Voir la source de Varnish

= Varnish : Présentation =
[https://varnish-cache.org/docs/6.5/tutorial/introduction.html Documentation]

Varnish est un reverse proxy pour le web servant de cache. Il reçoit des requêtes des clients et tente de leur répondre depuis le cache; si il n'y parvient pas, il forwarde la requête vers le backend, récupère la réponse et la garde en cache. Les réponse qu'il fournit depuis son cache sont très rapides. Il décide si il doit stocker la réponse ou non quand elle vient du backend; pour cela il utilise l'en-tête HTTP ''Cache-Control''. Il y'a certaines choses qui ne vont pas en cache, comme les cookies. 

Varnish utilise le VCL (Varnish Configuration Language) pour les confs de ses stratégies. 

== Installation et test ==

L'installation se fait via:
<source>
sudo apt install varnish
</source>

Nous allons ensuite activer Varnish:

<source>
systemctl enable varnish
systemctl start varnish
</source>

Puis tester son fonctionnement. Par défaut, Varnish sert sur le port 6081 ce qu'il trouve en local sur le port 8080. Je vais donc démarrer un serveur http.server avec Python (après avoir créé un fichier index.html avec du texte dedans):

<source>
echo "test" >> index.html
python3 -m http.server 8080
</source>

En allant sur l'IP de mon serveur et sur le port 6081 avec un navigateur, je vois ma page. Ça fonctionne !

== Configuration ==
=== Backend par défaut ===
Je peux changer de port pour le backend si je le souhaite. Il faut éditer /etc/varnish/default.vcl et modifier la section suivante :

<source>
vcl 4.0;

backend default {
    .host = "127.0.0.1";
    .port = "8080";
}
</source>

Je peux changer l'hôte et le port. Je teste avec mon wiki:

<source>
backend default {
    .host = "wiki.ju.lab";
    .port = "80";
}
</source>
...et ça marche bien (après un reload de varnish).

=== Paramètres du daemon ===
On peut avoir les options du daemon avec man varnishd.
==== Changer le port de Varnish ====
On va configurer Varnish pour qu'il tourne sur le port 80.

Je commence par arrêter Varnish:
<source>
systemctl stop varnish
</source>

!LEGACY!
Puis j'ouvre /etc/default/varnish (Debian) et je modifie le bloc suivant :
<source>
DAEMON_OPTS="-a :6081 \
             -T localhost:6082 \
             -f /etc/varnish/default.vcl \
             -S /etc/varnish/secret \
             -s default,256m"
## Vers
DAEMON_OPTS="-a :80 \
             -T localhost:6082 \
             -f /etc/varnish/default.vcl \
             -S /etc/varnish/secret \
             -s default,256m"
</source>

!RECENT!
La documentation recommande pour Debian 8+ / Ubuntu 15+ de créer /etc/systemd/system/varnish.service.d/customexec.conf (/etc/sysconfig/varnish pour RHEL), puis d'y mettre:
<source>
[Service]
ExecStart=
ExecStart=/usr/sbin/varnishd -a :80 -T localhost:6082 -f /etc/varnish/default.vcl -S /etc/varnish/secret -s default,256m
</source>
Pour cela, il faut d'abord créer le dossier : mkdir /etc/systemd/system/varnish.service.d.
...puis systemctl daemon-reload et restart de varnish.

Chez moi, ça ne fonctionne pas, mais la ligne suivante fonctionne:
<source>
[Service]
ExecStart=
ExecStart=/usr/sbin/varnishd -j unix,user=vcache -F -a :80 -T localhost:6082 -f /etc/varnish/default.vcl -S /etc/varnish/secret -s default,256m
</source>
La différence:
* -j <jail[,jailoptions]>: Specify the jailing mechanism to use.
* -F: Do not fork, run in the foreground.

==== RAM allouée ====
Dans la configuration ci-dessus, la mention 
<source>
-s default,256m
</source>
contrôle la quantité de RAM allouée. Je peux remplacer 256m par autre chose.
Le default signifie que l'on utilise le storage backend par défaut (si j'ai bien compris).

=== Backends ===
[https://www.varnish-software.com/wiki/content/tutorials/varnish/sample_vclTemplate.html Doc]

Je vais repasser la conf de /etc/varnish/default.vcl à ce qu'elle était à la base:
<source>
vcl 4.0;

backend default {
  .host = "127.0.0.1";
  .port = "8080";
}
</source>

Je pourrais me contenter de mettre un cache Varnish par site web, et ça fonctionnerait bien comme ça, par exemple:
* HaProxy reçoit les requêtes (et gère le SSL) sur 443 / 80, redirige sur :
* Varnish qui écoute sur 80, qui redirige sur
* Apache / Nginx qui écoute sur 8080.

La configuration de Varnish est assez complexe et mal documentée :(

Il doit aussi être possible d'avoir un seul Varnish pour plusieurs servs, et de faire plusieurs backends sur lesquels on redirige en fonction d'ACL. Mais je lâche l'affaire pour l'instant.

=== Commandes d'administration ===
==== Varnishadm ====
Varnish a une cli d'administration:
<source>
varnishadm
</source>
...mais ma doc ne dit rien de son fonctionnement

==== Varnishlog ====
Varnish ne logue pas dans un fichier, on peut voir ses logs avec la commande varnishlog.

= Configuration avancée avec des backends =
<source>
vcl 4.1;
# Based on: https://github.com/mattiasgeniar/varnish-6.0-configuration-templates/blob/master/default.vcl

import std;
import directors;

backend nas { # Define one backend
	.host = "rzprdnas1";	# IP or Hostname of backend
	.port = "8000";			 # Port Apache or whatever is listening
	.max_connections = 1000; # That's it

	.probe = {
	.url = "/"; # short easy way (GET /)
	# We prefer to only do a HEAD /

	.interval	= 5s; # check the health of each backend every 5 seconds
	.timeout	 = 1s; # timing out after 1 second.
	.window	= 5;	# If 3 out of the last 5 polls succeeded the backend is considered healthy, otherwise it will be marked as sick
	.threshold = 3;
	}

	.first_byte_timeout	 = 300s;	 # How long to wait before we receive a first byte from our backend?
	.connect_timeout		= 5s;	 # How long to wait for a backend connection?
	.between_bytes_timeout	= 2s;	 # How long to wait between bytes received from our backend?
}

backend ext1 { # Define one backend
	.host = "rzprdext1";	# IP or Hostname of backend
	.port = "8000";			 # Port Apache or whatever is listening
	.max_connections = 1000; # That's it

	.probe = {
	.url = "/"; # short easy way (GET /)
	# We prefer to only do a HEAD /

	.interval	= 5s; # check the health of each backend every 5 seconds
	.timeout	 = 1s; # timing out after 1 second.
	.window	= 5;	# If 3 out of the last 5 polls succeeded the backend is considered healthy, otherwise it will be marked as sick
	.threshold = 3;
	}

	.first_byte_timeout	 = 300s;	 # How long to wait before we receive a first byte from our backend?
	.connect_timeout		= 5s;	 # How long to wait for a backend connection?
	.between_bytes_timeout	= 2s;	 # How long to wait between bytes received from our backend?
}

backend plex { # Define one backend
	.host = "rzprdext1";	# IP or Hostname of backend
	.port = "32400";			 # Port Apache or whatever is listening
	.max_connections = 300; # That's it

	.probe = {
	.url = "/web/index.html"; # short easy way (GET /)
	# We prefer to only do a HEAD /

	.interval	= 5s; # check the health of each backend every 5 seconds
	.timeout	 = 1s; # timing out after 1 second.
	.window	= 5;	# If 3 out of the last 5 polls succeeded the backend is considered healthy, otherwise it will be marked as sick
	.threshold = 3;
	}

	.first_byte_timeout	 = 300s;	 # How long to wait before we receive a first byte from our backend?
	.connect_timeout		= 5s;	 # How long to wait for a backend connection?
	.between_bytes_timeout	= 2s;	 # How long to wait between bytes received from our backend?
}

acl purge {
	# ACL we'll use later to allow purges
	"localhost";
	"127.0.0.1";
	"::1";
}

sub vcl_init {
	# Called when VCL is loaded, before any requests pass through it.
	# Typically used to initialize VMODs.

	new vdir = directors.fallback();
    #Ces deux backends sont ajoutés en fallback
	vdir.add_backend(nas);
	vdir.add_backend(ext1);
	# vdir.add_backend(server...);
	# vdir.add_backend(servern);
}

sub vcl_recv {
	# Called at the beginning of a request, after the complete request has been received and parsed.
	# Its purpose is to decide whether or not to serve the request, how to do it, and, if applicable,
	# which backend to use.
	# also used to modify the request

	if (req.url == "/varnishcheck") {
		return (synth(200, "health check OK!"));
	}
    #Si le host == plex, on utilise plex
	if (req.http.Host == "plex.rznet.fr") {
		set req.backend_hint = plex;
	}
	else {
		set req.backend_hint = vdir.backend(); # send all traffic to the vdir director
	}

	# Normalize the header if it exists, remove the port (in case you're testing this on various TCP ports)
	if (req.http.Host) {
	 set req.http.Host = regsub(req.http.Host, ":[0-9]+", "");
	}

	# Remove the proxy header (see https://httpoxy.org/#mitigate-varnish)
	unset req.http.proxy;

	# Normalize the query arguments
	set req.url = std.querysort(req.url);

	# Allow purging
	if (req.method == "PURGE") {
	if (!client.ip ~ purge) { # purge is the ACL defined at the begining
		# Not from an allowed IP? Then die with an error.
		return (synth(405, "This IP is not allowed to send PURGE requests."));
	}
	# If you got this stage (and didn't error out above), purge the cached result
	return (purge);
	}

	# Only deal with "normal" types
	if (req.method != "GET" &&
		req.method != "HEAD" &&
		req.method != "PUT" &&
		req.method != "POST" &&
		req.method != "TRACE" &&
		req.method != "OPTIONS" &&
		req.method != "PATCH" &&
		req.method != "DELETE") {
	/* Non-RFC2616 or CONNECT which is weird. */
	return (pipe);
	}

	# Implementing websocket support (https://www.varnish-cache.org/docs/4.0/users-guide/vcl-example-websockets.html)
	if (req.http.Upgrade ~ "(?i)websocket") {
	return (pipe);
	}

	# Only cache GET or HEAD requests. This makes sure the POST requests are always passed.
	if (req.method != "GET" && req.method != "HEAD") {
	return (pass);
	}

	# Some generic URL manipulation, useful for all templates that follow
	# First remove the Google Analytics added parameters, useless for our backend
	if (req.url ~ "(\?|&)(utm_source|utm_medium|utm_campaign|utm_content|gclid|cx|ie|cof|siteurl)=") {
	set req.url = regsuball(req.url, "&(utm_source|utm_medium|utm_campaign|utm_content|gclid|cx|ie|cof|siteurl)=([A-z0-9_\-\.%25]+)", "");
	set req.url = regsuball(req.url, "\?(utm_source|utm_medium|utm_campaign|utm_content|gclid|cx|ie|cof|siteurl)=([A-z0-9_\-\.%25]+)", "?");
	set req.url = regsub(req.url, "\?&", "?");
	set req.url = regsub(req.url, "\?$", "");
	}

	# Strip hash, server doesn't need it.
	if (req.url ~ "\#") {
	set req.url = regsub(req.url, "\#.*$", "");
	}

	# Strip a trailing ? if it exists
	if (req.url ~ "\?$") {
	set req.url = regsub(req.url, "\?$", "");
	}

	# Some generic cookie manipulation, useful for all templates that follow
	# Remove the "has_js" cookie
	set req.http.Cookie = regsuball(req.http.Cookie, "has_js=[^;]+(; )?", "");

	# Remove any Google Analytics based cookies
	set req.http.Cookie = regsuball(req.http.Cookie, "__utm.=[^;]+(; )?", "");
	set req.http.Cookie = regsuball(req.http.Cookie, "_ga=[^;]+(; )?", "");
	set req.http.Cookie = regsuball(req.http.Cookie, "_gat=[^;]+(; )?", "");
	set req.http.Cookie = regsuball(req.http.Cookie, "utmctr=[^;]+(; )?", "");
	set req.http.Cookie = regsuball(req.http.Cookie, "utmcmd.=[^;]+(; )?", "");
	set req.http.Cookie = regsuball(req.http.Cookie, "utmccn.=[^;]+(; )?", "");

	# Remove DoubleClick offensive cookies
	set req.http.Cookie = regsuball(req.http.Cookie, "__gads=[^;]+(; )?", "");

	# Remove the Quant Capital cookies (added by some plugin, all __qca)
	set req.http.Cookie = regsuball(req.http.Cookie, "__qc.=[^;]+(; )?", "");

	# Remove the AddThis cookies
	set req.http.Cookie = regsuball(req.http.Cookie, "__atuv.=[^;]+(; )?", "");

	# Remove a ";" prefix in the cookie if present
	set req.http.Cookie = regsuball(req.http.Cookie, "^;\s*", "");

	# Are there cookies left with only spaces or that are empty?
	if (req.http.cookie ~ "^\s*$") {
	unset req.http.cookie;
	}

	#if (req.http.Cache-Control ~ "(?i)no-cache") {
	#if (client.ip ~ purge) {
		# Ignore requests via proxy caches and badly behaved crawlers
		# like msnbot that send no-cache with every request.
		#if (! (req.http.Via || req.http.User-Agent ~ "(?i)bot" || req.http.X-Purge)) {
		#set req.hash_always_miss = true; # Doesn't seems to refresh the object in the cache
		#return(purge); # Couple this with restart in vcl_purge and X-Purge header to avoid loops
		#}
	#}
	#}

	# Large static files are delivered directly to the end-user without
	# waiting for Varnish to fully read the file first.
	# Varnish 4 fully supports Streaming, so set do_stream in vcl_backend_response()
	if (req.url ~ "^[^?]*\.(7z|avi|bz2|flac|flv|gz|mka|mkv|mov|mp3|mp4|mpeg|mpg|ogg|ogm|opus|rar|tar|tgz|tbz|txz|wav|webm|xz|zip)(\?.*)?$") {
	unset req.http.Cookie;
	return (hash);
	}

	# Remove all cookies for static files
	# A valid discussion could be held on this line: do you really need to cache static files that don't cause load? Only if you have memory left.
	# Sure, there's disk I/O, but chances are your OS will already have these files in their buffers (thus memory).
	# Before you blindly enable this, have a read here: https://ma.ttias.be/stop-caching-static-files/
	if (req.url ~ "^[^?]*\.(7z|avi|bmp|bz2|css|csv|doc|docx|eot|flac|flv|gif|gz|ico|jpeg|jpg|js|less|mka|mkv|mov|mp3|mp4|mpeg|mpg|odt|otf|ogg|ogm|opus|pdf|png|ppt|pptx|rar|rtf|svg|svgz|swf|tar|tbz|tgz|ttf|txt|txz|wav|webm|webp|woff|woff2|xls|xlsx|xml|xz|zip)(\?.*)?$") {
	unset req.http.Cookie;
	return (hash);
	}

	# Send Surrogate-Capability headers to announce ESI support to backend
	set req.http.Surrogate-Capability = "key=ESI/1.0";

	if (req.http.Authorization) {
	# Not cacheable by default
	return (pass);
	}

	return (hash);
}

sub vcl_pipe {
	# Called upon entering pipe mode.
	# In this mode, the request is passed on to the backend, and any further data from both the client
	# and backend is passed on unaltered until either end closes the connection. Basically, Varnish will
	# degrade into a simple TCP proxy, shuffling bytes back and forth. For a connection in pipe mode,
	# no other VCL subroutine will ever get called after vcl_pipe.

	# Note that only the first request to the backend will have
	# X-Forwarded-For set.	If you use X-Forwarded-For and want to
	# have it set for all requests, make sure to have:
	# set bereq.http.connection = "close";
	# here.	It is not set by default as it might break some broken web
	# applications, like IIS with NTLM authentication.

	# set bereq.http.Connection = "Close";

	# Implementing websocket support (https://www.varnish-cache.org/docs/4.0/users-guide/vcl-example-websockets.html)
	if (req.http.upgrade) {
	set bereq.http.upgrade = req.http.upgrade;
	}

	return (pipe);
}

sub vcl_pass {
	# Called upon entering pass mode. In this mode, the request is passed on to the backend, and the
	# backend's response is passed on to the client, but is not entered into the cache. Subsequent
	# requests submitted over the same client connection are handled normally.

	# return (pass);
}

# The data on which the hashing will take place
sub vcl_hash {
	# Called after vcl_recv to create a hash value for the request. This is used as a key
	# to look up the object in Varnish.

	hash_data(req.url);

	if (req.http.host) {
	hash_data(req.http.host);
	} else {
	hash_data(server.ip);
	}

	# hash cookies for requests that have them
	if (req.http.Cookie) {
	hash_data(req.http.Cookie);
	}

	# Cache the HTTP vs HTTPs separately
	if (req.http.X-Forwarded-Proto) {
	hash_data(req.http.X-Forwarded-Proto);
	}
}

sub vcl_hit {
	# Called when a cache lookup is successful.

	if (obj.ttl >= 0s) {
	# A pure unadultered hit, deliver it
	return (deliver);
	}

	# https://www.varnish-cache.org/docs/trunk/users-guide/vcl-grace.html
	# When several clients are requesting the same page Varnish will send one request to the backend and place the others
	# on hold while fetching one copy from the backend. In some products this is called request coalescing and Varnish does
	# this automatically.
	# If you are serving thousands of hits per second the queue of waiting requests can get huge. There are two potential
	# problems - one is a thundering herd problem - suddenly releasing a thousand threads to serve content might send the
	# load sky high. Secondly - nobody likes to wait. To deal with this we can instruct Varnish to keep the objects in cache
	# beyond their TTL and to serve the waiting requests somewhat stale content.

# if (!std.healthy(req.backend_hint) && (obj.ttl + obj.grace > 0s)) {
#	 return (deliver);
# } else {
#	 return (miss);
# }

	# We have no fresh fish. Lets look at the stale ones.
	if (std.healthy(req.backend_hint)) {
	# Backend is healthy. Limit age to 10s.
	if (obj.ttl + 10s > 0s) {
		#set req.http.grace = "normal(limited)";
		return (deliver);
	}
	} else {
	# backend is sick - use full grace
		if (obj.ttl + obj.grace > 0s) {
		#set req.http.grace = "full";
		return (deliver);
		}
	}
}

sub vcl_miss {
	# Called after a cache lookup if the requested document was not found in the cache. Its purpose
	# is to decide whether or not to attempt to retrieve the document from the backend, and which
	# backend to use.

	return (fetch);
}

# Handle the HTTP request coming from our backend
sub vcl_backend_response {
	# Called after the response headers has been successfully retrieved from the backend.

	# Pause ESI request and remove Surrogate-Control header
	if (beresp.http.Surrogate-Control ~ "ESI/1.0") {
	unset beresp.http.Surrogate-Control;
	set beresp.do_esi = true;
	}

	# Enable cache for all static files
	# The same argument as the static caches from above: monitor your cache size, if you get data nuked out of it, consider giving up the static file cache.
	# Before you blindly enable this, have a read here: https://ma.ttias.be/stop-caching-static-files/
	if (bereq.url ~ "^[^?]*\.(7z|avi|bmp|bz2|css|csv|doc|docx|eot|flac|flv|gif|gz|ico|jpeg|jpg|js|less|mka|mkv|mov|mp3|mp4|mpeg|mpg|odt|otf|ogg|ogm|opus|pdf|png|ppt|pptx|rar|rtf|svg|svgz|swf|tar|tbz|tgz|ttf|txt|txz|wav|webm|webp|woff|woff2|xls|xlsx|xml|xz|zip)(\?.*)?$") {
	unset beresp.http.set-cookie;
	}

	# Large static files are delivered directly to the end-user without
	# waiting for Varnish to fully read the file first.
	# Varnish 4 fully supports Streaming, so use streaming here to avoid locking.
	if (bereq.url ~ "^[^?]*\.(7z|avi|bz2|flac|flv|gz|mka|mkv|mov|mp3|mp4|mpeg|mpg|ogg|ogm|opus|rar|tar|tgz|tbz|txz|wav|webm|xz|zip)(\?.*)?$") {
	unset beresp.http.set-cookie;
	set beresp.do_stream = true;	# Check memory usage it'll grow in fetch_chunksize blocks (128k by default) if the backend doesn't send a Content-Length header, so only enable it for big objects
	}

	# Sometimes, a 301 or 302 redirect formed via Apache's mod_rewrite can mess with the HTTP port that is being passed along.
	# This often happens with simple rewrite rules in a scenario where Varnish runs on :80 and Apache on :8080 on the same box.
	# A redirect can then often redirect the end-user to a URL on :8080, where it should be :80.
	# This may need finetuning on your setup.
	#
	# To prevent accidental replace, we only filter the 301/302 redirects for now.
	if (beresp.status == 301 || beresp.status == 302) {
	set beresp.http.Location = regsub(beresp.http.Location, ":[0-9]+", "");
	}

	# Set 2min cache if unset for static files
	if (beresp.ttl <= 0s || beresp.http.Set-Cookie || beresp.http.Vary == "*") {
	set beresp.ttl = 120s; # Important, you shouldn't rely on this, SET YOUR HEADERS in the backend
	set beresp.uncacheable = true;
	return (deliver);
	}

	# Don't cache 50x responses
	if (beresp.status == 500 || beresp.status == 502 || beresp.status == 503 || beresp.status == 504) {
	return (abandon);
	}

	# Allow stale content, in case the backend goes down.
	# make Varnish keep all objects for 6 hours beyond their TTL
	set beresp.ttl = 1y;
	set beresp.grace = 24h;

	return (deliver);
}

# The routine when we deliver the HTTP request to the user
# Last chance to modify headers that are sent to the client
sub vcl_deliver {
	# Called before a cached object is delivered to the client.

	if (obj.hits > 0) { # Add debug header to see if it's a HIT/MISS and the number of hits, disable when not needed
	set resp.http.X-Cache = "HIT";
	} else {
	set resp.http.X-Cache = "MISS";
	}

	# Please note that obj.hits behaviour changed in 4.0, now it counts per objecthead, not per object
	# and obj.hits may not be reset in some cases where bans are in use. See bug 1492 for details.
	# So take hits with a grain of salt
	set resp.http.X-Cache-Hits = obj.hits;

	# Remove some headers: PHP version
	unset resp.http.X-Powered-By;

	# Remove some headers: Apache version & OS
	unset resp.http.Server;
	unset resp.http.X-Drupal-Cache;
	unset resp.http.X-Varnish;
	unset resp.http.Via;
	unset resp.http.Link;
	unset resp.http.X-Generator;

	return (deliver);
}

sub vcl_purge {
	# Only handle actual PURGE HTTP methods, everything else is discarded
	if (req.method == "PURGE") {
	# restart request
	set req.http.X-Purge = "Yes";
	return(restart);
	}
}

sub vcl_synth {
	if (resp.status == 720) {
	# We use this special error status 720 to force redirects with 301 (permanent) redirects
	# To use this, call the following from anywhere in vcl_recv: return (synth(720, "http://host/new.html"));
	set resp.http.Location = resp.reason;
	set resp.status = 301;
	return (deliver);
	} elseif (resp.status == 721) {
	# And we use error status 721 to force redirects with a 302 (temporary) redirect
	# To use this, call the following from anywhere in vcl_recv: return (synth(720, "http://host/new.html"));
	set resp.http.Location = resp.reason;
	set resp.status = 302;
	return (deliver);
	}

	return (deliver);
}


sub vcl_fini {
	# Called when VCL is discarded only after all requests have exited the VCL.
	# Typically used to clean up VMODs.

	return (ok);
}
</source>