== link:index.html[Index] -> link:config.html[Configuration]
////
Last checked: 2010/04/26 Cherokee 0.99.46b
////

Information Sources
-------------------

Every information source to be used by the web server is configured
only once in this section. After that, any virtual server will be able
to use as many as desired without having to set up the same thing over
and over again for every one of the configured servers.

This section shows the configured information sources (_Known
sources_), the usage of sources by every defined rule in your virtual
servers (_Source usage_) and allows the addition and edition of
sources.

There are two possible settings as *information sources*:

. *Remote*: Cherokee will simply dispatch the request to the specified
  host. As many hosts as needed can be added, and Cherokee will simply
  apply the balancing strategy to which the information source is
  assigned among them. If for any reason the requests made to these
  hosts could not be performed, the processes would simply fail.

. *Local*: As before, many hosts can be specified (despite the name,
  the selection is not limited to the localhost). This works the same
  way as the previous option. But in this case, if a request cannot be
  responded then, instead of failing directly, the command specified
  in the field `Interpreter` would be launched before reconnecting to
  try again. The most common use of this would be launching the
  service that should handle the request, but it is a flexible
  mechanism that in theory could allow you to reroute the information
  to allow you to serve SSH sessions via your webserver or any other
  neat tricks you can think of. If this field is specified then the
  parameter `Spawning timeout` can also be specified. If this limit is
  exceeded and the interpreter has not yet been launched, Cherokee
  will give up. The default value is 3 seconds.

Note that because of Cherokee's flexibility and modularity, sometimes
and depending of the situation one or the other setting might not
make much sense. In such occasions `cherokee-admin` will simply hide
whatever options do not apply.

.PHP information source
image::media/images/admin_info_sources.png[Information Sources]

[[daemonization]]
[WARNING]
Do not provide daemonizing options as part of the command to launch a
given interpreter.
Cherokee's auto-spawning mechanism is very useful, since it can
re-launch a local interpreter if for whatever reason the process had
died unexpectedly. The launched interpreter is automatically
daemonized, and `stdout` and `stderr` are conveniently handled in
accordance to any logging setting you might have configured. Beware:
some applications accept a parameter to daemonize themselves, but
using such parameter will most likely interfere with Cherokee, so you
are advised against that. There is no need for such parameters,
stderr/stdout redirections, background-launching, etc.


[TIP]
The information about source usage allows you to see which information
sources are currently in use in whatever rule configured along your
list of virtual servers. This is intended as a tool to let you know
whether you can or cannot safely remove an information source from
your list. When you do so, the source is also removed from the rule.
For instance, if your PHP rule was balancing among two information
sources, by deleting one of the sources in this section you would also
be deleting it in the corresponding rule entry. Now the PHP rule would
balance only with one information source. This is perfectly safe. The
problem arises whenever you want to completely exhaust the list of
sources being balanced. Since every balancer (a generic balancer or a
FastCGI balancer, for example) needs at least one information source,
the rule would be broken and Cherokee would fail to start. This is
prevented by disabling the deletion of the last source of any rule. To
eliminate such sources you will have to manually fix the rule by
either setting up alternative information sources or by completely
eliminating the rule.

[[unix_sockets]]
Unix Sockets
~~~~~~~~~~~~

Note that information sources can be addressed by host:port or directly
through Unix sockets. You only have to specify a full path to the
socket instead of the `host:port`.

For instance an entry such as `/var/tmp/socket-tmp.1` should get
everything to simply work if that was a correct path.


[[further_insight]]
Further insight
~~~~~~~~~~~~~~~

This is the list of currently supported balancing strategies:

* link:modules_balancers_round_robin.html[Round Robin]
* link:modules_balancers_ip_hash.html[IP Hash]

And these are the handlers that use balancing:

* link:modules_handlers_proxy.html[Reverse HTTP proxy]
* link:modules_handlers_fcgi.html[FastCGI]
* link:modules_handlers_scgi.html[SCGI]
* link:modules_handlers_uwsgi.html[uWSGI]
* link:modules_handlers_dbslayer.html[MySQL bridge]