759724 - Add "Webapp Runtime" to supported products in Socorro

Status: RESOLVED FIXED

(Socorro :: General, task)

Description

[Robert Kaiser]

Unless the answer to bug 745980 comment #22 is that we want to merge those reports with Firefox, we'll need to add a "Webapp Runtime" product to Socorro. I hope to resolve that question today.

Comment 1

[Robert Kaiser]

I sat down with myk and jsmith today and talked about thism and the decision is to go forward with rewriting to "Webapp Runtime" and adding that as a new product to Socorro. (In case the space in the name might be a problem for Socorro, we'll go with "WebappRuntime".)

Laura, Rob, are we able to do that in a way that this is going to production within the next few weeks?

Comment 2

[[:jberkus] Josh Berkus]

I need a full specification on this before I go any further.  Is there a wiki page about this?

Comment 3

[Jason Smith [:jsmith]]

(In reply to [:jberkus] Josh Berkus from comment #2)
> I need a full specification on this before I go any further.  Is there a
> wiki page about this?

What information are you looking for on this in this wiki?

Comment 4

[Robert Helmer [:rhelmer]]

(In reply to Jason Smith [:jsmith] from comment #3)
> (In reply to [:jberkus] Josh Berkus from comment #2)
> > I need a full specification on this before I go any further.  Is there a
> > wiki page about this?
> 
> What information are you looking for on this in this wiki?

Josh and I were just discussing, and we wonder if it wouldn't make more sense to add "Webapp Runtime" as a new process type for Firefox (like browser, plugin, content) ?

Since it ships with Firefox it breaks some of our assumptions about what a "product" is, so we'd need to duplicate info from Firefox. 

This means less work on implementation side, but it's more important that this is useful to Socorro users, so let us know.

Comment 5

[Robert Kaiser]

I discussed this with Rob, we are having different ADU numbers and blocklists, and different add-ons (actually none at all) on Webapp Runtime, so adding it as a process type doesn't really make sense.
We need to do it as a new product but special-case it in the way that it uses the same build info as Firefox (desktop).

Comment 6

[[:jberkus] Josh Berkus]

Kairo,

I'll be working on this today and it should go into version 14 of socorro, whenever we release that.

Comment 7

[Robert Kaiser]

Cool, thanks, Josh, looking forward to that!

Comment 8

[[:jberkus] Josh Berkus]

OK, SQL is written for this:

https://github.com/mozilla/socorro/pull/655

I'd like Adrian to review that, but it's currently running on Crash-Stats-Dev for testing.

Comment 9

[[:jberkus] Josh Berkus]

This is written and ready to pull.

Comment 10

[[:jberkus] Josh Berkus]

Done.

QA: verify that WebRuntime is available as a product in Socorro.

Currently not working due to lack of data.  Fixing via backfill.

Comment 11

[Jason Smith [:jsmith]]

(In reply to [:jberkus] Josh Berkus from comment #10)
> Done.
> 
> QA: verify that WebRuntime is available as a product in Socorro.
> 
> Currently not working due to lack of data.  Fixing via backfill.

Could you clarify what backfilling means? And will the existing crashes in small number still show up? I see the product on crash-stats.mozilla.com, but as you've specified, I get an error saying no data was found.

Comment 12

[Robert Kaiser]

(In reply to Jason Smith [:jsmith] from comment #11)
> I see the product on crash-stats.mozilla.com,
> but as you've specified, I get an error saying no data was found.

I corrected that error. We need at least one "featured" version to be set for the front page to work, I did set the only versions we have right now, i.e. 16.0a1 as "featured", so we have *something* now.

Comment 13

[Jason Smith [:jsmith]]

(In reply to Robert Kaiser (:[e-mail]) from comment #12)
> (In reply to Jason Smith [:jsmith] from comment #11)
> > I see the product on crash-stats.mozilla.com,
> > but as you've specified, I get an error saying no data was found.
> 
> I corrected that error. We need at least one "featured" version to be set
> for the front page to work, I did set the only versions we have right now,
> i.e. 16.0a1 as "featured", so we have *something* now.

Good. Looks like we're getting top crash data too:

https://crash-stats.mozilla.com/topcrasher/byversion/WebappRuntime/16.0a1/7

I think using Firefox versioning is fine, unless someone else has a concern on it (as we follow the Firefox release train). Issues in the data right now I see is:

1. No data shown for Crashes per 100 Active Daily Users - ADU appears to be zero
2. Report for Nightly builds - shows nothing
3. Hang reports - nothing, but are we even collecting this?
4. Top Changes - nothing, but are we even analyzing this?

Other notes of possible issues:

- Process Type is called browser, but web apps have no concept of a browser - is this is a bug? Should this be web app?

Any ideas on the issues on the web runtime end? The socorro end? I know we recently just added support for blocklists in the web runtime on a per app basis, so could that be used to help calculate ADU?

Comment 14

[Robert Helmer [:rhelmer]]

(In reply to Jason Smith [:jsmith] from comment #13)
> - Process Type is called browser, but web apps have no concept of a browser
> - is this is a bug? Should this be web app?

The only process types that Socorro knows about are "plugin", "browser" and "content". 

In fact "browser" is what we use when no process type is specified, so we could perhaps change this if you want it to say something else, but it's currently not indicative of a problem and is expected behavior AFAICT.

Comment 15

[Robert Kaiser]

(In reply to Jason Smith [:jsmith] from comment #13)
> 1. No data shown for Crashes per 100 Active Daily Users - ADU appears to be
> zero

This is because  metrics doesn't separate out the ADU data yet (and not pushing that to Socorro), they are working on it.

> 2. Report for Nightly builds - shows nothing

Probably expected, as WebRT doesn't have separate nightly builds. If we want to change this, a separate bug should be filed.

> 3. Hang reports - nothing, but are we even collecting this?

This is only for Plugins, does WebRT even run those? Even then, we probably have too few reports in to see plugin hangs among them.

> 4. Top Changes - nothing, but are we even analyzing this?

That only comes with time when there has been a useful topcrash list for enough time that changes from one day to the other make sense.


> - Process Type is called browser, but web apps have no concept of a browser

That's OK and expected, should not disturb us. See Rob's comment on that as well. In the end, "browser" is not so wrong for WebRT as it might sound for Thunderbird, for example. ;-)
It just means it's a crash in our main application process ("plugin" or "content" would be different processes run from the main one).

Comment 16

[Robert Kaiser]

(In reply to Robert Kaiser (:[e-mail]) from comment #15)
> (In reply to Jason Smith [:jsmith] from comment #13)
> > 1. No data shown for Crashes per 100 Active Daily Users - ADU appears to be
> > zero
> 
> This is because  metrics doesn't separate out the ADU data yet (and not
> pushing that to Socorro), they are working on it.

Bug 770166 is tracking collection of that data, bug 765084 pushing it to Socorro.

Comment 17

[Jason Smith [:jsmith]]

(In reply to Robert Kaiser (:[e-mail]) from comment #15)
> > 3. Hang reports - nothing, but are we even collecting this?
> 
> This is only for Plugins, does WebRT even run those? Even then, we probably
> have too few reports in to see plugin hangs among them.

Only flash intends to be ran in the web runtime.