Launchpad itself

Bug watch updater will try to update invalid bug watches forever

Bug #67480 reported by Stuart Bishop
6
Affects Status Importance Assigned to Milestone
Launchpad itself
Fix Released
High
Graham Binns
Launchpad itself 10.03

Bug Description

The bug watch updater keeps no state between runs, so if someone enters an invalid bug id then that invalid bug watch will be probed every day for ever.

This also means that errors that we might need to deal with a buried in a torrent of "Didn't find bug" error messages.

First step in solving this is to reduce this error message down to an INFO - we expect this situation to happen so it isn't an error, and our code should eventually cope with it.

At the moment we determine whether or not to check a BugWatch based on the time elapsed since its lastchecked date. Once bug 538095 has landed we should alter BugTracker.getBugWatchesNeedingUpdate() to take account of BugWatchActivity, so that BugWatches which have errored more than a certain number of times in their last N checks get checked less often than those which have never errored.

NB: There are many smaller fixes that need to be made for this bug to be considered fixed and those fixes will be tracked individually. We can fix this bug when all other bugs tagged story-bug-watch-error-tracking[1] are complete.

 [1] https://bugs.edge.launchpad.net/malone/+bugs?field.tag=story-bug-watch-error-tracking

See original description
Diogo Matsubara (matsubara)
Changed in malone:
status: Unconfirmed → Confirmed
Revision history for this message
Christian Reis (kiko) wrote :

I don't agree with bumping down the errors into INFO until the code can cope with invalid watches. The reason is that I use the updater output to go in and garden the bug watches that are wrong. I agree that we need better reporting of these, but I don't think we're there yet, and I'm not sure if it will be a short-term priority.

I have been changing to zero the ID of bug watches that are /really/ invalid. If you could delete those, Stuart, I'd appreciate it. I can then vouch to keep the error log output at zero there!

Revision history for this message
Björn Tillenius (bjornt) wrote : Re: [Bug 67480] Re: Bug watch updater will try to update invalid bug watches forever

On Mon, Dec 04, 2006 at 08:46:23PM -0000, Christian Reis wrote:
> I don't agree with bumping down the errors into INFO until the code can
> cope with invalid watches. The reason is that I use the updater output
> to go in and garden the bug watches that are wrong. I agree that we need
> better reporting of these, but I don't think we're there yet, and I'm
> not sure if it will be a short-term priority.

I agree with kiko here, the first step isn't to hide the errors, it to
make them visible elsewhere. There should probably be some UI for
displaying all the errors, so that users can fix them manually, but
that's not going to happen anytime soon.

Stuart, if you think that the script produces too much output, maybe you
could move it out from nightly.sh and run it separately? That way you
can create a new mailman topic for it, so that people interested in the
output can subscribe to it, and the rest don't have to see it. I would
actually like that, since there are two other scripts in nightly.sh that
time from time produce way more output than the bug watches script.

Gavin Panella (allenap)
tags: added: story-reliable-bug-syncing
Revision history for this message
Gavin Panella (allenap) wrote :

We could have an error_count on a bug watch. For each error without an intervening successful update, it is incremented. Also, it is not considered eligible for checking until it is (24 * error_count) hours old. We should then have a report or some other way to deal with bug watches that have an error_count of greater than, say, 4 or 5.

Revision history for this message
Gavin Panella (allenap) wrote :

I'm marking this as High importance because it will have beneficial knock on effects: reduce the OOPS report noise, reduce the load on checkwatches and remote systems, help us garden bug watches because we'll know those that are persistently failing (at the moment we don't have any record of how long a bug watch has been failing).

Changed in malone:
importance: Undecided → High
status: Confirmed → Triaged
Graham Binns (gmb)
description: updated
tags: added: story-bug-watch-error-tracking
Graham Binns (gmb)
description: updated
Revision history for this message
Graham Binns (gmb) wrote :

We can consider this fixed when bug 545909 lands.

Changed in malone:
milestone: none → 10.03
status: Triaged → In Progress
Graham Binns (gmb)
Changed in malone:
status: In Progress → Fix Committed
assignee: nobody → Graham Binns (gmb)
Gavin Panella (allenap)
tags: added: qa-ok
Graham Binns (gmb)
Changed in malone:
status: Fix Committed → Fix Released
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.
Subscribing...

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.