Hi folks,
Status from our meting on Thursday June 12;
Morten * 3.4.0 debian package received from Werner. Now out on the tool boxes around the country. Discovered a bug in ip dev info, unfortunately. * Automatic database schema upgrade failed for the Arnold database. Some constraints have been automatically named by PostgreSQL 7.4, but the naming scheme has changed for PostgreSQL 8. We should explicitly name constraints, indexes etc. in the future to avoid naming confusions on upgrades. * Deleting a box from NAV takes a _very_ long time. The database delete operation blocks the web interface, causing the result page to load "forever", resulting in a bad user experience. The delete takes a long time because of all the cascading deletes necessary in PostgreSQL. Possible workarounds are being looked into; * do the delete as a background process. Inform the user that the delete will take effect in an hour (or so). Other improvement: * The key bottleneck (when deleting a switch) is most likely the cam table. It can become very large, thus updating hundreds of thousands of rows due to cascading from netbox deletes will be slow. A tested schema change (details will be reported by Morten) reduced the delete time from 43 minutes to 10 minutes for a box with 750.000 cam rows in a 12M row cam table. * We need a means of overriding the django settings file. I.e. django sends email using an awkward from address and django has timezone set to chicago time, which is a pain in many use cases... * possible solution: /etc/django.conf where sensible parameters are set and in turn imported into the django settings file(s).
John Magne * macwatch background process implemented. It runs just after the camlogger in time and looks for "wanted" mac addresses listed in a config file. If a "wanted" mac is detected an alarm is generated. * macwatch is checked into a separate repository. Needs further testing before being merged to the default branch. * A nav cricket contrib is made for temperature monitoring. The contrib adds a database table for storing temperature sensors. A script fills this table with data based on collected temperature sensors seen on the netboxes. Another script expands the cricket config tree to include the temperature monitoring. The latter script is also embedded in makecricketconfig, it needs adjustment to the .nav file in the cricket config tree - a rather unocumented feature as of now. * This contrib is not on the default branch yet, either.
Magnus: * Started working with alert profiles. Focus on bug fixing. Currently working with resolving error messages seen with php5. Fixes need to support php4 as well. * Is considering a rewrite of alert profiles to python. Many advantages; 1) wanted fixes in the user interface can then be resolved (not worth while fixing the current php code to be more NAV look a like). 2) the same django model as the new alert engine uses can be adopted. 3) There are cases today where alert engine and alert profiles have a different view of the world, unfortunately (read: bug[ish]).
Jørgen: * Not much to report - studying for his exams. Will start next Tuesday.
Tomas * A reported bug from UiT in alert engine is not possible to reproduce, not by UiT either. Bug case closed. * Has studied the alert engine code in debth. Tried for hours to resolve a bug, but changed to a more positive mode and started rewriting the code. * Has the framework for a new alert engine ready with lots of django models implemented. A complete pyhthon alert engine may be finished by the end of next week!!!
Kristian * Good progress with the new network explorer. A lot af javascript. Basic skeleton in place, currently implementing search functions, sysname search in place. * Responded to several bug reports regarding the new network map.
Possible new task to someone [tm] * Find a bug tracker replacement since sourceforge is horribly slow. Alternatives could be launchpad.net or roundup (on a UNINETT install).
- Vidar