This file is part of oopstools
Copyright 2009-2011 Canonical Ltd. All rights reserved.
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU Affero General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU Affero General Public License for more details.
You should have received a copy of the GNU Affero General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>.
This is a Django application to analyze OOPS reports
Please see https://dev.launchpad.net/LazrStyleGuide and
https://dev.launchpad.net/Hacking for how to develop in this
OOPStools uses buildout for configuration.
All dependencies (except for the database packages) are available from the
lazr-source-dependencies branch on Launchpad. Simply typing make will set-up
the download-cache of dependencies eggs and then create the buildout.
If you already have that shared eggs cache branch available locally (to share
between branches of the Lazr project), simply set the LAZR_SOURCEDEPS_DIR to
the directory containing the download-cache and eggs directories before
To run under apache with mod_wsgi (recommended)
* libapache2-mod-wsgi package
To run under apache with mod_python
* libapache2-mod-python package
Deployment using mod_wsgi
Create a custom cfg file - start with production.cfg and take a copy. Update
your copy with your database configuration and the paths to your OOPS
db-name = /path/to/your/oops.db
index-template = 'index.html'
oopsdir = /path/where/rsynced/oopses/are/found
Update settings.py setting a custom SECRET_KEY, and appropriate AMQP connection
details (so that the oops-tools web site itself can report errors, should they
To deploy oops tools make sure all the dependecies are installed.
* bin/buildout -c yourfilename.cfg
* Run bin/django syncdb
* Run bin/django migrate
* Copy apache/oops-tools.dev.mod_wsgi to /etc/apache2/sites-available/
The script bin/amqp2disk is an AMQP handler that will receive OOPS reports over
AMQP and publish them locally to disk as well as loading the metadata directly
into the oops-tools database. To use this you will need to config your OOPS
creation to publish over AMQP. If you are using Python then the oops-amqp
module will help you do this.
The Django web UI also integrates with AMQP to report errors that occur
rendering or searching for OOPS reports. The settings to control this are in
settings.py (or in your configuration file which buildout compiles into a
Sometimes Launchpad developers want to run oops-tools to help analyse locally
generated OOPS reports.
The first step is to setup the local database.
Local PostgreSQL setup
If you do Launchpad development, you probably have PostgreSQL already running
on port 5432. You'll need another cluster to run the oops-tools database.
Create a new cluster for lpoops:
$ LC_ALL=C sudo pg_createcluster 9.1 lpoops --encoding UNICODE --port 5433
Apply the following patch to /etc/postgresql/9.1/lpoops/pg_hba.conf. Note that
this configuration change needs to be before the "DO NOT DISABLE!" line in the
sudo patch /etc/postgresql/9.1/lpoops/pg_hba.conf <<'EOF'
--- pg_hba.conf 2005-11-02 17:33:08.000000000 -0800
+++ /tmp/pg_hba.conf 2005-11-03 07:32:46.932400423 -0800
@@ -58,7 +58,9 @@
# on a non-local interface via the listen_addresses configuration parameter,
# or via the -i or -h command line switches.
+# Oops-tools users
+local all all trust
+host all all 127.0.0.1/32 trust
In case localhost is using IPv6, you'll need an extra line just below the one above.
host all all ::1/128 trust
Start the cluster.
$ sudo pg_ctlcluster 9.1 lpoops start
Add your own user to the cluster:
$ sudo -u postgres createuser -a -d $USER --port 5433
To create the new db:
$ createuser lpoops --createdb --no-superuser --no-createrole --port 5433
$ createdb -O lpoops lpoops --port 5433
Note that this command creates a password-less database. Use --password to
create a db with a password.
To delete the db:
$ dropdb lpoops --port 5433
$ dropuser lpoops --port 5433
If you used a different port than 5433 for the PostgreSQL setup, pass the
DBPORT variable to make and it'll build the configuration files using the
* `make DBPORT=$PORT`
* Run bin/django syncdb
* Run bin/django migrate
* `make run` at the root of the oops-tools directory
* Point your browser to http://localhost:8000/oops
NEWS for oopstools
* Permit multiple project/project groups to be supplied to prune.
(Robert Collins, #1003627)
The brown paper bag edition.
* Fix typos that prevented the new timeline-django code installing or working
properly. (Robert Collins)
* Dependency bump to non-broken version of timeline-django. (Robert Collins)
* Oops-tools now depends on newer djangorecipe, django 1.3.3 and
oops-datedir-repo 0.0.19. The -repo dependency bump gets the analogous fix
for 1050722 for the on-disk purging. (Robert Collins).
* Oops reports generated by oops-tools itself will now include django event
data from timeline_django. This can be disabled if needed by disabling the
oops_timeline hook in oopstools.wsgi. (Robert Collins)
* When pruning, look for OOPS references up to now rather than up to
prune_until. Otherwise we miss any references created in the last
week, probably only finding references within 24 hours of the OOPS
occurring. (William Grant, #1050722)
* Reports have a per-report email address that can be set, to allow different
reports to go to different addresses. (Robert Collins)
* HTML reports can be created for non-ascii OOPS topics.
(Robert Collins, #1048470)
* Corrupt timeline objects are handled more gracefully if they are not a list,
or the list rows are too short or too long. (Robert Collins, #890001)
* Failures in processing an AMQP sourced OOPS will now write out the path to
the disk copy of the failed OOPS, to ease debugging.
(Robert Collins, #892914)
* Failures in the Django web UI will now raise an OOPS over AMQP, making
oops-tools an oops-tools consumer. (Robert Collins)
* Flush stdout when logging an OOPS receipt in amqp2disk.
(Robert Collins, #884569)
* Hostnames are now shown in the OOPS detail page. This makes it possible to
figure out if one host in particular is having trouble. (Robert Collins)
* Long oops ids - which are likely hashes - no longer have the old appserver id
heuristic applied to them: if they have a missing or empty reporter, they are
associated with the instance 'unknown'. (Robert Collins, #889982)
* Mixed case OOPS reports can now be looked up in the web UI without their
OOPS-prefix (if they had one). (Robert Collins, #884571)
* Now uses the lp source dependencies rather than lazr, as part of simplifying
the maintenance of the many related LP projects. (Robert Collins)
* Old OOPS reports can be cleaned out using the new script bin/prune.
* OOPS reports that don't meet the normal rules for req_vars are handled
a bit better.
(Robert Collins, William Grant, Roman Yepishev, #885416, #884265)
* Report timelines now include a backtrace for each event, making it easier to
determine the origin of the event (e.g. why something is looked up).
Timelines without this information will show 'unknown' in the relevant
column on the OOPS page. (Robert Collins, #890976)
* The OOPS reporter field - or prefix as it was known - is now shown
in the OOPS details page. (Robert Collins)
* The prefix and appinstance values can now be up to 100 characters long.
(Robert Collins, #892917)
* The req_vars variable in OOPS reports may now be a dict.
(Robert Collins, #888866)
* Added AMQP support via the bin/amqp2disk script. (Robert Collins)
* Bumped oops-amqp rev to 0.0.3 for bugfixes. (Robert Collins)
* amqp2disk now creates a fanout exchange if --bind-to is supplied.
* amqp2disk -v will print the received OOPS ids on the console, for
entertainment and delight. (Robert Collins)
* OOPS reports with non-ascii URL values are handled by url escaping the URL
bytestring (this is separate to handling of unicode URL values).
(Roman Yepishev, Robert Collins, #881400)
* OOPS reports with unicode URL values are now handled during oops loading: the
unicode string is utf8 encoded (an arbitrary choice) and url escaped.
(Robert Collins, #879309)
* OOPS reports with no topic (formerlly called pageid) will no longer use their
url instead. Rather '' is used, and reports will show Unknown for empty
pageids. (Robert Collins, #880641)
* Initial release
* Add dependency on testtools and fixtures for cleaner tests.
* Fix race conditions with oops loading and non-sequential oopses. These can
occur when e.g. rsync is writing at the same time as oops loading, or if a
non-sequential publisher is used. (Robert Collins, #667373)
* Update to oops-datedir-repo 0.0.7 and add reading of bson serialized oopses
(which are more extensible and more easily parsed). (Robert Collins)
TODO: Brief introduction on what you do with files - including link to relevant help section.