I am working on an app that needs to return an image file in the controller. I send over some request parameters to the app, it goes to the controller which pulls out the params and then creates an image based on those parameters. So instead of rendering the view, the app would need to return a png file. The send_data method was designed to do exactly this. Here is the code:
data = File.open(your_image_path,’rb’).read
send_data(data , :filename => your_image_name, :type=>’image/png’)
I recently installed Nginx on another server and forgot how difficult it was the last time I did it. You need to make sure you have all the dependencies before you can configure the source. Here are the steps to get a successful install on CentOS 5 using Nginx 0.7.61.
Nginx should be installed after that. Then you can edit your config in /usr/local/nginx/conf and start the web server using:
I recently moved to a master/slave setup with MySQL. At first I was thinking.. “Oh this is great, I have a backup database now in case anything breaks on the master. “ Don’t go there.. think about it for a second. If something gets corrupted on the master, then it is going to push the changes over to the slave and then you will have 2 corrupted databases. It is not that hard to create a backup, though, using mysql_dump + shell script + crontab. That way you can have a nightly backup and reload it into MySQL if anything gets corrupted. Here is what you do:
Create a file with your favorite editor called backup.sh and add the following to it:
The 2nd line in this file sets a variable for the current date in format like this: Aug-26-09. The last line uses the mysql_dump command to output all your databases to a file that appends the current date to the file name using the NOW variable. If you did not append the date to the mysql backup, then it would get overwritten every night. NOTE: you may also want to create a delete script, so this cron does not fill up your filesystem.
Next thing you need to do is change the permission on the file so that you can execute it. I am just assigning RWX for everyone, because I am too lazy to do it the right way (you probably just want to make this executable by the crontab user)
The last thing you have to do is add it to your crontab. 0 0 * * * *, makes the cron execute it at midnight every night.
Now this script will execute every night and export your entire database. Now, dont you feel better?? Thought so.
If you are using MySQL with InnoDB tables, you should really look at the innodb_buffer_pool_size. This tells MySQL how much memory it should use to cache data on your InnoDB tables. This cuts down on disc IO. It can save you alot of lookup time if you are pulling out of the cache as opposed to making another call to the database table. As a good rule of thumb this value should be set to 10% larger than the size of your database… that is assuming you have that much memory available on your server. By default innodb_buffer_pool_size is set to 8MB. If you have a decent sized data set, and most of you probably do, then you are going to want to increase this. My database is about 1GB, so I set my innodb_buffer_pool_size to 2GB, so that I can account for some growth in the future. Just be sure not to set the value too high. If you only have 2GB of memory on your box, then don’t set it to 2GB. This can cause paging at the OS level and that is bad news.
So here is how you set that parameter in mysql
open up your /etc/my.cnf in your favorite editor and add the parameter in the mysqld section
Then you restart mysql and the changes should be picked up.
You can double check by issuing the following command in mysql client. It will show you the variable and its value
NOW THATS A QUICK AND EASY PERFORMANCE TUNE!!
I have been running a web app on 1 server since 2005 and we have seen a good bit of traffic. We are currently ranked 8200 according to quantcast.com with just over 200,000 visitors a month. Traffic is picking up and I expect it to almost double in Q4. NewRelic shows all green right now with about 250ms response time which is fine. But I am worried that if traffic doubles, the database may become a little stressed. So I have decided to be pro-active and go ahead and add another server to the database layer. Right now, I have Nginx in front of mongrel cluster (4 instances) and MySQL as the database. All of this is running on the same server. So my new setup will be:
Server 1: Nginx > Mongrel Cluster > MySQL master
Server 2: MySQL slave
NOTE: RailsEnvy has a really good video series on scaling your database (and many other great videos on other subjects), so you may want to check that out just to get an overview of the process.
In order to get this working, Rails will need to be able to perform writes only on master and reads on both master and slave. But, lets talk about setting up the replication first. Here is how you set it up:
1. Set binary logging on the master server. Edit /etc/my.conf
2. Set the server ids on master and slave in /etc/my.conf
3. Grant the slave permission to pull the data from the master (replace XX.XX.XX.XX with your slave’s IP address)
4. Perform a read-lock on the master
5. Record the file and position on the master. It should give you output like this. Record the file and position to use later
6. Export data on master using mysql_dump
7. Unlock the tables on master
8. FTP the tar file of the mysql snapshot to your slave box
9. Stop MySQL on the slave
10. You need to start slave using –skip-slave option. I changed the init.d script to do this. Probably an easier way to do it, but whatever.. It was line 61 in /etc/init.d/mysqld. Just make a backup of that file, so you can replace it after you start mysql
11. Start mysql on slave. Then change /etc/init.d/mysql back to the original (without –skip-slave option)
12. Import the data into the slave
13. Configure master on the slave box using the values for your host,user, and password. Fill in the last 2 with the values you recorded from step 5
14. Start the slave replication on the slave server
Now your slave should be getting updates from the master. Go ahead and change some stuff on the master and see if the slave replicates. You can also view the status of the slave by using this mysql command:
The next step in the process is to get rails to communicate with the master/slave setup. The easiest way to do this is to use the machoism plugin. You can get it at github. There are a few different branches of this, but I am using the one from technoweenie (I had problems getting the mislav version to work)
http://github.com/technoweenie/masochism/tree/master
It is easy to install. Just go to your rails app root directory and issue the following command
Next you need to configure your database.yml to for masochism. One interesting thing I found out that I was not aware of is that you can set the login as a variable, so you can reuse it later in the config file. Here is my config
# setup for masochism (master)
master_database:
<<: *login
host: XX.XX.XX.XX
Then add the following lines to your environment.rb file
# in environment.rb
config.after_initialize do
if Rails.env.production?
ActiveReload::ConnectionProxy::setup!
end
endRestart your app server and you should be good to go.
I thought I was until I figured out that there was a conflict between masochism and another plugin I use called use_db. Masochism seems to be overriding the database config for use_db, so it is restricting me to only the master/slave databases. I will post when I figure out a solution to this.
I have 2 database servers. One server runs a script that updates about 8 million products. This takes a long time 
And I can’t use MySQL master/slave because the data is not production worthy until the script fully completes. Instead of using Master/Slave, I push the binary files from database server 1 to database server 2 when the script completes. So each of the databases will be turned off at some point. Therefore, I need to use HAProxy to forward the requests to the working server when the other one is turned off. And when the product script is not running and binary files are not being pushed, I want both of them running in the cluster. So here is application flow:
web server requests -> haproxy -> database servers 1 & 2
In order to make this work we need HAProxy installed and we need a health check script for both databases, so HAProxy knows when one is not running (no requests will be forwarded to this server). Here are the steps:
First I need to install haproxy on my web server (CentOS Linux), so that it can forward the requests to the 2 MySQL instances. Go here:
Download the latest stable source and make it.
You dont have to ./configure and make. Just “make install”. Next step is to set up a config file. Most people choose /etc/haproxy.cf
Here is a what my final config looks like:
global
log 127.0.0.1 daemon debug
stats socket /tmp/stats
maxconn 4096
pidfile /var/run/haproxy.pid
daemon
defaults
log global
mode tcp
option dontlognull retries 3 option redispatch
maxconn 2000
contimeout 5000
clitimeout 50000
srvtimeout 50000
listen MySQL XX.XX.XX.XX:3305
mode tcp
option httpchk
stats enable
stats uri /haproxy-stats
balance roundrobin
server mysql_slave XX.XX.XX.XX:3306 check port 9200 inter 12000 rise 3 fall 3
server mysql_slave XX.XX.XX.XX:3306 check port 9200 inter 12000 rise 3 fall 3
Of course, you need to replace XX.XX.XX.XX with your server IPs. The first IP in the “listen” section is the IP address where HAProxy is installed. The last 2 lines in this file are where you put the IP addresses for your database servers.
After your config is set up, you can start haproxy with this command:
Then use your SQL client (I am using SQLyog) to login to mysql using port 3305. If you are able to login, then that means that HAProxy is listening on port 3305 and forwarding the requests to the slaves.
The next thing you need to do is add some health checks, so that if one of the slaves goes down, then HAProxy will stop forwarding requests to the non-working slave. For this I referenced a tutorial written by Unai Rodriquez - Having HAProxy check mysql status through a xinetd script Essentially, you just create a xinetd service that listens on port 9200 and returns an HTTP status 200 or 500 depending on the state of the MySQL server. HAProxy stops forwarding requests if it sees a status 500.
You can test this by stopping and starting MySQL and watching http://yourhost/haproxy-stats. This will show if the servers are up or down. After you verify that it is working correctly, it is time to change your Ruby on Rails (or whatever framework you are using) configuration to use port 3305. Deploy your code, restart your app server and you should have a working mysql cluster.
Well… I have been ignoring this for a few weeks, because it is not a huge deal, but I thought I might as well fix this.
Problem: The created_at and updated_at dates do not match my system time. Every time I save a record, ActiveRecord records the date 5 hours ahead of my local time. This was messing up some of my cronjobs that run 5 minutes before midnight. Some of the records were not being found, because the date was recorded five hours in the future.
Solution: After some digging on Google, I figured out that this is set in config/environment.rb.
I did some more digging and found that UTC is 5 hours ahead of ET, so ActiveRecord was using this timezone instead of my local timezone. The first fix I tried did not work, which was to chagne config.time_zone =”Eastern Time (US & Canada)’ I restarted my app server and still saw created_at was five hours ahead of my server’s system time. So then I just commented it out and restarted the app again. Walah.. it works now. So the correct config to use system timezone is as follows:
No more worries
I am starting this blog, because I thought it might be useful for some Rails developers out there. Other rails bloggers have saved my ass plenty of times, so why not write about my problems & solutions and maybe I can help out some other frustrated developer. I own a web startup called CouponShack.com. We have been around for a while and we just finished rewriting our site in Rails 2.3. We use Nginx in front of Mongrel and also use Sphinx search.
