Copyright (c) 2002-2005 Infrae. All rights reserved.
See also LICENSE.txt

Meta::
  
  Valid for:  Silva 1.4.x
  Author:     Martijn Faassen
  Email:      faassen@infrae.com
  CVS:        $Revision: 1.4 $

=======================
Silva maintenance guide
=======================

This document contains information about maintaining a Silva instance. Its
target are system administrators. For information about installing Silva,
see INSTALL.txt. For information about non-maintenance issues, see
TROUBLESHOOTING.txt.

Solving version bloat
---------------------

Each time a user creates a new version of any VersionedContent object (e.g.
Silva Document) a new Version object is created inside that object. This can
potentially result in a large number of Version objects getting stored in
the ZODB, and since rather than diffs the whole contents of a Version is
copied when a new Version is created, this might add up to a firm amount of
disk space used.

Since version 1.2 Silva provides functionality to remove Version objects
from VersionedContent, but the interface is not very useful if you have a
large setup and want to get rid of all unused Versions, so there is some
additional functionality available, in the form of an adapter (so part of
the API, not directly callable) to clean up all older versions at once. To
trigger this functionality, you will have to write a Python script (on the
Silva root, from the ZMI) with the following contents:

  from Products.Silva.adapters.cleanup import getCleanupVersionsAdapter
  root = context.get_root()
  adapter = getCleanupVersionsAdapter(root)
  adapter.cleanup()

  return '\n'.join(['%s: %s' % items for items in result.items()])

When called (note that it requires Manager permissions) it will traverse the
full Silva tree to find versions other than the current editable, approved,
published or last closed ones of each VersionedContent object and remove them.
Note that this can take a LOT of time to execute.

Triggering publication and expiration
-------------------------------------

The Silva application is HTTP based, and doesn't have a scheduling system to
trigger maintenance scripts out-of-the-box, which makes that scheduling
tasks is impossible. HTTP applications generally only do anything if there's
an HTTP request, so maintenance tasks are usually ran on HTTP requests too.
For certain tasks, this can lead to problems, for instance consider the 
following issue with Silva's workflow:

When a document is approved and the publication date has passed, the document
is not yet published, since this requires an HTTP request to the document to 
get triggered. This means that the status of certain documents may not always
be up-to-date, which for certain bits of functionality (e.g. searching) might
have incorrect results.

To solve this problem, you can call an external URL from a cron (or similar)
application:

  http://localhost/silva/status_update

The script behind this URL requires Manager permissions, and will check
both the publication date and expiration date for each relevant Silva object,
fixing the publication status when appropriate.
