collective.warmup has been created to warm up web application's caches upon start and restart.
It works by reading a configuration file containing a list of urls that are then invoked.
collective.warmup is inspired by Gil Forcada's warmup_plone.py script.
While the basic use case is very simple, collective.warmup offers the following extra features:
- Verifies the correctness of the response body (e.g. contains a certain string)
- Can operate in crawl mode, following links in pages to warm up all related pages
- Can filter out certain urls according to a general definition
collective.warmup can be installed in two ways:
To install it (using virtualenv, on a GNU/Debian machine):
$ sudo apt-get install build-essential python-dev python-lxml python-virtualenv libxml2-dev libxslt1-dev $ virtualenv warmup && cd warmup && souce bin/activate (warmup) $ pip install collective.warmup
and launch it with:
(warmup) $ bin/warmup <path/file.ini>
To integrate collective.warmup in a buildout with Plone, add this to your configuration:
[buildout] ... parts = ... warmup [instance] ... eggs += collective.warmup [instance] environment-vars += WARMUP_BIN ${buildout:directory}/bin/warmup WARMUP_INI ${buildout:directory}/warmup.ini WARMUP_HEALTH_THRESHOLD 50000 [warmup] recipe = zc.recipe.egg:scripts eggs = collective.warmup
After executing the buildout you will find the warmup script
in the bin
directory.
In this example, the script will be executed automatically by the Zope instance each time it is started.
This is a sample warmup.ini
configuration:
[warmup] enabled = True sleep = 2 base_url = http://localhost logfile = /path/to/warmup.log urls = Home page [config] max_attempts = 2 base_path = mysite [Home page] path = ${config:base_path}/ max_attempts = ${config:max_attempts} check_exists = Welcome check_not_exists = p0wned by follow_links = True ignore_middle = @@ ++theme++ # ? ignore_end = .css .js .png .jpg .jpeg .gif .xml RSS .ico
- enable :
True
orFalse
- If
False
, the script will do nothing when invoked. - sleep : integer
- The number of seconds the script waits between url retrievals.
Defaults to
2
. - base_url : a valid URL (don't forget the protocol!)
- The base URL to check (all paths in the various URL sections are relative to this URL).
- log_file : a filesystem path
- The file where the logs will be written.
- urls : a list of sections (separated by newline)
- The URLs that we want to check. Each URL must have its own section in the configuration file (see below) and we reference these sections here (do not put real URLs here!). It also set an order for the checks (which are executed sequentially).
- max_attempts : integer
- The maximum number of attempts to check the url.
Defaults to
2
- path : the path to check
- The path will be added to the
base_url
parameter in order to retrieve the page url - check_exists : list of strings
- A list of strings that must be present in the page
- check_not_exists : list of strings
- A list of strings that must not be present in the page
- follow_links :
True
orFalse
- If
True
the script will follow the links in the page and will perform the same checks for each link. - ignore_middle : list of strings
- If
follow_links
isTrue
, the links containing one of these strings will be ignored - ignore_end : list of strings
- If
follow_links
isTrue
, the links ending with one of these strings will be ignored
In order not to mark backend healthy too early by the load-balancer, before proper
warmup, this package defines a browser view called @@health.check
which can be
used within your load-balancer probe mechanism. See bellow a Varnish configuration
example:
backend instance_1 { .host = "localhost"; .port = "8081"; .probe = { .url = "/health.check"; .interval = 5s; .timeout = 1s; .window = 5; .threshold = 3; } }
This way Varnish will mark the Zope instance backend healthy when
ZODB cache-size is bigger than WARMUP_HEALTH_THRESHOLD
. If you do not define
the WARMUP_HEALTH_THRESHOLD
environment variable, the Zope instance backend
will be marked healthy as soon as Zope will be able to handle requests.