timeout in project code page

[Diogo Matsubara]

As seen in OOPS-989EB101 the +code-index page for bzr timed out. Some people reported that page was constantly timing out.

Some other occurences: OOPS-990B3705, OOPS-990D1962, OOPS-990F150
Related: OOPS-1038EA132, OOPS-1038EA133, OOPS-1038EB142, OOPS-1038ED115

After 2.1.11 rollout:OOPS-1055EA104, OOPS-1055EA105, OOPS-1055EA116, OOPS-1059S691, OOPS-1059S686, OOPS-1059S689

Recent soft timeouts: OOPS-1063EA15, OOPS-1063EA16, OOPS-1063EA162, OOPS-1063EA185, OOPS-1063EA25, OOPS-1063EA93, OOPS-1063EB106, OOPS-1063EB154, OOPS-1063EB35, OOPS-1063EC152

Recent hard timeouts: edge(r7374): OOPS-1068EA193, OOPS-1068EA207, OOPS-1068EA208, OOPS-1068EB172, OOPS-1068EB194, OOPS-1068EB196, OOPS-1068ED154, OOPS-1068ED183, OOPS-1068EA186, OOPS-1068EA187, OOPS-1068EA198, OOPS-1068EC155, OOPS-1068EC157, OOPS-1068EC169, OOPS-1068EC229

[Tim Penhey (thumper)]

Fixed in RF 7035.

[Joey Stanford (joey)]

Thumper: Landed before IR was required thus no IR required for this one.

[Jonathan Lange (jml)]

This doesn't look fixed to me. https://code.edge.launchpad.net/launchpad times out (Error ID: OOPS-998EC32).

[Diogo Matsubara (matsubara)]

Hi Tim, can you check this bug again?

I have some samples of post rollout hard time outs: OOPS-996S7, OOPS-996A2672 and lots of soft time outs: OOPS-996S8, OOPS-997A1735, OOPS-997A1546, OOPS-997A1674, OOPS-997B1937

[Tim Penhey (thumper)]

This appears to be two different things now:

The first is not caching the security checks. This is being worked on by Michael and the branch is currently being reviewed.

The second is a slow query for getting the revisions. It is much slower than I expected, so I'll chase it up with stub or jtv.

[Tim Penhey (thumper)]

I'm downgrading to High as we have stopped the project listing for active projects timing out every time.

Now it is private branch listings and speeding up the revision count checks.

[Tim Penhey (thumper)]

I have a branch that will save up to another 2s, however the vast majority of the time is in the query for the latest revisions.

Stuart, can we talk again about how to optimise this query?

[Tim Penhey (thumper)]

OK. I'll look to change the query to do this.

Also the older OOPSes that you were looking at with all the repeated branch subscription queries was due to the security adapter not caching checks. This was fixed several releases ago.

[Stuart Bishop (stub)]

The lp:~stub/launchpad/pending-db-changes branch contains the indexes in patch-2109-14-0.sql