My Note Cloud Service is a part of My Notes solution to access your IBM Lotus Notes applications on iPhone or Android developed by Inexika Inc. The service is used by My Notes Desktop and My Notes mobile apps connected over the internet.

• Cloud of servers in different parts of the world. The closest server is chosen for connection to ensure the maximum speed.
• Ability to run more then one instance on each server
• Non-blocking Tornado based HTTP Server listening of particular port as instances
• Silent switching between instances and servers
• Ability to run additional servers and instances on the fly

• Nginx as a front reverse proxy with Non-buffered upload, round robin forwarding of connection requests to instances
• RRDTool-based graphing and monitoring (optional)

• Install nginx

• Standard Nginx doesn't support Non-buffered upload. You can either install Tengine instead (web server based on the Nginx originated by Taobao which does) or just apply patch to your Nginx version. That's the way we followed because proxy_request_buffering was the only option we needed in addition. Use commands like below to apply the patch.

    patch -p 1 -i nginx-1.4.2-no_buffer-v8.patch
    ./configure --with-http_ssl_module --with-http_gzip_static_module
    make
    make install
  

• Install Python 2.7

• Install RRDTool - optional RRA-stats and monitoring

• Clone repository git clone git://github.com/Inexika/MyNotes-Service.git mynotes && cd mynotes

• Install required python packages (you can do this just using pip install -r requirements.txt):

  • Tornado
  • PyRRD - Python bindings for RRDTool

• Fix PyRRD bug

• Make sample environment

• Configure server

• Configure instances

• Configure nginx site

• Start nginx service start nginx

• Start instanses

• Start monitor (optional)

• My Notes Desktop - Desktop application running on client computer with IBM Lotus Notes installed
• Mobile app - Mobile app for iPhone or Android
• Service - Cloud of several servers, Desktop is connected to the closest one.
• Server - Server where more than one instance usually running
• Instance - Process on a server, listening to a particular port
• Master Instance - Special instance in the cloud, to control distribution of CustomerID ranges between other instances.
• Monitor - Script running periodically to update charts and alert about instances being overloaded or down.

Make a sample configuration:

# rename sample files and directories
mv mynotes.sample.conf mynotes.conf
mv instance.sample instance
mv range.sample range
# create directories for range and log files
mkdir range/8082
mkdir log log/8081 log/8082

The sample configuration is:

• Nginx as a reverse proxy server
• The only server mynotes.your_domain.com is in the cloud.
• There are 2 instances 8081, 8082 running by default
• 8081 is considered to be a master
• RRDTool-based graphing and monitoring is enabled

mynotes.conf file contains common parameters for all instances of current server, specify necessary options.

Timeouts and buffer-size for data transferring and logging settings:

# --Timeout settings
timeout_agent = 15
timeout_cache = 10
timeout_client = 10
timeout_no_reply = 30

# --buffer size: use upper bound Content-Length as a key
buffer_size = {102400:4096, 512000:8192, 1048576:16384, float('inf'):32768}

# --max clients option for AsyncHTTPClient
http_max_clients = 60

# failed connections before removing instance/server
max_instance_failed = 5

# --logging settings
logging= 'INFO'
stats_enabled = True

RRDTools settings. Defines parameters for Round-Robin-Archives to be created:

# --RRD settings
rrd_enabled = True

#set True if rrd file should be recreated
rrd_reset = False

# stats receiving interval
# rra settings depend on this
# 1 minute
stats_period = 60   

# Round Robin Archives (RRA) settings
rrd_rra = [
(1,720),    #60->720 samples of 1 minute |(hour)--> (12 hours)
(5,1440),   #1440 samples of 5 minutes (day)
(60,168),   #168 samples of 1 hour (week)
(1440,60),  #60 samples of 1 day (2 months)
(43200,12), #12 samples of 1 month (year)
]

Current server name, default instances and Master:

# current server name
server = 'mynotes.your_domain.com'

# all your default cloud instances are listed here 
sites = [
('mynotes.your_domain.com','8081'),
('mynotes.your_domain.com','8082'),
]

# master instance of your cloud
master = ('mynotes.your_domain.com','8081')

Charts and alert parameters. The section is used to config monitor:

# --monitor settings
# path to graph images
monitor_graph_path = '../imgs'

# --chart settings - graphical representations to be created
# graphics = {size: ([graph_periods], [graph_types])}
# size: 'S','M','L' are acceptable values
# graph_periods: RRDTool time offset specifications: '1h', '1day', '1m', etc.
# graph_types: hardcoded internal graph names. None means all of existing types.
graphics = {
'S': [ (['6h'], ['cpu-max', 'bytes-max', 'agents_u-max', 't_completed-max', 't_unique-max', 'duration-max']),],
'M': [
(['1h','3h','6h','12h','1day','3day','7day','1m'],['cpu','bytes','agents_u','t_completed','t_unique','duration']),
(['6h'], ['cpu-max', 'bytes-max', 'agents_u-max', 't_completed-max', 't_unique-max', 'duration-max']),
]
}

# --alert settings
alert_over = 900000     #900000 mcs/sec = 90%
alert_threshold = 0.3
alert_grace_period = 1800
alert_from = 'MyNotes cloud'
alert_sender = 'mynotes@your_domain.com'
alert_receivers = ['mynotes_admin@your_domain.com']
alert_log = 'log/monitor.sent'
alert_smtp = 'mail.your_domain.com'

Sample config files for all instances are located in ./instance directory. Use listened port as a file name. Instance config file contains parameters, related to particular instance. Such as log file, stats file, etc.

There are 2 sample files: 8081.conf, 8082.conf in the directory. Rename them according to your ports if needed. Create another files for all of your instances or remove extra ones if needed.

• range_file, master_range:

  The first file contains instance CustomerID range for registration . No need to create it manually. The second one is for master-instance only. It contains cloud CustomerID range distributed among instances.

  ./range.sample contains 8081/master.ini file as a template empty directory 8082. You can either rename it to range use that structure or relocate range files where you wish.

• log_file_prefix, stats_file_prefix, rrd_file:

  ./log.sample contains directories. You can either rename it to log or locate the files where you wish.

• Check the correct locations are specified in the instances' config files .

8081.conf

# log file name
log_file_prefix = 'log/8081/mynotes.log'

# stats file name
stats_file_prefix = 'log/8081/mn_stats.log'

# range ID file
range_file = 'range/8081/range.ini'

# master range ID (for master instance only!!)
master_range = 'range/8081/master.ini'

# optional: if Round-Robin-Achive is used for stats and monitoring
rrd_file = 'log/8081/stats.rrd'

To configure nginx site use mynotes.your_domain.com file in sites-available. See example.

upstream config:

# round-robin upstream
upstream mynotes  {
    server 127.0.0.1:8081 weight=1;
    server 127.0.0.1:8082 weight=1;
}

server context:

server {
    server_name mynotes.your_domain.com;
    listen 80;
    ....
    client_max_body_size 32m;
    proxy_buffering off;
    proxy_read_timeout 300s;

    # Tengine feature - http://tengine.taobao.org/document/http_core.html
    proxy_request_buffering off;    

location to pass to upstream:

# All other requests except those:
# mache "port number" regex
# are transfered (proxy) to MyNotes upstream (Tornado)
location / {
    proxy_set_header        Host            $host;
    proxy_set_header        X-Real-IP       $remote_addr;
    proxy_set_header        X-Forwarded-For $proxy_add_x_forwarded_for;
    proxy_pass http://mynotes;
}

locations to pass to particular instances:

# Requests wich mache "port number" regex are transfered to corresponding instance of My Notes service (tornado)
location ~ /8081$ {
    proxy_set_header        Host            $host;
    proxy_set_header        X-Real-IP       $remote_addr;
    proxy_set_header        X-Forwarded-For $proxy_add_x_forwarded_for;
    proxy_pass http://127.0.0.1:8081;
}
location ~ /8082$ {
    proxy_set_header        Host            $host;
    proxy_set_header        X-Real-IP       $remote_addr;
    proxy_set_header        X-Forwarded-For $proxy_add_x_forwarded_for;
    proxy_pass http://127.0.0.1:8082;
}
.....

or use pyrrd.patch from sources

pyrrd.backend.external.prepareObject()

if function == 'fetch':
    validParams = ['resolution', 'start', 'end']
    params = common.buildParameters(obj, validParams)
    + return (obj.filename, [obj.cf] + params)
    - return (obj.filename, obj.cf, params)

Each of the instances can be run independently from the others. To run instances and listen particular ports:

python mn_service.py --port=8081
python mn_service.py --port=8082

We recommend to configure periodic start monitor script. Every time monitor runs it updates files with stats for charts. It also checks whether overloading or downtime is happening, and sends email alert if needed.

python mn_monitor.py

Charts can be shown somewhere, e.g. special site or page can be configured. See one of our server stats as example