Helion-Prime Solutions blog » ruby

Deployment of Ruby on Rails applications on OpenBSD

alex.shapovalov — Mon, 05 Apr 2010 12:35:56 +0000

Preamble

In this post I will define typical production environment on OpenBSD OS for deployment of Ruby on Rails applications.

There are few common things:
1. The Ruby on Rails framework doesn’t support concurrent running in multiple threads within the same process, and so to scale it and fully utilize available hardware we need to execute application in several processes.
2. We need load-balancer to spread incoming requests between application instances.
3. We need separate web-server to serve static content.

Overall configuration

At first all incoming HTTP requests from a clients come to httpd web-server, it servers all static content, and send other requests to HAProxy.
HAProxy receives requests and selects free Thin instance, forwards the request to it, receives a response and passes it back to httpd.

Following diagram should give you basic understanding about common work of components :

OpenBSD Httpd – standard OpenBSD web-server

I suggest OpenBSD standard web-server as you can find it as part of OpenBSD base installation, it checked for security issues and being updated as part of OpenBSD. We will use it to serve static content and don’t bother our Thin servers.

file: httpd.conf

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35

BindAddress SERVER_IP_ADDRESS

# Dynamic Shared Object (DSO) Support
# caching proxy
LoadModule proxy_module /usr/lib/apache/modules/libproxy.so

# allow Symbolic Links for root of our static content and all sub-directories

Options +FollowSymLinks
ServerAdmin ADMIN_EMAIL

# path to root of our static content
DocumentRoot /var/www/railsdocs/RAILS_PROJECT/public
ServerName SERVER_NAME
ServerAlias www.SERVER_NAME

# directories that contain static content (they excluded from dispatching to HAProxy)
ProxyPass /images !
ProxyPass /stylesheets !
ProxyPass /javascripts !
ProxyPass /500.html !
ProxyPass /503.html !

# address where to send and from receive requests (HAProxy listens that address)
ProxyPass / http://127.0.0.1:4000/
ProxyPassReverse / http://127.0.0.1:4000/

# Disallows remote servers to be mapped into the space of the local server.
ProxyRequests Off

# Don't use incoming Host HTTP request header for proxy request.
ProxyPreserveHost Off

ErrorLog logs/SERVER_NAME-error_log
CustomLog logs/SERVER_NAME-access_log common

see for configuration details: man httpd

For OpenBSD 4.6/4.6 -Stable
It’s the hard part, OpenBSD4.6 has a bug in mod_proxy module so ‘!’ directive doesn’t work.
You have to edit following file: /usr/src/usr.sbin/httpd/src/modules/proxy/mod_proxy.c

Find method: static int proxy_trans(request_rec *r)
in that method after condition: if (len > 0) {
add 2 string:

1 2	if (ent[i].real[0] == '!' && ent[i].real[1] == '\0') return DECLINED;

so final part of code:
…

1
2
3

if (len > 0) {
if (ent[i].real[0] == '!' && ent[i].real[1] == '\0')
return DECLINED;

…

Then recompile your system, it’s common procedure for following -stable so you should already know it otherwise see for details for building instructions: http://www.openbsd.org/faq/faq5.html

Thin – high performance ruby web server

We need some Ruby web-server, and it seems that at this time Thin provides best performance.
At least we see such results on Thin homepage: http://code.macournoyer.com/thin/

file: start.sh

1 2	# start work instances thin start -C thin-production.yml

file: thin-production.yml

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

---
environment: production

port: 4001
address: 127.0.0.1
daemonize: true
servers: 4

chdir: /var/www/railsdocs/RAILS_PROJECT
pid: tmp/pids/thin.pid
log: log/thin.log

user: myuser
group: mygroup

require: []

see for configuration details: http://code.macournoyer.com/thin/usage/

HAproxy – TCP/ HTTP load balancer

As Rails doesn’t support concurrent running each incoming request should be assigned to a separate process. HAProxy can be configured to send only one request at a time to every Thin server, it will always pick instance that is not busy with something.

It provides bunch of other useful things like:
– route HTTP requests depending on statically assigned cookies ;
– switch to backup servers in the event a main one fails ;
– accept connections to special ports dedicated to service monitoring ;
– add/modify/delete HTTP headers both ways ;
– block requests matching a particular pattern ;
for full documentation see: http://haproxy.1wt.eu/#docs

note:
If someone thinks that we could use nginx for that purpose check following performance comparison of HAProxy and Nginx:
http://affectioncode.wordpress.com/2008/06/28/another-comparison-of-haproxy-and-nginx/

file: haproxy.cfg

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45

defaults
log global
mode http

# provides more detailed information about HTTP contents, such as the request and some cookies
option httplog
# do not to log any session which didn't transfer any data
option dontlognull
# allow the proxy to break their persistence and redistribute connections in case of failure
option redispatch

# the number of attempts to reconnect after a connection failure to a server
retries 3

# the time we accept to wait for a connection to establish on a server
contimeout 100000
# the time we accept to wait for data from the client, or for the client to accept data
clitimeout 100000
# the time we accept to wait for data from the server, or for the server to accept data
srvtimeout 100000

listen project_proxy 127.0.0.1:4000
balance roundrobin

# creates an HTTP 'X-Forwarded-For' header which contains the client's IP address.
# This is useful to let the final web server know what the client address was
option forwardfor

# using “maxconn 1″ improves performance with Rails.
# As Rails instance can process only 1 request “maxconn 1″ force HAProxy to select next free instance

server app1_1 127.0.0.1:4001 check inter 60000 rise 2 fall 5 maxconn 1
server app1_2 127.0.0.1:4002 check inter 60000 rise 2 fall 5 maxconn 1
server app1_3 127.0.0.1:4003 check inter 60000 rise 2 fall 5 maxconn 1
server app1_4 127.0.0.1:4004 check inter 60000 rise 2 fall 5 maxconn 1

# httpd web-server will handle it in case of 503, 504 errors due to it's static content
errorloc 503 http://DOMAIN_NAME/503.html
errorloc 504 http://DOMAIN_NAME/504.html

# statistics page thru http://127.0.0.1:8080
listen stats 127.0.0.1:8080
balance roundrobin
mode http
stats uri /

see for configuration details: http://haproxy.1wt.eu/#docs

Tuning of Postgresql under OpenBSD

alex.shapovalov — Thu, 25 Feb 2010 15:04:00 +0000

Preamble

I assume that you already made your best with help of your favorite programming language
and recommended postgresql performance tips: http://www.postgresql.org/docs/8.4/static/performance-tips.html

Postgresql resources

No doubt that standard postgresql configuration is far from modern production environments.
Therefore you need to spend enough time with following sources.

1. resource consumption documentation:
http://www.postgresql.org/docs/8.4/static/runtime-config-resource.html

The most important parameters are:
work_mem (integer)
shared_buffers (integer)

2. Query Planning documentation:
http://www.postgresql.org/docs/current/static/runtime-config-query.html

The most important parameters are:
effective_cache_size (integer)
random_page_cost (floating point)

OpenBSD resources

The default sizes in the GENERIC kernel are insignificant also and waiting for your tuning as well.
Posgtresql doesn’t start without enough memory size so always know when you need to increase kern.shminfo.shmmax.

Setting that we can change in /etc/sysctl.conf
the maximum number of System V IPC system-wide semaphore sets (and identifiers) which can exist at any given time:
kern.seminfo.semmni

the maximum total individual System V IPC semaphores which can be assigned by applications:
kern.seminfo.semmns

the amount of shared memory available in the system (bytes):
kern.shminfo.shmmax

the maximum number of shared memory segments:
sysctl kern.shminfo.shmseg

Full list of setting you can see with:
# man sysctl

OpenBSD kernel parameters
So, there are set of parameters that can be tuned only with kernel rebuild.

You should tune them only if system works unstable with default values and you have:
kernel warnings: “uvm_mapent_alloc: out of static map entries”
or panics like: “panic: malloc: out of space in kmem_map”

NKMEMPAGES
This option defines number of pages in kernel kmem_map structure.

MAX_KMAPENT
It defines number of static entries in kernel kmem_map (kernel virtual memory).

They can be changed in:
/usr/src/sys/arch/conf/GENERIC

As start you need to recheck ‘Building the System from Source’ part of OpenBSD documentation:
http://openbsd.org/faq/faq5.htm

Usually administrators select these parameters using set of tests on dedicated testing box where
they emulate load of production servers.

Example

our test server: 1x Intel Quad core CPU, 2GB RAM
software: Ruby on Rails application, postgresql DB, memcached.
load: about 15.000 users/day, peak load: 10 users/sec.

postgresql_dir/data/postgresql.conf

1
2
3
4
5
6
7
8
9

# RESOURCE USAGE
shared_buffers = 738MB
max_prepared_transactions = 30
work_mem = 16MB
max_fsm_pages = 2000000

# QUERY TUNING
effective_cache_size = 512MB
random_page_cost = 1.7

/etc/sysctl.conf

1
2
3

kern.seminfo.semmni = 256
kern.seminfo.semmns = 2048
kern.shminfo.shmmax = 805306368 # Shared memory segment size is 768M

/usr/src/sys/arch/conf/GENERIC

1
2
3

## custom settings
option MAX_KMAPENT = 3072
option NKMEMPAGES = 32768

Log rotation in Ruby on Rails applications

alex.shapovalov — Thu, 19 Nov 2009 12:30:04 +0000

preamble

Even on mid-size applications logs grow enormously fast, it’s a pity Rails doesn’t provide built-in functionality for that like Apache web-server does. Fortunately there is an easy and fast solution for that issue.

1. Install cronolog

“cronolog [http://cronolog.org/] is a simple filter program that reads log file entries from standard input and writes each entry to the output file specified by a filename template and the current date and time.”

For details, and list of parameters see: [http://cronolog.org/usage.html]

On GNU/Linux Debian:

# apt-get install cronolog

2. Setup your Rails application

Edit necessary environment file: like ‘config/environments/production.rb’ file

1
2
3
4
5

# system logger configuration
# config.logger = SyslogLogger.new

config.logger = Logger.new(IO.popen("FULL_PATH_TO_CRONOLOG log/production.log.%Y%m%d", "w"))
config.logger.level = Logger::INFO

That’s all, fast and easy .. Enjoy!

Easy image scaling for RoR developers

alex.shapovalov — Mon, 02 Nov 2009 16:21:40 +0000

We have common task: image scaling [http://en.wikipedia.org/wiki/Image_scaling], and correct understanding that: we can’t ask users to provide image of appropriate sizes, just use width and height CSS parameters for images, and thumbnail is always required feature.

Luckily we have ImageMagic library [http://www.imagemagick.org] and program interfaces for popular languages.

install RMagic gem
# gem install rmagick

add small code:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30

require 'rubygems'
require 'RMagick'

class ImageController < ApplicationController

def index

# determine basic params
w = params[:w]
h = params[:h]
path = params[:path]

if !w.to_s.empty? && !h.to_s.empty? && File.exists?(path)
begin
image = Magick::Image.read(path).first
transformed_image = image.resize_to_fill(w.to_i, h.to_i)

# send image
send_data(transformed_image.to_blob, :disposition => 'inline',
:type => "image/#{image.format.downcase}")

rescue Exception => e
Rails.logger.error e.message
Rails.logger.error e.backtrace.join("\n")
render :text => "Error processing image #{type}/#{aliaz}", :code => 404, :layout => false
end

end

end

Here to re-size images we use ‘resize_to_fill’ ImageMagic method:
re-size the image to fit within the specified dimensions while retaining the aspect ratio of the original image, and if necessary, crop the image in the larger dimension.

Sure you understand that resizing takes time and so you should cache re-sized images.

For additional info look thru:
[http://studio.imagemagick.org/RMagick/doc/]

Multi-domain applications in Ruby on Rails

alex.shapovalov — Thu, 03 Sep 2009 14:34:58 +0000

preamble

Ruby on Rails is a great framework that still luck some common features, among them: multi-domain support.
Here I will describe fast solution that doesn’t work for every browser and another one that do the work.

Ok, you did your homework and google something like ‘ruby on rails multi-domain’.
Very often provided solution:

edit environment.rb

1	config.action_controller.session = {:domain => '.mydomain.com'}

With that parameter Rails always read cookies from same domain. In real some browsers forbid for applications to read cookies from other domain due to insecurity of that operation. And as Mozilla Firefox in set of browsers that forbid that we just need another solution.

Now it’s clear that we should implement necessary functionality other way.

Here we go:

1. will store session ID in the database(by default Rails2 store it in cookies)

environment.rb:

1	config.action_controller.session_store = :active_record_store

and then execute rake task that create necessary DB migration:

1	rake db:sessions:create

Apply it with: rake db:migrate.

2. setup session parameters

environment.rb:

1
2
3
4
5
6
7
8
9

require "rubygems"
require "active_support"

config.action_controller.session = {
:session_key => '_myapp_session_id', # session identification key
:secret => '89sHslddfsd98klasdKd', # hash code of session generator(make it random and longer)
:cookie_only => false,
:expire_after => 1.week, # TTL (time to live)
}

3. let’s create session creation handler

application_controller.rb:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

before_filter { |controller|
opts = controller.request.session_options
key = opts[:key]

# use session ID if it's passed
if controller.params[key]
# session initialization with old ID
controller.session[:nothing]

opts[:id] = controller.params[key]
controller.request.session_options = opts

# session initialization with new ID
controller.session[:nothing]
end
}

4. session pass

We need to pass session ID as parameter when user change domain name, and within domain we still use cookie.

1	“#{request.session_options[:key]}” => request.session_options[:id]

It’s wise to use session initialization before request.session_options[:id] invocation due to Rails use lazy loading, and session can be uninitialized.
Use something like: session[:nothing].

Pay attention that information that Rails get from cookie has bigger priority, and so if you have some parameter in cookie Rails will use it firstly.

Note:
An attacker can still steal you session ID by sniffing the network, or exploiting javascript, he/she gets the value from the cookie itself. If you care about security so much just use HTTPS.