Virtualmin reconfigured all php-fpm configs after upgrade to 6.07 GPL

Due to etckeeper after upgrade to 6.07 virtualmin changed port of all websites.

This caused error because virtualmin did not reconfigure apache configuration to use new ports for every php-fpm enabled website.

This was an serious issue for us.

Any idea?

Yep, I had same issue last night… I think I’m going back to FCGID and will let FPM mature a while longer.

The solution (workaround) is to change all domains to FCGID and back to FPM:

Step 1:

mkdir /var/run/php
chown -R www-data:www-data /var/run/php
/usr/sbin/virtualmin modify-web --all-domains --mode fcgid
/etc/init.d/apache2 restart
/etc/init.d/php7.0-fpm restart

Step 2:

mkdir /var/run/php
chown -R www-data:www-data /var/run/php
/usr/sbin/virtualmin modify-web --all-domains --mode fpm
/etc/init.d/apache2 restart
/etc/init.d/php7.0-fpm restart

(you might need to adjust the init.d lines, per your configuration)

Yeah the upgrade took out all of the php driven websites on my server. I changed from fpm to fcgi and back again on just one website which resolved this for the entire server. I think fpm just needed restarting actually…something not quite right in the update script perhaps?

I had all kinds of issues with PHP-FPM last week, but they disappeared when I disabled fail2ban. I’m using CSF Firewall and have it configured to much of what I understand fail2ban does, so I disabled fail2ban. But perhaps coincidentally, I haven’t had any PHP-FPM problems since.

I don’t know if there’s any connection because I know almost nothing about fail2ban or how it works, and I don’t have the time to learn it right now. All I know is that I haven’t had any PHP-FPM problems since I disabled it, which was for reasons completely unrelated to PHP-FPM.

Got the same problem with nginx configs!
After upgrade on two servers - php_from_port numbers changed, but in nginx configs are stayed old port numbers.
Example of php-fpm port changes: from 8000 to 8076, 8001 to 8079, 8002 to 8080, etc (so randomly).

Trick from helps for virtualhost configs, that not modified manually, great thanks @jimdunn!

Does any way to bulk rollback old port numbers in php_fpm exists?

Seems this feature caused to port changes?

I see the same problem here. It seems I have to go FCGID and back FPM for all virtualhosts, doing it just for one is not enough. I’ll try jimdunn’s trick on other hosts but yes, it’s a real problem especially in the middle of august with a lot of people away in vacations :frowning:
Maybe it’s also the case for Virtualmin’s people … enjoy ! :slight_smile:

On each “Re-Check Configuration” action Virtualmin 6.07 continue changing php-fpm ports of each virtualserver to new numbers with message “Fixing port clash for PHP-FPM version 7.2.3”!! So all manual-configured web servers need to be updated manually :frowning:
How can I tell Virtualmin to stop change ports?

For the time being I switched back to FCGID for all my hosts … and that’s a lot ! I’m surprised that this problem does not get more attention/reaction from users … does it mean that only a few people are using FPM ?

Or didn’t update yet , you can the security part solved with some manual edits. :wink:

Or using php fpm different way.

OK i think a lot of us are waiting for updated solution

Frankly, I’ve always had problems with PHP-FPM on any server I’ve ever used since it came out, regardless of what control panel was installed, or even if no control panel was installed. It doesn’t matter whether I configure it to listen on a socket or a port. At some point, usually several times a week, it poops the bed. The only difference is which error it throws.

I gave up on it a long time ago. Sometimes I try it for a while on one of the test sites I have for just such troubleshooting. I think the record has been six days without an error that shut down the whole site. So I just use FCGID. My servers are lightly-loaded. You can’t even tell the difference – except that FCGID doesn’t crash.


And with virtualmin 6.05 box centos 7 remiphp and codeit ( having for port 443 then openssl 1.1.1c and tls 1.3) repo also none probs after manual set it right. once.

That why i don’t try the update yet :wink: ( changepas renamed and that one part in the conf deleted and so for security meantime as workarround)

This one to

It was with install already a problem but takes pretty much time to have it here known at support as BUG , i did do it manualy at commandline to that setting for pass and so with mariadb 10.3 , and replied on some here in forum with the official mariadb cookbook for that. ( YUP it did take much time and trying while one don’t expect such)

IT is YUP real hard to see and look and search the forum, the docs, and github info’s to know where are bugs, what is status for those and so on not so good.

Issue’s that are going private doesn’t make knowing about status / bugs / info’s easier . :wink:

So for both : Security related and a Schema “” BUGS" / in work should be found on one place much more easy then it is now?