1088354 - Consider switching builds-4hr.js to builds-2hr.js

Status: RESOLVED WONTFIX

(Infrastructure & Operations :: RelOps: General, task, P2)

Description

[Jonathan Griffin (:jgriffin)]

Treeherder currently reads builds-4hr.js frequently, and has to parse the entire file, figure out what's new, and process that.  There's significant overhead to this; it would be easier for Treeherder if this file were smaller, e.g., builds-2hr.js.  In an earlier conversation with catlee, I recall him saying that this would make generation of this file easier on the relops side as well.

Can we consider making this change?

In order for this to be successful, both Treeherder and TBPL would have to be modified with to use a new url (unless builds-4hr.js wasn't renamed, but changed to contain just 2 hours of data).  We'd also want to implement a fallback in Treeherder ala bug 1069467 for the case where we were down for more than 2 hours, so we'd have a mechanism for populating data not contained in builds-2hr.js.

Ultimately, we want builds-4hr to go away (see bug 1026109), and to rely on pulse instead, but there are several issues that would need to be fixed first, and switching to a 2-hour format might reduce pressure on both relops and Treeherder in the interim.

Comment 1

[Dustin J. Mitchell [:dustin] (he/him)]

We've already considered it -- it was my idea :)

I never got a thumbs-up, though -- is this bug that thumbs-up?

Assuming it is..

Putting 2 hours of builds into builds-4hr.js seems like the kind of mistake we make all the time -- "this is just temporary", but makes no sense and is still in place 3 years later.  So, let's put the two hours' data into builds-2hr.js.  And, magically, symlink builds-4hr.js to that file for compatibility.  Then we can make the TBPL and Treeherder changes, and monitor for hits on builds-4hr.js to make sure we got them all.  Then delete the file.

So, I'm just waiting on a green light, and I nominate :jgriffin for that :)

Comment 2

[Ed Morley [:emorley]]

(In reply to Jonathan Griffin (:jgriffin) from comment #0)
> In order for this to be successful, both Treeherder and TBPL would have to
> be modified with to use a new url (unless builds-4hr.js wasn't renamed, but
> changed to contain just 2 hours of data).  We'd also want to implement a
> fallback in Treeherder ala bug 1069467 for the case where we were down for
> more than 2 hours, so we'd have a mechanism for populating data not
> contained in builds-2hr.js.

In addition to this, I believe we'll need bug 765451, since otherwise this "fallback import of the daily archive, run by IT oncall in case of emergency" won't have the partial day's jobs, and so leave a gap.

Comment 3

[Dustin J. Mitchell [:dustin] (he/him)]

Well, that was easily fixed :)

Comment 4

[Jonathan Griffin (:jgriffin)]

(In reply to Dustin J. Mitchell [:dustin] from comment #1)
> We've already considered it -- it was my idea :)
> So, let's put the two hours' data into
> builds-2hr.js.  And, magically, symlink builds-4hr.js to that file for
> compatibility.  Then we can make the TBPL and Treeherder changes, and
> monitor for hits on builds-4hr.js to make sure we got them all.  Then delete
> the file.
> 
> So, I'm just waiting on a green light, and I nominate :jgriffin for that :)

Agreed. Let's do it as soon as bug 1069467 is fixed.

Comment 5

[:kanban-engops]

A Pivotal Tracker story has been created for this Bug: https://www.pivotaltracker.com/story/show/81866200

Comment 6

[:kanban-engops]

Dustin J. Mitchell added a comment in Pivotal Tracker:   
   
pending a dependent bug

Comment 7

[Dustin J. Mitchell [:dustin] (he/him)]

Well, six months later, I guess this wasn't a thumbs-up..

Comment 8

[Ed Morley [:emorley]]

I'm pretty sure we may now need do this as part of the switch to Heroku, given bug 1151806 comment 0. Bug 1069467 is still a prerequisite for this bug, but now has higher priority.

Comment 9

[Ed Morley [:emorley]]

Bug 1178240 (landed since comment 8) made the need for this much less, let's just wontfix.