CHAPTER ONE LOAD-BALANCING

1.1 Synopsis

In this part, we will explain how to create a load-balancer withnginxfor a lot of OpenERPservers.The

load-balancing

and the

multi-process

mode of OpenERPonly work with gunicorn.And in this case, the cron are not launched by the system.

1.2 Nginx

Nginx is a free, open-source, high-performance HTTP server and reverse proxyYou can copy/paste the code in

/etc/nginx/sites-enabled/

with a filenameYou will find the documentation about the upstream keyword at

http://wiki.nginx.org/HttpUpstreamModule

 

upstream
openerp_servers{
server
openerp-server1.back.local max_fails=3 fail_timeout=30s;
server
openerp-server2.back.local max_fails=3 fail_timeout=30s;
server
openerp-server3.back.local max_fails=3 fail_timeout=30s;}
serfver
{
listen
80;
server_name
openerp.example.com;
rewrite
^ https://$server_name$request_uri?permanent;}
server
{
listen
443;
server_name
openerp.example.com;
access_log
/var/log/nginx/openerp.example.com-access.log;
error_log
/var/log/nginx/openerp.example.com-error.log debug;
ssl
on;
ssl_certificate
/etc/nginx/ssl/server.crt;
ssl_certificate_key
/etc/nginx/ssl/server.key;
ssl_session_timeout
5m;
ssl_protocols
SSLv3 TLSv1;
ssl_ciphers
ALL:!ADH:!EXPORT56:RC4+RSA:+HIGH:+MEDIUM:+LOW:+SSLv3:+EXP;
ssl_prefer_server_ciphers
on;
location
/{
 proxy_pass
http://openerp_servers;
 proxy_set_header
Host$http_host;
 proxy_set_header
X-Real-IP$remote_addr;
 proxy_set_header
X-Forwarded-For$proxy_add_x_forwarded_for;}}

1.3 Gunicorn

Gunicornis a Python WSGI HTTP Server for UNIX.

We will use it for the multi-process mode of OpenERP.On the OpenERP servers, just install gunicorn via apt-get

# apt-get install python-gunicorn

1.4 OpenERP

You should configure OpenERP with the gunicorn mode.In the directory of the server, you can execute gunicorn with the following command

cd/home/openerp/servergunicorn openerp:wsgi.proxied.application -c gunicorn.conf.py

2.1 PostgreSQL

2.1.1 Synopsis

PostgreSQLis a free and open source database server.In this documentation, we will use PostgreSQL 9.x because this is the first release with theStreaming Replication.In PostgreSQL, there are two kind of replication.•

Warm Standby

It’s a simple replication without any access.•

Hot Standby

, in this mode, you can connect to the database but only in read access.We will use the

Hot Standby

mode for the Streaming Replication.In this tutorial, we suppose that you use an unix account, named

openerp

.Firstly, we have to configure two PostgreSQL servers, the master and the slave.

2.1.2 Streaming Replication

Here is the directory structure for the tutorial

root@vm:/# tree homehome|-> openerp|-> postgresql|-> master|-> standbymkdir /home/openerp/postgresql

Master

Via this command we initialize a new PostgreSQL cluster.

cd/home/openerp/postgresql/usr/lib/postgresql/9.1/bin/initdb -D master

The

wal_level

parameter specifies the

hot_standby

mode and the

max_wal_senders

is the number of slaveservers. In this case, the value is one.

cat >> master/postgresql.conf<< _EOF_ > include ’replication.conf’> _EOF_ cat >> master/replication.conf<< _EOF_ > wal_level = ’hot_standby’> max_wal_senders = 1> archive_mode = on> #archive_command = ’cp %p /home/stephane/postgresql/test/archives_xlog/%f’> _EOF_

We start the new PostgreSQL cluster

/usr/lib/postgresql/9.1/bin/pg_ctl -D master start

Authentication

We will use the replication via the network, in this case, we have to configure the PostgreSQL server to use a dedicateduser for the replication. In our case, this is the openerp user. It’s a normal postgresql user, don’t forget to set all therights for this user and a password.

createuser openerppsql -d postgres -c"ALTER USER openerp PASSWORD ’secret’;"

Note:

Here is a part of the content of the

master/pg_hba.conf

file

# The replication database is a virtual database used by the replication processhost replication openerp 127.0.0.1/32 md5

# Don’t forget to restart the master server before the slave because without that, the slave can’t b

/usr/lib/postgresql/9.1/bin/pg_ctl -D master restart

Slave

Tocreateabackupofthemasterfortheslave, youcanusethe

pg_basebackup

commandprovidedwithPostgreSQL9.1.

cd/home/openerp/postgresql/usr/lib/postgresql/9.1/bin/pg_ctl/pg_basebackup -D standby -v -P -Fp -l backup -h 127.0.0.1 -U open

We configure the standby server to listen to the port 5433 (because we are on the same physical server). If the standbyserver is another server, don’t forget to use 5432.

cat > standby/replication.conf<< _EOF_ > port = 5433> hot_standby = on> _EOF_

The

primary_conninfo

is used by the Streaming Replication, the parameter have the hostname of the master withthe port and the pair user/password.

 

cat > standby/recovery.conf<< _EOF_ > standby_mode = ’on’> primary_conninfo = ’host=127.0.0.1 port=5432 user=openerp password=secret’> trigger_file = ’/tmp/standby’> _EOF_

In this config file, the trigger_file key will be used by the server, if there is a touch on this file, the standby server willautomatically pass in the master mode with the read/write accesses.

2.2 PG-Pool II

For the load-balancing and the failover features, we will use PgPool2.

2.2.1 Installation

# apt-get install pgpool2# apt-get install postgresql-9.1-pgpool2

2.2.2 Configuration

Here is the basic configuration for the load-balancing and the failover.The

failover_command

is important because if pgpool can’t make a connection to the master, it will

touch/tmp/standby

file and this file will be triggered by the standby server.If you want to use this feature with a remote standby server, you can use

ssh

to touch a file on a remote server. Butdon’t forget to change the config file of the standby server and set the right value for the

trigger_file

parameter.

cat >> /etc/pgpool2/pgpool.conf << _EOF_ > port = 5434> load_balance_mode = on> backend_hostname0 = ’localhost’> backend_port0=5432> backend_weight0=1> backend_flag0 = ’ALLOW_TO_FAILOVER’> backend_hostname1 = ’localhost’> backend_port1=5433> backend_weight0=1> backend_flag1 = ’ALLOW_TO_FAILOVER’> failover_command = ’touch /tmp/standby’> _EOF_