ZopeHealthWatcher allows you to monitor the threads of a Zope application, wether it's a Zeo client, wether it's a plain Zope server.
For each thread running on your server, you will know if it's active or idling. When it's active, you will get an execution stack.
It's also useful to debug in case of a locked thread : you'll know where the problem is located.
You can monitor it through your browser or through a console script.
In addition, your front-end servers can efficiently check the status (health checking) of a Zope server without blocking the Zope application threads.
ZopeHealthWatcher is based on DeadlockDebugger code, see http://plone.org/products/deadlockdebugger.
If you run zc.buildout, add the Products.ZopeHealthWatcher product into
your buildout file.
For example
[buildout]
parts =
zhw
[zhw]
recipe = zc.recipe.egg
eggs = Products.ZopeHealthWatcher
scripts = zHealthWatcher
You can also install it using pip or easy_install.
Once the package is installed, create the config file zopehealthwatcher.ini module located in
directory of the zope instance and add following lines, so
the tool is activated:
[ZopeHealthWatcher] SECRET=secret
There are three ways to query the tool: with the zHealthWatcher script, through the browser or http based health check clients.
zHealthWatcher takes the root URL of the Zope server to run:
$ zHealthWatcher http://localhost:8080 OK - 0/4 thread(s) are working (ready)
It will return the number of idling and busy threads.
In case your server is on high load (e.g. 4 busy threads), the tool will return some relevant infos like the time, the sysload (only linux), the memory information (only linux) and for each busy thread, the current stack of execution, the query, the url and the user agent:
$ zHealthWatcher http://localhost:8080
Information:
Time 2009-05-18T18:23:34.415319
Sysload
Meminfo
Dump:
Thread -1339518976
QUERY: GET /test?
URL: http://localhost:8080/test
HTTP_USER_AGENT: Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; en-US; rv:1.9.0.10) Gecko/2009042315 Firefox/3.0.10
File "/Volumes/MacDev/bitbucket.org/zopewatcher/parts/zope2/lib/python/ZServer/PubCore/ZServerPublisher.py", line 25, in __init__
response=b)
...
roles = getRoles(container, name, value, _noroles)
Thread -1340051456
QUERY: GET /test?
URL: http://localhost:8080/test
HTTP_USER_AGENT: Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; en-US; rv:1.9.0.10) Gecko/2009042315 Firefox/3.0.10
File "/Volumes/MacDev/bitbucket.org/zopewatcher/parts/zope2/lib/python/ZServer/PubCore/ZServerPublisher.py", line 25, in __init__
response=b)
...
roles = getRoles(container, name, value, _noroles)
Thread -1341648896
not busy
Thread -1341116416
QUERY: GET /test?
URL: http://localhost:8080/test
HTTP_USER_AGENT: Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; en-US; rv:1.9.0.10) Gecko/2009042315 Firefox/3.0.10
File "/Volumes/MacDev/bitbucket.org/zopewatcher/parts/zope2/lib/python/ZServer/PubCore/ZServerPublisher.py", line 25, in __init__
response=b)
...
roles = getRoles(container, name, value, _noroles)
Thread -1340583936
QUERY: GET /test?
URL: http://localhost:8080/test
HTTP_USER_AGENT: Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; en-US; rv:1.9.0.10) Gecko/2009042315 Firefox/3.0.10
File "/Volumes/MacDev/bitbucket.org/zopewatcher/parts/zope2/lib/python/ZServer/PubCore/ZServerPublisher.py", line 25, in __init__
response=b)
...
roles = getRoles(container, name, value, _noroles)
WARNING - 4/5 thread(s) are working (high load)
If the server is down or unreachable, the script will return a failure:
$ zHealthWatcher http://localhost:8080 FAILURE - [Errno socket error] (61, 'Connection refused')
zHealthWatcher is also returning the right exit codes, so it can be used by third party programs like Nagios:
- OK = 0
- WARNING = 1
- FAILURE = 2
- CRITICAL = 3
An HTML version is accessible through the web, using the url http://host:port/manage_zhw?secret. This url has to be changed depending on the values entered in zopehealthwatcher.ini.
Beware that this URL is not password protected.
In general, before performing traffic by a front-end server (e.g. load balancer) forwarding to back-end Zope servers, it is recommended to tell the front-end server to check the health of the Zope threads hosted by each Zope server of a farm. The not password protected url http://host:port/manage_zhw can simply used without blocking a Zope thread for health checks by common known web front-end servers like nginx, varnish or haproxy. The availability of a Zope thread is encoded in the http response of the url:
If the server has idle Zope threads, it will return $ curl -I http://localhost:8080/manage_zhw HTTP/1.0 200 OK ... If the server is busy (all Zope threads are busy), it will return $ curl -I http://localhost:8080/manage_zhw HTTP/1.0 404 Not Found ... If the server has zombie Zope threads or rather is in an undefined state, it will return $ curl -I http://localhost:8080/manage_zhw HTTP/1.0 500 Internal Server Error ...