helion-prime
home about us blogs contacts

Posts Tagged ‘RoR’

Deployment of Ruby on Rails applications on OpenBSD

Monday, April 5th, 2010

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 :
deployment diagram

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

Thursday, February 25th, 2010

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

Thursday, November 19th, 2009

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!

RubyConf2008 conference

Tuesday, January 6th, 2009

If you are real Ruby follower you surely know about RubyConf2008 conference that was held in November 2008, or perhaps you even participated. If you still miss all the fun you can go thru whole list of videos from that conference on [http://rubyconf2008.confreaks.com/].

Or at least you can watch videos I’ve selected for busiest men:


name: Reasons behind Ruby
duration: 31 minutes
description: introductory speech from Yukihiro Matsumoto, father of Ruby where he speaks about Ruby, its future, and community in general without any technical stuff.
URL: [http://rubyconf2008.confreaks.com/matzs-keynote.html]


name: Fork Ruby
duration: 48 minutes
description: speech from Dave Thomas that helped write Agile Manifesto, and Programming Ruby: A Pragmatic Programmer’s Guide. He share thoughts about possible ideas for Ruby language and its development.
URL: [http://rubyconf2008.confreaks.com/keynote.html]


name: Ruby 1.9: What to Expect
duration: 50 minutes
description: Dave Thomas and David Black author of popular book Ruby for Rails, Ruby core contributor and the creator and maintainer of RCRchive show to us differences between Rubys with 2 irb windows step by step.
URL: [http://rubyconf2008.confreaks.com/ruby-19-what-to-expect.html]


name: Recovering from Enterprise
duration: 45 minutes
description: Comparison between Ruby and Java worlds from Ruby, and Ruby on Rails contributer Jamis Buck. He describe fundamental differences between Java and Ruby and his mistakes, and ways to avoid them. Also provide ideas how to write in real Ruby and not in Java with Ruby syntax.
URL: [http://rubyconf2008.confreaks.com/recovering-from-enterprise.html]





Next conference will be held on 13-14 March of 2009 in Salt Lake City.
See details on official site: [http://mtnwestrubyconf.org/]


©2010 Helion-Prime Solutions Ltd.
Custom Software Development Agile Company.