New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 891929 link

Starred by 1 user

Issue metadata

Status: Started
Owner:
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Windows
Pri: 1
Type: Feature

Blocking:
issue 867155



Sign in to add a comment

Install IntelPowerGadget 3.5 on Windows 10 Intel GPU bots

Project Member Reported by zmo@chromium.org, Oct 3

Issue description

Blocking: 867155
Cc: nednguyen@chromium.org
Ken, Ned, how do we find out the list of Win10/Intel bots on perf waterfall? We also want to install IntelPowerGadget 3.5 on these bots, so we could add power perf tests on them.
zmo@: you can see this file:

https://chrome-internal.googlesource.com/infradata/config/+/master/configs/chrome-swarming/bots.cfg

  bot_id: "build{117..121}-a7"  # win-10-perf
  bot_id: "build{255..281}-a7"  # win-10-perf

Do you also want to install IntelPowerGadget 3.5 for pinpoint?
Thanks Ned. Yes, we do if we have pinpoint for Win10/Intel.
The pinpoint ones are:

  bot_id: "build{219..223}-a7" # win-10-perf
  bot_id: "build{282..292}-a7" # win-10-perf
Labels: -Type-Bug Type-Feature
Let me recategorize this as a feature, but it's still quite urgent for our team. Labs team, thanks in advance for your help getting this software package rolled out.

Owner: actodd@chromium.org
Status: Assigned (was: Untriaged)
Status: Started (was: Assigned)
In summary, you guys are requesting the following sets of bots have IPG 3.5 installed.

https://chromium-swarm.appspot.com/botlist?c=id&c=os&c=task&c=status&f=gpu%3A8086%3A5912-24.20.100.6286&f=os%3AWindows-10&f=cpu%3Ax86-64&f=pool%3AChrome-GPU&l=100&s=id%3Aasc

bot_id: "build{117..121}-a7"  # win-10-perf
bot_id: "build{255..281}-a7"  # win-10-perf
bot_id: "build{219..223}-a7"  # win-10-perf
bot_id: "build{282..292}-a7"  # win-10-perf

#7. you are correct
I am starting this process now. While doing it I am going to explore some potential automation so it could take a couple days before its completely done. I will update the bug as I go.
Thank you!
Can you list those win10 perf bot hostnames in a more programmatic way instead of hostname or is that really the best way?
Expanding on what I said in #11 ... I mean can you list them using swarming dimensions or do we have to track them using hostnames?
Cc: -nednguyen@chromium.org jbudorick@chromium.org nedngu...@google.com dtu@chromium.org
#12: for now, https://chrome-swarming.appspot.com/botlist?c=id&c=os&c=task&c=status&c=pool&f=os%3AWindows-10&f=gpu%3AIntel%20(8086)&l=100&q=pool&s=id%3Aasc seems to work well (align with #7)

Later we need to do similar query but limit to the perf pools.

Interestingly, it seems like there are inconsistencies in term of machine configs :-o
Project Member

Comment 15 by bugdroid1@chromium.org, Oct 11

The following revision refers to this bug:
  https://chrome-internal.googlesource.com/infra/puppet/+/9f0fbe8df8283ecfc01cadb9a9abce02f7ed8a31

commit 9f0fbe8df8283ecfc01cadb9a9abce02f7ed8a31
Author: Adam Todd <actodd@google.com>
Date: Thu Oct 11 16:07:44 2018

If you can provide me with a more sustainable set of swarming query dimensions I would love to use them instead. The benefit to you will be that all your machines that meet these query parameters will get this software no matter what kind of churn happens to specific hosts.
The first attempt at puppet+cipd package deployment has been committed. Later today i will spot check the hosts to see if it has worked. I would expect there to be hick ups since this was an essentially undocumented process.
I don't fully understand the terminologies you used in #17, so I am asking: are you saying by later today Intel Power Gadget might be installed on all the machines listed in this bug?
Project Member

Comment 19 by bugdroid1@chromium.org, Oct 11

The following revision refers to this bug:
  https://chrome-internal.googlesource.com/infra/puppet/+/5f288653c366ee9e615b2f593b5e4deaba932ed4

commit 5f288653c366ee9e615b2f593b5e4deaba932ed4
Author: Adam Todd <actodd@google.com>
Date: Thu Oct 11 21:55:36 2018

#18: Yes.

As expected we made a mistake on our first try and have pushed out a change to the puppet config. Hopefully this configuration will work properly.
Let us know when the bots are ready. Then I'll add tests using IPG to these bots.
Gentle ping: any updates?
It should be ready by now.  Can you test and see if it works?

How can I tell if it's installed properly?

It's only working on swarming, I just noticed a bug in the way the hardcoded hosts were entered.  Fixing that now...
Cc: actodd@chromium.org
Owner: zmo@chromium.org
I will upload a CL to run some tests on Win/Intel bots. If they pass, we know they install correctly.

Let me take this bug from this point since the next action is in my hand.
By the way, thanks a lot for getting these bots set up.
Project Member

Comment 26 by bugdroid1@chromium.org, Oct 13

The following revision refers to this bug:
  https://chrome-internal.googlesource.com/infra/puppet/+/44dddb4743d3ad35bfd19e1a20831a1cfc862ce1

commit 44dddb4743d3ad35bfd19e1a20831a1cfc862ce1
Author: Elliott Friedman <friedman@google.com>
Date: Sat Oct 13 00:00:53 2018

It's not working:
E2018-10-12_17:02:10 24]  Error: Could not retrieve catalog from remote server: Error 400 on SERVER: Could not find class chomre_infra::packages::cipd::intel_power_gadget for build117-a7.labs.chromium.org on node build117-a7.labs.chromium.org

Give me a few to fix.
Owner: friedman@chromium.org
Assign to you. When you are done, assign back to me for testing.
Project Member

Comment 29 by bugdroid1@chromium.org, Oct 13

The following revision refers to this bug:
  https://chrome-internal.googlesource.com/infra/puppet/+/2bea4cf90afcaad7901d450161b6e1b6788c0fc5

commit 2bea4cf90afcaad7901d450161b6e1b6788c0fc5
Author: Elliott Friedman <friedman@google.com>
Date: Sat Oct 13 00:08:52 2018

Project Member

Comment 30 by bugdroid1@chromium.org, Oct 13

The following revision refers to this bug:
  https://chrome-internal.googlesource.com/infra/puppet/+/e890e98f68803034c5436cf2136311d845b557b5

commit e890e98f68803034c5436cf2136311d845b557b5
Author: Elliott Friedman <friedman@google.com>
Date: Sat Oct 13 00:15:47 2018

More errors:
$ C:/chrome-infra/labs/intel_power_gadget/.cipd/cipd_v_017882833e28fc8ce98ad825eb90de3776d1ca22.exe ensure -root C:/chrome-infra/labs/intel_power_gadget -list C:/chrome-infra/labs/intel_power_gadget/.cipd/tmp/ensure_packages -service-url https://chrome-infra-packages.appspot.com -service-account-json C:/creds/service_accounts/service-account-cipd-generic-auth.json
Error: resolving package: failed to resolve package version (line 1): prefix "infra_internal/labs/intel_power_gadget/windows-amd64" doesn't exist or the caller is not allowed to see it.

Still digging...
Project Member

Comment 32 by bugdroid1@chromium.org, Oct 13

The following revision refers to this bug:
  https://chrome-internal.googlesource.com/infra/puppet/+/43334be9f8dae7d6ef47b31f1e912865dbdbc251

commit 43334be9f8dae7d6ef47b31f1e912865dbdbc251
Author: Elliott Friedman <friedman@google.com>
Date: Sat Oct 13 00:30:45 2018

Project Member

Comment 33 by bugdroid1@chromium.org, Oct 13

The following revision refers to this bug:
  https://chrome-internal.googlesource.com/infra/puppet/+/edc0a106c77cc5eb22d6195a55b96c1179467e65

commit edc0a106c77cc5eb22d6195a55b96c1179467e65
Author: Elliott Friedman <friedman@google.com>
Date: Sat Oct 13 00:37:59 2018

ok finally!

chrome-bot@build117-a7 /cygdrive/c/chrome-infra-logs
$ ls ../chrome-infra/labs/intel_power_gadget/
'Intel Power Gadget EULA.rtf'   IntelPowerGadgetUserGuide.pdf   PowerGadget.msi

It should be fully rolled out within 1 hour.
Owner: zmo@chromium.org
zmo: please see https://chrome-internal-review.googlesource.com/c/infra/puppet/+/698868 for the exact reason that I'd rather not use static hostnames.  Can swarming dimensions not be used for this?
Project Member

Comment 37 by bugdroid1@chromium.org, Oct 16

The following revision refers to this bug:
  https://chrome-internal.googlesource.com/infra/puppet/+/e1b4cda2fc95761269519f8fa7b6e96683ac800f

commit e1b4cda2fc95761269519f8fa7b6e96683ac800f
Author: Bryce Albritton <dba@google.com>
Date: Tue Oct 16 18:34:34 2018

Re #36 -- the CL referenced actually targets the proper pools + dimenisons that were mentioned earlier in the bug (#1, #4) instead of hostnames as it was noticed that some machines that were targeted were not in scope (wrong platform/not even deployed).

Sign in to add a comment