=US; O=Let's Encrypt; CN=R3
>> *  SSL certificate verify ok.
>> * using HTTP/2
>> * [HTTP/2] [1] OPENED stream for https://<fqdn>/
> 
> 
> So it tries with http/3, gets a wrong certificate (the server default one), then it switches to http/2 and goes on. 
> The first attempt with http/3 probably gives problems with some browsers?
> If this is the case, is there a way to avoid this or should I simply configure http/3 on all domains?

Why would you (or client) connect to an http/3 port for a server which does not support http/3?
What would be your expectation in this case? The same will happen for http/1 and http/2 as well.
You can connect to an http/1 ssl port, request a missing server and receive a default server certificate.
It's just there is no fallback for http/1 and client will just give up.

The add_header suggests that http/3 is supported for a certain server.
It should not be sent for servers which do not support http/3, and clients should not use http/3 for them.

>> If it does, this will mislead clients by offering them to switch to unsupported http/3.
>> 
>>> Nginx version: nginx/1.26.1 on ubuntu 22.04
>>> 
>>> Thanks.
>>> Riccardo
>>> _______________________________________________
>>> nginx mailing list
>>> nginx at nginx.org <mailto:nginx at nginx.org>
>>> https://mailman.nginx.org/mailman/listinfo/nginx
>> 
>> ----
>> Roman Arutyunyan
>> arut at nginx.com <mailto:arut at nginx.com>
>> 
> 
> Riccardo
> 
>> 
>> 
>> 
>> _______________________________________________
>> nginx mailing list
>> nginx at nginx.org <mailto:nginx at nginx.org>
>> https://mailman.nginx.org/mailman/listinfo/nginx
> 
> _______________________________________________
> nginx mailing list
> nginx at nginx.org <mailto:nginx at nginx.org>
> https://mailman.nginx.org/mailman/listinfo/nginx

----
Roman Arutyunyan
arut at nginx.com

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.nginx.org/pipermail/nginx/attachments/20240626/7315a1fe/attachment.htm>