OMERO.web Nginx and Gunicorn deployment (Unix/Linux)
====================================================
.. note:: Since OMERO 5.2, the OMERO.web framework no longer bundles
a copy of the Django package, instead manual installation of
the Django dependency is required. It is highly recommended to use
`Django 1.8`_ (LTS) which requires Python 2.7. For more information
see :ref:`python-requirements` on the
:doc:`/sysadmins/version-requirements` page.
OMERO.web uses the
`Web Server Gateway Interface (WSGI) <http://wsgi.readthedocs.org/en/latest/learn.html>`_
as a primary deployment platform. The Python standard
`PEP 3333 <https://www.python.org/dev/peps/pep-3333/>`_
that describes how a web server communicates with web applications.
.. _gunicorn_default_configuration:
Gunicorn default configuration (Unix/Linux)
-------------------------------------------
Install `Django 1.8`_ and `Gunicorn >= 19.3 <http://gunicorn.org>`_
using the package requirements file:
::
$ pip install -r share/web/requirements-py27-nginx.txt
.. note:: For more details refer to
:djangodoc:`how to install Django 1.8 <topics/install/#install-the-django-code>`
or :djangodoc:`upgrade Django to 1.8 <topics/install/#remove-any-old-versions-of-django>`.
Additional settings can be configured by changing the following properties:
- :property:`omero.web.application_server.max_requests` to 500
- :property:`omero.web.wsgi_workers` to (2 x NUM_CORES) + 1
.. note::
**Do not** scale the number of workers to the number of clients
you expect to have. OMERO.web should only need 4-12 worker
processes to handle many requests per second.
- :property:`omero.web.wsgi_args` Additional arguments. For more details
check `Gunicorn Documentation <http://docs.gunicorn.org/en/stable/settings.html>`_.
.. _nginx_gunicorn_wsgi_configuration:
Nginx configuration (Unix/Linux)
--------------------------------
If you have installed Nginx, OMERO can automatically generate a
configuration file for your web server. The location of the file
will depend on your system, please refer to your web server's manual.
See :ref:`customizing_your_omero_web_installation_unix`
for additional customization options.
To create a site configuration file for inclusion in a system-wide nginx
configuration redirect the output of the following command into a file:
::
$ bin/omero web config nginx
.. literalinclude:: nginx-omero.conf
.. note::
OMERO.web requires ``body_in_file_only`` adjusted in your default nginx
config because nginx must buffer incoming data. Make sure you have that
set to the following config:
::
http {
...
sendfile on;
send_timeout 60s;
client_max_body_size 0;
...
}
To configure an HTTPS server follow
`the nginx documentation <http://nginx.org/en/docs/http/configuring_https_servers.html>`_.
Running OMERO.web (Unix/Linux)
------------------------------
Start the Gunicorn worker processes listening by default on 127.0.0.1:4080:
::
$ bin/omero web start
... static files copied to '/home/omero/OMERO.server/lib/python/omeroweb/static'.
Starting OMERO.web... [OK]
The Gunicorn workers are managed **separately** from other OMERO.server
processes. You can check their status or stop them using the
following commands:
::
$ bin/omero web status
OMERO.web status... [RUNNING] (PID 59217)
$ bin/omero web stop
Stopping OMERO.web... [OK]
Django WSGI workers (PID 59217) killed.
.. _download_limitation:
Download limitations
^^^^^^^^^^^^^^^^^^^^
In order to offer users the ability to download data from OMERO.web you have
to deploy using :ref:`gunicorn_async_configuration`.
OMERO.web is able to handle multiple clients on a single worker
thread switching context as necessary while streaming binary data from
OMERO.server. Depending on the traffic and scale of the repository you should
configure connections and speed limits on your server to avoid blocking
resources. We recommend you run benchmark and performance tests.
It is also possible to apply :ref:`download_restrictions` and
offer alternative access to binary data.
.. note::
Handling streaming request/responses requires proxy buffering
to be turned off. For more details refer to
`Gunicorn deployment <http://docs.gunicorn.org/en/stable/deploy.html>`_
and
`Nginx configuration <http://nginx.org/en/docs/http/ngx_http_proxy_module.html#proxy_buffering>`_.
.. note::
:property:`omero.web.application_server.max_requests` should be set to 0
.. _nginx_wsgi_benchmark:
Benchmark
---------
We run example benchmarks on rendering thumbnails and 512x512 pixels planes
for 100 concurrent users making 5000 requests in total::
$ ab -n 5000 -c 100 https://server.openmicroscopy.org/omero/webclient/render_thumbnail/size/96/1234/
This is ApacheBench, Version 2.3 <$Revision: 1430300 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Licensed to The Apache Software Foundation, http://www.apache.org/
Server Software: nginx/1.9.9
Server Hostname: server.openmicroscopy.org
Server Port: 80
Document Path: /omero/webclient/render_thumbnail/size/96/31851/
Document Length: 1880 bytes
Concurrency Level: 100
Time taken for tests: 224.488 seconds
Complete requests: 5000
Failed requests: 0
Write errors: 0
Total transferred: 10450000 bytes
HTML transferred: 9400000 bytes
Requests per second: 22.27 [#/sec] (mean)
Time per request: 4489.763 [ms] (mean)
Time per request: 44.898 [ms] (mean, across all concurrent requests)
Transfer rate: 45.46 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 0 0 0.3 0 3
Processing: 435 4446 685.2 4363 7644
Waiting: 432 4446 685.3 4362 7644
Total: 435 4446 685.1 4363 7644
Percentage of the requests served within a certain time (ms)
50% 4363
66% 4553
75% 4670
80% 4750
90% 5072
95% 5398
98% 6795
99% 6955
100% 7644 (longest request)
::
$ ab -n 5000 -c 100 http://server.openmicroscopy.org/omero/webclient/render_image/1234/20/0/
This is ApacheBench, Version 2.3 <$Revision: 1430300 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Licensed to The Apache Software Foundation, http://www.apache.org/
Server Software: nginx/1.9.9
Server Hostname: server.openmicroscopy.org
Server Port: 80
Document Path: /omero/webclient/render_image/1234/20/0/
Document Length: 24293 bytes
Concurrency Level: 100
Time taken for tests: 247.154 seconds
Complete requests: 5000
Failed requests: 0
Write errors: 0
Total transferred: 122515000 bytes
HTML transferred: 121465000 bytes
Requests per second: 20.23 [#/sec] (mean)
Time per request: 4943.080 [ms] (mean)
Time per request: 49.431 [ms] (mean, across all concurrent requests)
Transfer rate: 484.09 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 0 0 0.4 0 4
Processing: 482 4898 855.3 4737 8303
Waiting: 476 4898 855.3 4737 8303
Total: 482 4898 855.2 4737 8303
Percentage of the requests served within a certain time (ms)
50% 4737
66% 5041
75% 5250
80% 5397
90% 5862
95% 6621
98% 7301
99% 8062
100% 8303 (longest request)
.. _nginx_troubleshooting:
Troubleshooting
---------------
In order to identify why OMERO.web is not available run:
::
$ bin/omero web status
Then consult nginx :file:`error.log` and :file:`OMERO.server/var/log/OMEROweb.log`
For more details check :ref:`troubleshooting-omeroweb`.
Debugging
^^^^^^^^^
To run the WSGI server in debug mode, enable
`Gunicorn logging <http://docs.gunicorn.org/en/stable/settings.html#logging>`_
using :property:`omero.web.wsgi_args`:
::
$ bin/omero config set omero.web.wsgi_args -- "--log-level=DEBUG --error-logfile=/home/omero/OMERO.server/var/log/error.log".
.. _gunicorn_advance_configuration:
EXPERIMENTAL: Gunicorn advance configuration (Unix/Linux)
---------------------------------------------------------
.. note:: Experimental configurations are not ready for production use.
Configuration may change. Features may not work properly.
OMERO.web deployment can be configured with sync and async workers.
Sync workers are faster and recommended for a data repository with
:ref:`download_restrictions`. If you wish to offer users the ability
to download data then you have to use async workers; read more about
:ref:`download_limitation`.
For more details refer to
`Gunicorn design <http://docs.gunicorn.org/en/stable/design.html>`_.
.. _gunicorn_sync_configuration:
EXPERIMENTAL: Sync workers
^^^^^^^^^^^^^^^^^^^^^^^^^^
.. note:: Experimental configurations are not ready for production use.
Configuration may change. Features may not work properly.
- Install `futures <https://pypi.python.org/pypi/futures>`_ ::
$ pip install futures
Additional settings can be configured by changing the following properties:
- The number of worker threads for handling requests, see
`Gunicorn threads <http://docs.gunicorn.org/en/stable/settings.html#threads>`_ ::
$ bin/omero config set omero.web.wsgi_worker_class
$ bin/omero config set omero.web.wsgi_threads $(2-4 x NUM_CORES)
.. _gunicorn_async_configuration:
EXPERIMENTAL: Async workers
^^^^^^^^^^^^^^^^^^^^^^^^^^^
.. note:: Experimental configurations are not ready for production use.
Configuration may change. Features may not work properly.
- Install `Gevent >= 0.13 <http://www.gevent.org/>`_ ::
$ pip install "gevent>=0.13"
Additional settings can be configured by changing the following properties:
- The maximum number of simultaneous clients, see
`Gunicorn worker-connections <http://docs.gunicorn.org/en/stable/settings.html#worker-connections>`_ ::
$ bin/omero config set omero.web.wsgi_worker_class gevent
$ bin/omero config set omero.web.wsgi_worker_connections 1000
$ bin/omero config set omero.web.application_server.max_requests 0