bugmessage.index is not populated

[Robert Collins]

BugMessage.index exists in the schema to let BugTask:+index be more efficient, but its not populated.

To populate it we need a multistep process:
 step 1: a garbo hourly job that does a small amount of conversions. It needs to do at least 175 per garbo invocation if we want a 6 month completion date. I think that most bugs are small enough we could even set the minimum to 350 : the auto loop tuner will do more if more is acceptable given db load and time taken.
  --- Done

 step 2: start setting BugMessage.index when adding BugMessages - this will require calculating and setting the BugMessage.index for all other messages on the bug at the same time for consistency.
  --- Done

 step 3: when fully populated, make the bugmessage.index column NON-NULL, remove the garbo job to set bugmessage.index and remove the calculation of bugmessage.index for existing messages when adding a new message.
  --- Done

 step 4: We can fix bug 607935

[Launchpad QA Bot (lpqabot)]

Fixed in stable r12289 <http://bazaar.launchpad.net/~launchpad-pqm/launchpad/stable/revision/12289>.

[Robert Collins (lifeless)]

step 1 runs fine on qastaging (after we sorted a bunch of qastaging yaks out. Thanks spm!). Its doing > 175 - a lot more. So we'll be able to move forward soon.

[Robert Collins (lifeless)]

On staging one run of garbo hourly populates:

launchpad_qastaging=> select count( bugmessage.id ) from bugmessage where index is not null limit 50;
 count
-------
 68532
(1 row)

[Robert Collins (lifeless)]

We're 90% through the indexing process; I'm going to start prepping the next phase now.

[Robert Collins (lifeless)]

200K left to index.

[Robert Collins (lifeless)]

Completely indexed.

[Launchpad QA Bot (lpqabot)]

Fixed in stable r12365 (http://bazaar.launchpad.net/~launchpad-pqm/launchpad/stable/revision/12365) by a commit, but not testable.

[Robert Collins (lifeless)]

bug 607935 can now be worked on: the live code will always set bugmessage.index and reindexes if needed before doing anything else. There are nounindexed bugmessages.

[William Grant (wgrant)]

"always" except for cocoplum and germanium, who add bugmessages but are outside nodowntime.

[Robert Collins (lifeless)]

On Tue, Feb 15, 2011 at 11:09 AM, William Grant <email address hidden> wrote:
> "always" except for cocoplum and germanium, who add bugmessages but are
> outside nodowntime.

oh grrrrrrrrr

Matt, can I have a downtime window for these two services please, just
a simple deploy of the new code which will keep this column intact.

[Launchpad QA Bot (lpqabot)]

Fixed in stable r12393 (http://bazaar.launchpad.net/~launchpad-pqm/launchpad/stable/revision/12393) by a commit, but not testable.