Tristan's questions

After a small update ran by an Ansible playbook from MariaDB 10.5.4 to MariaDB 10.5.5, the custom configuration file got silently overriden by the default configuration causing the server to refuse connection because the valid custom configuration was no longer present.

Our custom configuration file was stored under /etc/mysql/my.cnf

My question is: What is the best practice to store the custom configuration so it never again gets replaced by the default one when an update comes out

Is it /etc/mysql/conf.d/ ?

Distribution: Debian 10 installed with the official repository: https://downloads.mariadb.com/MariaDB/mariadb-10.5/repo/debian

I have 2 scenarios:

• /media/cache/test.jpg or stylesheet/theme/xxxx.css is generated by the application. It's not a static file present on the filesystem.

• /assets/test.jpg is a static file present on the host

I want both of them to be cached by the browser adding expires 6M header.

So I did:

location / {
    try_files $uri @rewriteapp;
}

location @rewriteapp {
     rewrite ^(.*)$ /index.php/$1 last;
}

location ~ ^/(index)\.php(/|$) {
    fastcgi_pass unix...;
    internal;
}

location ~ \.php$ {
    return 404;
}

# assets, media
location ~* \.(?:css(\.map)?|js(\.map)?|jpe?g|png|gif)$ {
    expires 6M;
    access_log off;

    # try accessing the file directly, and if not found 
    # it means the application has to generate it, 
    # so reroute to the @rewriteapp rule.
    try_files $uri @rewriteapp;
}

Problem is nginx detects a loop:

*194369 rewrite or internal redirection cycle while redirect to named location "@rewriteapp", client: 127.0.0.1, server: , request: "GET /stylesheet/theme/3441-1484735061.css

What am I missing please ?

I'd like to redirect all traffic from

www.*.*.<gTLD> to *.*.<gTLD>

If possible, I'd like to do that outside the block that serves the application. This block does not use a server_name because it listens to 5,000+ hostnames on the same app.

Do you have an idea on how to achieve that please ? All the example I find is only for named hosts....

Thank you

My application generates dynamic content and I try to cache them using browser cache. So If I try to add to my server block the following:

location ~*  \.(jpg|jpeg|png|gif|ico|css|js)$ {
    expires 365d;
}

Nginx no longer pass the request to the backend to generate the assets and it ends up with http status code 404

Instead, adding those 2 following conditions could solve my problem (taken from this question):

location / {
    if ($upstream_http_content_type ~ "(image/jpeg)|(image/png)|(image/gif)|(image/svg+xml)|(text/css)|(text/javascript)|(application/javascript)") {
      expires 90d;
    }

    if ($sent_http_content_type ~ "(image/jpeg)|(image/png)|(image/gif)|(image/svg+xml)|(text/css)|(text/javascript)|(application/javascript)") {
      expires 90d;
    }
    try_files $uri @rewriteapp;
}

location @rewriteapp {
    rewrite ^(.*)$ /site.php/$1 last;
}

location ~ ^/(site|site_dev|admin|admin_dev)\.php(/|$) {
    # it's PHP that generates some of my assets
    fastcgi_pass unix:/var/run/php/php7.1-fpm.sock;
    fastcgi_split_path_info ^(.+\.php)(/.*)$;
    include fastcgi_params;
    fastcgi_param SCRIPT_FILENAME $document_root$fastcgi_script_name;
    fastcgi_param HTTPS on;
}

The problem is that I can't see in the http response headers any date that indicates that it's working and images will be served from browser cache for 90 days.

What am I missing ?

I just upgraded to Debian 9 (stretch) but system seems to not support ALPN negociation (after running a HTTP2 test). All guides over the internet still refers to Jessie (Debian 8) but none talks about Stretch (Debian 9).

OpenSSL> version
OpenSSL 1.1.0f  25 May 2017

nginx -v
nginx version: nginx/1.13.1

nginx was installed from official repository

Is there a simple way (without recompiling manually nginx) to get ALPN with Debian 9 & nginx ?

Thanks.

I upgraded to Jessie last recently and when doing a lsb_release -a it reports that the release is 7 (instead of 8) and codename Jessie.

I did an update && upgrade && dist-upgrade when all sources point to jessie but no updates are available.

Another server I recently upgraded reports release 8 and Codename Jessie.

How can I fix this wrong information reported inside Release number ? Thanks

Why in Varnish 4.0 can't I see my custom header set on the response ?

sub vcl_recv {
   // disable varnish for now
   return (pass);
}

sub vcl_backend_response {
    set beresp.http.X-Backend = "my_custom_hostname";
    return (deliver);
}

It was working some times ago, (in varnish 3.x or without http2 on nginx backend, don't know which factor is it).

Do you know what's causing the fact that i can't see my X-Backend custom header in the http response ?

Thanks.

We have currently a wildcard vhost accepting all domain request.

I want to redirect subdomain.domain.com to www.subdomain.domain.com so my wildcard will answer

One exception, I dont want to redirect www.domain.com to www.www.domain.com

How to achieve that ? I've put together (outside the wildcard block) but it's not dealing with the exception and not sure about the variable thing

if ($http_host ~ (.*).domain.com) {
    set $test P;
}

if ($http_host != www.domain.com) {
    set $test  "${test}C";
}

if ($http_host != www.*.domain.com) {
    set $test  "${test}D";
}

if ($test = PCD) {
    rewrite ^(.*)$ www.%1.domain.com$1 permanent;
}

Thanks

For now I have crons like :

45 * * * * www-data php /var/www/console cmd:dispatch --env=prod 2>&1 > /var/log/cron/cmd_dispatch.log

And I'd like to disable email sending. In order to achieve that can I add &> /dev/null like :

45 * * * * www-data php /var/www/console cmd:dispatch --env=prod 2>&1 > /var/log/cron/cmd_dispatch.log; &> /dev/null

Would it disable only the email sending ?

I can't access to sub-subdomains.

For example http://electricien.synerciel.fr is working.

But, for sub-subdomains http://electricite-lemonnier.electricien.synerciel.fr/ returns DNS_PROBE_FINISHED_NXDOMAIN

Here's the DNS configuration :

I don't know what's going on. I can't find any valid resource to explain what the error message means.

I just switched from Apache to debian 7.6 nginx php-fpm and after running all kind of optimisations it seems that it takes a lot longer than usual to display images :

The application is extremely fast (PHP side), I went from 1sec load time, to <500ms, the only problem are the images takes too much time to load for no valid reason.

Examples :

50KB => 800ms recieving

or sometimes, just for an 32x32 icon, it takes 700ms "waiting" before recieving. It feels like a 56K modem displaying "progressivly" the images.

• My nginx vhost can be found here
• My nginx configuration can be found here

My guess would be that I screwed up some settings since the hardware is not suspected (2servers 8CPU Xeon, 32Gb Ram, 1Gbps network, load avg: 0.13, SSD servers)

The 2 server are load balanced, meaning that all images are randomly downloaded from one or the other server (no difference) for the same master request.

PS: Before switching to nginx, we had bad hardware, bad network and still, the images were loading a lot faster than now.

Why the image are taking so long to load? An example can be found here (http://sarl-hensgen.fr/)

I did optimize nginx with what I found on the internet My specs are 2 servers load balanced :

• Intel Xeon E5-1620v2 (4c/8t 3,7 GHz+/3,9 GHz+)
• 32 Go DDR3 ECC 1600MHz
• 3x 160Go SSD Intel DC S3500 SATA3 6Gbps (raid 1)
• 1 Gbps

My optimisation for nginx are:

• worker_processes 10;
• worker_rlimit_nofile 20000;
• keepalive_timeout 15;
• worker_connection 4096;
• The rest of the optimizations can be found here

For php5-fpm (php 5.5.x):

• pm_type: dynamic
• pm_max_children: 300
• pm_start_servers: 5
• pm_min_spare_servers: 5
• pm_max_spare_servers: 35
• pm_max_requests: 500
• pm_process_idle_timeout: 10s

Others:

• Modify the maximum opened file for debian (uopen)
• Modify the maximum connection per socket

The problem is, while on heavy load the 8CPU are 100%, the memory usage is very low (3Go used on 32Go available).

The results of 800 users hitting the application for 1 minutes is available. Caution, keep in mind that there are 2 servers serving responses.

• http://screencast.com/t/vnpRRk12Ry8
• http://screencast.com/t/W126GJPYZH7

Is there any other tuning options for nginx or php-fpm to allow better performances ? The low memory used kind of tells me i'm doing something wrong.

On my staging server, I have Varnish + Nginx. They are on port 80 and 8080.

Given that I have to install jenkins, on which port I have to bind it ? (which is usually 8080 but it's already taken)

My configuration is very simple :

To avoid duplicate content, I want to prevent user to go on my website by hitting directly apache (which is running on port 8080).

I have setup a Varnish server listening on port 80, so I want to use only this to avoid bot indexing the same website on different port which may cause duplicate content issue.

I'm using a dedicated server with Debian 6.

My virtual host looks like :

<VirtualHost *:8080>
    ServerAdmin webmaster@localhost
    ServerName www.seek-team.com

    DocumentRoot ...
    DirectoryIndex app.php

    <Directory "/var/www/seek-team.com/current/web">
        Options -Indexes FollowSymLinks SymLinksifOwnerMatch
        AllowOverride All
        Allow from All
    </Directory>
</VirtualHost>

How to prevent user to directly access to the website using port 8080 ? (but I still need varnish to hit apache correctly).

Thanks.

I need to access to my website on port :8080 But for a reason, I can't :

That's what looks like my iptable -L https://gist.github.com/28e7a48d91e933c6f377

After searching on serverfault, I've just added :

iptables -I INPUT 1 -i lo -p tcp --dport 8080 -j ACCEPT 

With no effet (even after a iptable restart)

How to allow access of the port 8080 please (both locally and externally because of the setting i'm going to put for varnish / apache2) ?

EDIT it seems that after a restart of iptable, the rule of allowing the port 8080 has gone:

after restart.

Thanks.

I was interested in installing the SPDY mod for apache2, in order to improve the speed / latency of my website.

My major question is : does all the websites on my dedicated server will have to switch to SSL, or will SPDY be enabled only for the SSL enabled website ?

I hope it'll not mess up all my production website.

i have a dedicated server (on which <10 websites from 10 visitors to 500 visitors per day) :

PIV 2,8HT
2Go RAM
1x80 Go SATA 100 Mbs mutualized / 10 Mbs dedicated

for the same price I have another possibility to switch on :

Atom Dual Core 510 HT (1,66 Ghz)
4 Go DDR2
250 Go SATA
BP 100 Mbs fullduplex

What do you think ? What would you choose : stay on my P4 dedicated server, or switch on the other server ? I'm a little afraid of the 1,66 GHz CPU. Am I wrong ?

Thanks

on my dedicaced server, i host multiple website.

Simple question : what do i need to enter in /etc/hostname ?

Why choose

website1.com instead of website2.com ?

What do i need to put in there ?

Thanks

All is in the title,

GA : 260 visitors for one day (30/04/2010) Urchin 6 : 750 visitors (30/04/2010)

Then, who believe ?

Thanks

could you pelase tell me what's the difference (i never heard of lighttpd before) ? pro / cons ?

and what would you pick for a website who have to deal with a lot of querys (like 20,000 min per day) ?

Thanks