Input Advisor http2 button

[Bonsai78]

The Advisor should at least have the posibillity to siwtch of http2, if it is on .... and it should probably explain what is behind ....

I am new with Plesk an lerned now the advisor is horrible dangerous. If I switched on http2, all my websites only produced errormessages. To find out then, that I need to be root to switch this off on commandline costed me half an hour, as I did not have the root password at this time. I took me another half an hour to find out, that this ngix somehow was switched on, without a working configuration. (Dont' know if it was on before, but if so, the configuration before worked).

As two productive onlineshops got killed for an hour, I have to say this button is not good for the health of the administrator. The adrenaline levels raised to critical values. Adrenaline is quite unhealthy if you sit in front of a screen and do not move ....

I read two hours about http2 in the web before I gave this button a try ..... but I really did not expect it has no undo function .... otherwise I never would have tried thi on a productive machine.

 

[Ruslan Kosolapov]

Hi @Bonsai78 !

It's the very first time I've got feedback the http2 breaks websites

Could you share the configuration of your server (OS, virtualisation type etc) and the exact error messages with me?
Another thing I'm interested in - how did you fix the problem? Plesk provides Webserver Configurations Troubleshooter right in the UI, have you tried it?

 

[Hex]

I did that exact same thing from the advisor and nothing happened to the websites, it got faster. (just to share my experience not to false your statement) it could be true to some configurations.

AWS
CentOS7 KVM
Plesk Onyx Version 17.8.11 Update #14
Nginx enabled
Firewall enabled
Fail2ban enabled
IPv6 enabled bty

and yes how did you fix the problem and what is the error massages on the websites?

 

[Bonsai78]

@Ruslan Kosolapov Sorry, it was at least two weeks ago.
In Firefox came for all sites "SSL_ERROR_RX_RECORD_TOO_LONG"

I switched of http2 by command line

plesk bin http2_pref disable

I recomend to add this line behind a button in the Advisor. It saves time to google this.

After http2 was off, I got 502 Bad Gateway in the browser.

Then I stopped the ngix service and everything started to work again.

My System:
CentOS Linux release 7.5.1804 (Core) running on a virtual machine (no Idea on what, I only have access to the VMs)
Plesk Onyx Version 17.8.11 Update #14
Firewall enabled
Fail2ban enabled
PHP 5.6.22 by OS-Vendor is running as PHP-FRM
Not sure if ngix was running before, now it is off. My configurations of the sites also telling me, ngix is not active.
Where I see if IPV6 is enabled? In IP-Addresses configuration I see only a IPV4, so I guess it is off.
A pfSense Firewall is running between the public IP and my Plesk machine.
At the time when I switched on HTTP2 there was no public IP in the configuration of Plesk. Field was empty. Now as I have set this also the Filezilla is not any longer complainig about the IP. I found this setting today as I try to find out why SNI fails on my machine.

No I did not try to use the Webserver Configurations Troubleshooter, as I only had a full machine backup, and there was absolutely no time to do backups of all databases changed over day and all files changed over day. I had to come online asap.

@Hextrator: I fully agree that this is a configuration based error, otherwise the forum would be full of compliants. My problem here is, I did not setup the system. My Provider did, and I have no idea what they changed from default. I also have absolutely no experience with any reverse proxy servers. I never used one. I only have some experience with Apaches, Php and Mysql/Mariadb. I am new to Plesk, too.

 

[Bonsai78]

@Ruslan Kosolapov: I tested again, also the Webserver Configurations Troubleshooter can't resolve this nginx problem. It might be a good idea to have HTTP2 served from Apache only. I can't understand why I need a proxy as Apache normally can http2 out of the box.

 

[Ruslan Kosolapov]

@Bonsai78, thanks! We will analyse your configuration and the error message, and we'll try to repeat the bug (to fix the bug we need to repeat it )

Regarding http2 and Apache: you can vote for this functionality on our UserVoice Support for HTTP/2 on APACHE (only) without using nginx as proxy! And Server Push!

PS: actually, I have no doubt that this functionality makes sense (and we have it in the backlog actually), but there are another features/requests with higher priority.

 

[Bonsai78]

There's no bug with the http2, it is just nginx fails on my system. No idea what my provider did with this system before handing over to me ....

 

[zanuda]

Hello, we were able to reproduce your issue.
We create a bug EXTADVISOR-439 which should be fixed in one of the nearest releases.

 

[Bonsai78]

Wow great! Thank you very much!

 

[Tristan]

In my case the cipher suite was not correct in ssl.conf just a tip!

 

[Tristan]

Resolved - "ssl_ciphers" directive is duplicate in /etc/nginx/conf.d/ssl.conf