International PHP Conference - Fall 2012 : Varnish Cache

Mike Willbanks | Barnes & Noble Varnish Cache

Housekeeping… •  Talk –  Slides will be posted after the

talk. •  Me –  Sr. Web Architect Manager at NOOK Developer –  Prior MNPHP Organizer –  Open Source Contributor –  Where you can find me: •  Twitter: mwillbanks G+: Mike Willbanks •  IRC (freenode): mwillbanks Blog: http://blog.digitalstruct.com •  GitHub: https://github.com/mwillbanks

Agenda •  Varnish? •  The Good : Getting Started • 

The Awesome : General Usage •  The Crazy : Advanced Usage •  Gotchas

WHAT IS VARNISH? Official Statement What it does General use

case

Official Statement “Varnish is a web application accelerator. You install

it in front of your web application and it will speed it up significantly.”

None

None

None

You can cache… Both dynamic and static files and contents.

A Scenario •  System Status Server – Mobile apps check current

status. – If the system is down do we communicate? – If there are problems do we communicate? – The apps and mobile site rely on an API •  Trouble in paradise? Few and far in between.

The Graph - AWS 0 10000 20000 30000 40000 50000

60000 70000 80000 Small X-Large Small Varnish Requests Requests 0 50 100 150 200 250 300 350 400 450 500 Small X-Large Small Varnish Time Time 0 2 4 6 8 10 12 14 Small X-Large Small Varnish Peak Load Peak Load 0 100 200 300 400 500 600 700 Small X-Large Small Varnish Req/s Req/s

The Raw Data Small   X-­‐Large   Small  Varnish  

Concurrency   10   150   150   Requests   5000   55558   75000   Time   438   347   36   Req/s   11.42   58   585   Peak  Load   11.91   8.44   0.35   Comments   19,442  failed   requests  

Traditional LAMP Stack HTTP Server Cluster Database Load Balancer

LAMP + Varnish * Varnish can act as a load

balancer. HTTP Server Cluster Database Varnish Cache Load Balancer Cache Hit Yes No

THE GOOD – JUMP START Installation General Information Default VCL

Installation rpm --nosignature -i http://repo.varnish-cache.org/redhat/varnish-3.0/el5/ noarch/varnish-release-3.0-1.noarch.rpm yum install varnish curl

http://repo.varnish-cache.org/debian/GPG-key.txt | sudo apt-key add - echo "deb http://repo.varnish-cache.org/ubuntu/ lucid varnish-3.0" | sudo tee -a /etc/apt/sources.list sudo apt-get update sudo apt-get install varnish git clone git://git.varnish-cache.org/varnish-cache cd varnish-cache sh autogen.sh ./configure make && make install

Varnish Daemon •  varnishd –  -a address[:port] listen for client

–  -b address[:port] backend requests –  -T address[:port] administration http –  -s type[,options] storage type (malloc, file, persistence) –  -P /path/to/file PID file –  Many others; these are generally the most important. Generally the defaults will do with just modification of the default VCL (more on it later).

•  varnishd -a :80 \ -T localhost:6082 \ -f /path/to/default.vcl

\ -s malloc,512mb •  Web server to listen on port 8080 General Configuration

Setup a backend! backend default { .host = “127.0.0.1” .port

= “8080” }

So what’s actually caching? •  Any requests containing – GET /

HEAD – TTL > 0 •  What cause it to miss? – Cookies – Authentication Headers – Vary “*” – Cache-control: private

Request Response vcl_recv req. vcl_hash vcl_hit vcl_miss vcl_fetch vcl_deliver req.

req. obj. resp. req. bereq. beresp. req. bereq. vcl_pipe req. bereq. vcl_pass req. bereq.

HTTP Caching •  RFC 2616 HTTP/1.1 Headers – Expiration •  Cache-Control

•  Expires – Validation •  Last Modified •  If-Modified-Since •  ETag •  If-None-Match

TTL Priority •  VCL – beresp.ttl •  Headers – Cache-control: s-max-age – Cache-control:

max-age – Expires – Validation

Use Wordpress? backend default { .host = "127.0.0.1“; .port =

"8080"; } sub vcl_recv { if (!(req.url ~ "wp-(login|admin)")) { unset req.http.cookie; } } sub vcl_fetch { if (!(req.url ~ "wp-(login|admin)")) { unset beresp.http.set-cookie; } }

THE AWESOME – VCL, DIRECTORS AND MORE VCL Directors Verifying

VCL

Varnish Configuration Language •  VCL State Engine –  Each Request

is Processed Separately & Independently –  States are Isolated but are Related –  Return statements exit one state and start another –  VCL defaults are ALWAYS appended below your own VCL •  VCL can be complex, but… –  Two main subroutines; vcl_recv and vcl_fetch –  Common actions: pass, hit_for_pass, lookup, pipe, deliver –  Common variables: req, beresp and obj –  More subroutines, functions and complexity can arise dependent on condition.

Request Response vcl_recv req. vcl_hash vcl_hit vcl_miss vcl_fetch vcl_deliver req.

req. obj. resp. req. bereq. beresp. req. bereq. vcl_pipe req. bereq. vcl_pass req. bereq.

VCL - Process VCL Process Description vcl_init Startup routine (VCL

loaded, VMOD init) vcl_recv Beginning of request, req is in scope vcl_pipe Client & backend data passed unaltered vcl_pass Request goes to backend and not cached vcl_hash Creates cache hash, call hash_data for custom hashes vcl_hit Called when hash found in cache vcl_miss Called when hash not found in cache vcl_fetch Called to fetch data from backend vcl_deliver Called prior to delivery of response (excluding pipe) vcl_error Called when an error occurs vcl_fini Shutdown routine (VCL unload, VMOD cleanup)

VCL – Variables •  Always Available –  now – epoch

time •  Backend Declarations –  .host – hostname / IP –  .port – port number •  Request Processing –  client – ip & identity –  server – ip & port –  req – request information •  Backend –  bereq – backend request –  beresp – backend response •  Cached Object –  obj – Cached object, can only change .ttl •  Response –  resp – response information

VCL - Functions VCL Function Description hash_data(string) Adds a string

to the hash input regsub(string, regex, sub) Substitution on first occurrence regsuball(string, regex, sub) Substitution on all occurrences ban(expression) Ban all items that match expression ban(regex) Ban all items that match regular expression

DEFAULT VCL Walking through the noteworthy items. Request Response vcl_recv

req. vcl_hash vcl_hit vcl_miss vcl_fetch vcl_deliver req. req. obj. resp. req. bereq. beresp. req. bereq. vcl_pipe req. bereq. vcl_pass req. bereq.

None

vcl_recv •  Received Request •  Only GET & HEAD by

default – Safest way to cache! •  Will use HTTP cache headers. •  Cookies or Authentication Headers will bust out of the cache.

None

vcl_hash •  Hash is what we look for in the

cache. •  Default is URL + Host – Server IP used if host header was not set; in a load balanced environment ensure you set this header!

None

vcl_fetch •  Fetch retrieves the response from the backend. • 

No Cache if… – TTL is not set or not greater than 0. – Vary headers exist. – Hit-For-Pass means we will cache a pass through.

GENERAL ADJUSTMENTS Common adjustments to make.

Cache Static Content No reason that static content should not

be cached.

Remove GA Cookies GA cookies will cause a miss; remove

them prior to going to the backend.

Allow Purging Only allow from localhost or trusted server network.

DIRECTORS Leveraging backend servers

Directors – The Types Director Type Description Random Picks based

on random and weight. Client Picks based on client identity. Hash Picks based on hash value. Round Robin Goes in order and starts over DNS Picks based on incoming DNS host, random OR round robin. Fallback Picks the first “healthy” server.

Director - Probing •  Backend Probing •  Variables –  .url

–  .request –  .window –  .threshold –  .intial –  .expected_response –  .interval –  .timeout

Load Balancing Implementing a simple varnish load balancer. Varnish does

not handle SSL termination.

Grace Mode Request already pending for update; serve grace content.

Backend is unhealthy. Probes as seen earlier must be implemented.

Saint Mode Backend may be sick for a particular piece

of content Saint mode makes sure that the backend will not request the object again for a specific period of time.

Purging •  The various ways of purging – varnishadm – command

line utility – Sockets (port 6082) – HTTP – now that is the sexiness

Purging Examples varnishadm -T 127.0.0.1:6082 purge req.url == "/foo/bar“ telnet

localhost 6082 purge req.url == "/foo/bar telnet localhost 80 Response: Trying 127.0.0.1... Connected to localhost. Escape character is '^]'. PURGE /foo/bar HTTP/1.0 Host: bacon.org curl –X PURGE http://bacon.org/foo/bar

Distributed Purging •  curl multi-request (in php) •  Use a

message queue –  Use workers to do the leg work for you •  You will need to store a list of servers “somewhere”

Logging •  Many times people want to log the requests

to a file – By default Varnish only stores these in shared memory. – Apache Style Logs •  varnishncsa –D –a –w log.txt – This will run as a daemon to log all of your requests on a separate thread.

Logging Apache style logging using: varnishncsa -O -a -w log.txt

VERIFY YOUR VCL You likely want to ensure that your

cache is: 1.  Working Properly 2.  Caching Effectively

What is Varnish doing… Varnishtop will show you real time

information on your system. •  Use -i to filter on specific tags. •  Use -x to exclude specific tags.

Checking Statistics… Varnishstat will give you statistics you need to

know how you’re doing.

THE CRAZY ESI – Edge-Side Includes Varnish Administration VMOD

ESI – Edge Side Includes •  ESI is a small

markup language much like SSI (server side includes) to include fragments (or dynamic content for that matter). •  Think of it as replacing regions inside of a page as if you were using XHR (AJAX) but single threaded. •  Three Statements can be utilized. –  esi:include – Include a page –  esi:remove – Remove content –  <!-- esi --> - ESI disabled, execute normally

ESI Diagram Page Content <esi:include src="header.php" /> B a c

k e n d V a r n i s h Varnish detects ESI, requests from backend OR checks cached state.

Using ESI •  In vcl_fetch, you must set ESI to

be on – set beresp.do_esi = true; – Varnish refuses to parse content for ESI if it does not look like XML •  This is by default; so check varnishstat and varnishlog to ensure that it is functioning like normal.

ESI Usage <html> <head><title>Rock it with ESI</title></head> <body> <header> <esi:include

src=”header.php" /> </header> <section id="main">...</section> <footer></footer> </body> </html>

Embedding C in VCL •  Before getting into VMOD; did

you know you can embed C into the VCL for varnish? •  Want to do something crazy fast or leverage a C library for pre or post processing? •  I know… you’re thinking that’s useless.. – On to the example; and a good one from the Varnish WIKI!

Embedded C for syslog C{ #include <syslog.h> }C sub vcl_something

{ C{ syslog(LOG_INFO, "Something happened at VCL line XX."); }C } # Example with using varnish variables C{ syslog(LOG_ERR, "Spurious response from backend: xid %s request %s %s \"%s\" %d \"%s\" \"%s\"", VRT_r_req_xid(sp), VRT_r_req_request(sp), VRT_GetHdr(sp, HDR_REQ, "\005host:"), VRT_r_req_url(sp), VRT_r_obj_status(sp), VRT_r_obj_response(sp), VRT_GetHdr(sp, HDR_OBJ, "\011Location:")); }C

Varnish Modules / Extensions •  Taking VCL embedded C to

the next level •  Allows you to extend varnish and create new functions •  You could link to libraries to provide additional functionality

VMOD - std •  toupper •  tolower •  set_up_tos • 

random •  log •  syslog •  fileread •  duration •  integer •  collect

ADMINISTERING VARNISH Management Console Cache Warm up

Management Console •  varnishadm –T localhost:6062 – vcl.list – see all

loaded configuration – vcl.load – load new configuration – vcl.use – select configuration to use – vcl.discard – remove configuration

Cache Warmup •  Need to warm up your cache before

putting a sever in the queue or load test an environment? – varnishreplay –r log.txt

GOTCHAS Having Keep-Alive off No SSL Termination No persistent cache

ESI multiple fragments Cookies*

QUESTIONS? These slides will be posted to SlideShare & SpeakerDeck.

SpeakerDeck: http://speakerdeck.com/u/mwillbanks Slideshare: http://www.slideshare.net/mwillbanks Twitter: mwillbanks G+: Mike Willbanks IRC (freenode): mwillbanks Blog: http://blog.digitalstruct.com GitHub: https://github.com/mwillbanks