803248 - buildbot config changes to support panda_android*

Status: RESOLVED FIXED

(Infrastructure & Operations Graveyard :: CIDuty, task)

Comment 1

[Hal Wine [:hwine] use NI!]

This bug is most closely associated with putting the first pandas-for-android in production (they are running in staging now). Adding a few interested parties to the cc list for "up to the minute" status.

Comment 2

[Kim Moir [:kmoir] ET]

Attachment: buildbotcustom changes to support android on pandas

Comment 3

[Kim Moir [:kmoir] ET]

Attachment: buildbot config changes to support android on pandas

Not sure what slaves will be allocated to staging yet.  Also, I'm unsure of how to only enable a tests on the cedar branch.

Comment 4

[Armen [:armenzg]]

An example that might help is Win7x64 which is only enabled on mozilla-central:
http://hg.mozilla.org/build/buildbot-configs/file/default/mozilla-tests/config.py#l741
   747             'enable_opt_unittests': False,
   749             'enable_debug_unittests': False,

http://hg.mozilla.org/build/buildbot-configs/file/default/mozilla-tests/config.py#l955
   956 # XXX hacking warning - this code could get out of date easily
   957 BRANCHES['mozilla-central']['platforms']['win64']['enable_opt_unittests'] = True
   958 for suite in SUITES.keys():
   959     options = SUITES[suite]['options']
   960     if options[1] == ALL_PLATFORMS:
   961         options = (options[0], ALL_PLATFORMS + PLATFORMS['win64']['slave_platforms'])
   962     if options[1] == NO_MAC:
   963         options = (options[0], NO_MAC + PLATFORMS['win64']['slave_platforms'])
   964     if not SUITES[suite]['enable_by_default']:
   965         # Suites that are turned off by default
   966         BRANCHES['mozilla-central'][suite + '_tests'] = (0, True) + options
   967     else:
   968         # Suites that are turned on by default
   969         BRANCHES['mozilla-central'][suite + '_tests'] = (1, True) + options

I hope it helps; even though "cedar" additions are best suited inside of project_branches.py:
http://hg.mozilla.org/build/buildbot-configs/file/default/mozilla/project_branches.py#l238

It won't be an easy patch from my experience.

Comment 5

[Kim Moir [:kmoir] ET]

Attachment: buildbot config to enable tests for android + panda in cedar branch

Comment 6

[Aki Sasaki (not active)]

Comment on attachment 677493 [details] [diff] [review]
buildbot config to enable tests for android + panda in cedar branch

> -PLATFORMS['android']['slave_platforms'] = ['tegra_android']
> +PLATFORMS['android']['slave_platforms'] = ['tegra_android','panda_android']

Style nit: please put a space after the comma.

> +        [('panda-%04i' % x, {'http_port': '30%03i' % x, 'ssl_port': '31%03i' % x}) for x in range(22,34) + range(46,81)]
> +    ),             

Pet peeve: please remove trailing whitespace.

Comment 7

[Justin Wood (:Callek)]

Comment on attachment 677493 [details] [diff] [review]
buildbot config to enable tests for android + panda in cedar branch

Review of attachment 677493 [details] [diff] [review]:
-----------------------------------------------------------------

::: mozilla-tests/config.py
@@ +161,4 @@
>  PLATFORMS['android']['env_name'] = 'android-perf'
>  PLATFORMS['android']['is_mobile'] = True
>  PLATFORMS['android']['tegra_android'] = {'name': "Android Tegra 250"}
> +PLATFORMS['android']['panda_android'] = {'name': "Android Panda"}

Bikeshed-NIT: I'm more a fan of "Android ICS Panda" or "Android ICS Opt" or something

But this bikeshed should not block landing!

::: mozilla-tests/production_config.py
@@ +15,5 @@
>          if x not in range(122,129) + [30,31,33,34,43,44,49,65,69,77,131,137,143,147,\
>              153,156,161,175,176,180,184,185,186,193,197,198,202,203,204,205,222,224,\
>              226,239,241,268,275,289,291,292,301]]), # decommissioned tegras
> +     'panda_android': dict(
> +        [('panda-%04i' % x, {'http_port': '30%03i' % x, 'ssl_port': '31%03i' % x}) for x in range(22,34) + range(46,81)]

nit: *when* we get into 4 digit panda numbers here, these port specs will fail to work properly, we should plan to find a good solution there. but WFM for now

Comment 8

[Ed Morley [:emorley]]

(In reply to Justin Wood (:Callek) from comment #7)
> >  PLATFORMS['android']['tegra_android'] = {'name': "Android Tegra 250"}
> > +PLATFORMS['android']['panda_android'] = {'name': "Android Panda"}
> 
> Bikeshed-NIT: I'm more a fan of "Android ICS Panda" or "Android ICS Opt" or
> something

The proposal Clint made in bug 800047 was: "Android 4.0 Panda" (we would later also change "Android Tegra 250" to "Android 2.2 Tegra 250"). I tend to get confused with the codenames, so would prefer the version.

This also must have something between the "Android" and "Panda", or we'll have to make the regexes a bit hacky in TBPL to avoid causing false-positives against b2g panda or android 2.2.

Comment 9

[Aki Sasaki (not active)]

Ok.

I think we're ok to land anyway, as long as we fix up naming while we're still on Cedar.  Correct?

Comment 10

[Justin Wood (:Callek)]

(In reply to Aki Sasaki [:aki] from comment #9)
> Ok.
> 
> I think we're ok to land anyway, as long as we fix up naming while we're
> still on Cedar.  Correct?

correct, naming for a cedar-landing is not blocking, but still a nice-to-have. We'll discuss it amongst ourselves [releng] tomorrow with myself, kim, armen and coop, and if we have an agreement we'll land this with that, otherwise we'll land this as-is and adjust later without worry for TBPL.

Comment 11

[Kim Moir [:kmoir] ET]

Attachment: buildbot config to enable tests for android + panda in cedar branch

updated with correct name for pandas + excluded devices from chassis 2 (jmaher testing) and 6 (not working) from production

Comment 12

[Justin Wood (:Callek)]

Comment on attachment 677833 [details] [diff] [review]
buildbot config to enable tests for android + panda in cedar branch

r+ per interdiff

Comment 13

[Kim Moir [:kmoir] ET]

Attachment: fixed to exclude running non-cedar branch builds on android pandas

I forgot to explicitly exclude builds from running in non-cedar branches in the previous patch.  Here's the interdiff

> @@ -1028,16 +1030,22 @@ for branch in ('mozilla-central', 'mozil
>                  BRANCHES[branch]['platforms'][pf][slave_pf]['debug_unittest_suites'] += [('jetpack', ['jetpack'])]
>                  continue
>              if pf == "macosx64" and slave_pf == "leopard":
>                  BRANCHES[branch]['platforms'][pf][slave_pf]['opt_unittest_suites'] += [('jetpack', ['jetpack'])]
>                  continue
>              BRANCHES[branch]['platforms'][pf][slave_pf]['opt_unittest_suites'] += [('jetpack', ['jetpack'])]
>              BRANCHES[branch]['platforms'][pf][slave_pf]['debug_unittest_suites'] += [('jetpack', ['jetpack'])]
>  
> +#exclude android builds from running on non-cedar branches on pandas
> +for branch in ['mozilla-aurora', 'mozilla-beta', 'mozilla-release', 'mozilla-esr10'] + ACTIVE_PROJECT_BRANCHES:
> +    if 'android' in BRANCHES[branch]['platforms'] and branch != "cedar" :
> +        del BRANCHES[branch]['platforms']['android']['panda_android']
> +        BRANCHES[branch]['platforms']['android']['slave_platforms'] = ['tegra_android']
> +
>  # Let's load Marionette for the following branches:
>  for branch in ('mozilla-central', 'mozilla-inbound', 'try', 'fx-team', 'services-central', ):
>      for pf in PLATFORMS:
>          if 'android' in pf:
>              # this is just for desktop Firefox
>              continue
>          config_file = "marionette/prod_config.py"
>          if pf.startswith('win'):

Comment 14

[Aki Sasaki (not active)]

Comment on attachment 677855 [details] [diff] [review]
fixed to exclude running non-cedar branch builds on android pandas

That's not pretty, but it'll go away when we're ready.

Comment 15

[Kim Moir [:kmoir] ET]

Attachment: patch to revert addition of esr17

Comment 16

[Justin Wood (:Callek)]

Comment on attachment 678417 [details] [diff] [review]
patch to revert addition of esr17

Review of attachment 678417 [details] [diff] [review]:
-----------------------------------------------------------------

Sorry for the think-o when reading this part of the patch and making that suggestion

Comment 17

[Kim Moir [:kmoir] ET]

In production

Two things 
-Android 2.2 builds are runing in cedar builds, we don't know why, will investigate tomorrow
-Android builders appear on the bm29, this should be disabled, only panda ones should appear

Comment 18

[Aki Sasaki (not active)]

Attachment: fix pandas

Comment 19

[Kim Moir [:kmoir] ET]

Last patch is in production

Comment 20

[Kim Moir [:kmoir] ET]

Attachment: patch to add 66 pandas to production + enable bm44

As imaged in bug 820825 (pandas-[0808-0873] which are panda-chassis-[073-078])

Comment 21

[Justin Wood (:Callek)]

Comment on attachment 691511 [details] [diff] [review]
patch to add 66 pandas to production + enable bm44

Review of attachment 691511 [details] [diff] [review]:
-----------------------------------------------------------------

Beyond skimming I did not do a close sanity check (e.g. are the panda-relays/id's correct, is there a typo in there, etc. I leave as a 'I trust you did it right')

I did however verify that all relay's used are identified as "Configured" in Bug 809932 properly.

::: buildfarm/mobile/devices.json
@@ +2306,5 @@
> +    "panda-0840": {
> +        "foopy": "foopy99",
> +        "relayhost": "panda-relay-075.p9.releng.scl1.mozilla.com",
> +        "relayid": "2:7"
> +    },    

nit whitespace EOL

Comment 22

[Kim Moir [:kmoir] ET]

Attachment: expand range of pandas in prod + enable builds on m-c

Comment 23

[Justin Wood (:Callek)]

(In reply to Justin Wood (:Callek) from comment #21)
> Comment on attachment 691511 [details] [diff] [review]
> patch to add 66 pandas to production + enable bm44

So, nick just hit probs while reconfig'ing this host, since its auth_keys was out of date. And I also noticed the master was not even started.

While I was there I also noticed it didn't have any keys on the host, which is needed for log uploads. for histroy. I added the keys (recursive scp from bm22), but did not start the master.

Comment 24

[Kim Moir [:kmoir] ET]

Attachment: patch to fix how foopies line up

Comment 25

[Justin Wood (:Callek)]

Comment on attachment 692318 [details] [diff] [review]
patch to fix how foopies line up

Review of attachment 692318 [details] [diff] [review]:
-----------------------------------------------------------------

::: buildfarm/mobile/devices.json
@@ +2308,3 @@
>          "relayhost": "panda-relay-075.p9.releng.scl1.mozilla.com",
>          "relayid": "2:7"
>      },    

nit whitespace EOL

Comment 26

[Kim Moir [:kmoir] ET]

Attachment: add new pandas in racks 7 and 8 as requested in bug 824766

Note:   This patch is just to assign the pandas to foopies in racks 7 and 8.  There are still some more to be assigned in rack 9.  

I didn't realize that we should be assigning 12 or 13 boards to each foopy when I set up the 66 pandas (rack 9) earlier. Previously, I assigned 11 boards so each chassis would map to a foopy. But this doesn't scale because we don't have enough foopies in the matching network to do this.

I'll reorganize the board assignments for the foopies in rack 9 once these new boards are online.

Comment 27

[Armen [:armenzg]]

Comment on attachment 697467 [details] [diff] [review]
add new pandas in racks 7 and 8 as requested in bug 824766

I attached in some bug a script that helps sorting out the assignment of boards to foopies taking into consideration the 88 to 7 foopies ratio.

Tangential to the patch:
Is rack #6 going to be fully used? I only see some of its devices listed on devices.json.

Comment 28

[Kim Moir [:kmoir] ET]

Thanks for the review Armen.

Right now Joel is just using some of the devices in rack 6, see bug 825876 for details.  I assume we'll use the rest if them too.

Comment 29

[Armen [:armenzg]]

What is left on this bug? (trying to do new year cleanup)

Comment 30

[Kim Moir [:kmoir] ET]

The only thing that's left is to do is to update the devices.json as I put more panda boards into production so I can attach those patches to other bugs.  Closing.