What is the status on this?  We are ready to start up our LUCI CI builders, once we have the machines?

We would also like to learn to do more of this ourselves, and learn all the systems involved.  Are these just started as project instances, or is there a manager that starts and names them?  Then I know there is a puppet configuration for them, and then the configuration in our swarming pool, which we have done in the past.  I wonder what other steps there are.

Was out of office yesterday.

Still working on making Puppet work with the new VMs. Currently only our GCE project is recognized, so I'm working on making it recognize dart-ci as well.

The following revision refers to this bug:
  https://chrome-internal.googlesource.com/infradata/config/+/ef5d3dc3ee133d6fedc7b0b3dadeb92e61056460

commit ef5d3dc3ee133d6fedc7b0b3dadeb92e61056460
Author: smut <smut@google.com>
Date: Wed Nov 01 20:47:09 2017

I think we managed to get Puppet working, but it looks like you only have 8 IP addresses:
https://pantheon.corp.google.com/iam-admin/quotas?project=dart-ci&service=compute.googleapis.com&metric=In-use%20IP%20addresses&usage=USED

Please request at least 60 IP addresses-- each VM will need its own public IP. I'll also need to know the full list of IP addresses so we can whitelist them for access to our Puppet master.

Er, actually all your quota is too low. For example you only have 24 CPUs in us-central1, which is enough for 3 8-core VMs. For 40 Linux and 20 Windows you will need:
1. >= 60 IP addresses
2. >= 480 CPUs
3. >= 18,000 GB persistent disk

All three of which need to be in us-central1 specifically.

I have made a request for twice that amount.  I requested static IP addresses, since I did not know whether this is needed or not.  When I get the request back, I'll follow up.

Google Compute Engine API

A request (ID:500f200001AdraKAAR) has been made for the following quotas
CPUs - us-central1
CPUs (all regions)
Static IP addresses - us-central1
In-use IP addresses - us-central1
Static IP addresses global
In-use IP addresses global
Persistent Disk Standard (GB) - us-central1

Quota is now approved:
Hello,

Your quota request for project '410721018617' has been approved and your quota has been adjusted accordingly.

The following quotas were increased:
+---------------------+-----------+----------------+-------------+------------------+------------------+
| Region: us-central1 |    CPUS   | DISKS_TOTAL_GB |  INSTANCES  | IN_USE_ADDRESSES | STATIC_ADDRESSES |
+---------------------+-----------+----------------+-------------+------------------+------------------+
|       Changes       | 24 -> 960 | 4096 -> 36000  | 240 -> 9600 |     8 -> 120     |     8 -> 120     |
+---------------------+-----------+----------------+-------------+------------------+------------------+

+------------------+------------------+------------------+------------------+
| GLOBAL Attribute | CPUS_ALL_REGIONS | IN_USE_ADDRESSES | STATIC_ADDRESSES |
+------------------+------------------+------------------+------------------+
|     Changes      |    64 -> 960     |     8 -> 120     |     8 -> 120     |
+------------------+------------------+------------------+------------------+


Please visit https://console.cloud.google.com/iam-admin/quotas?project=410721018617&service=compute.googleapis.com to review your updated quota.

Happy Computing!

I see that the 40 linux machines have now been created in our GCE project.  They have an empty /b directory at this point.  Should I add them to the Dart.LUCI pool config, or is that done automatically?  Is there any step I need to take, or will they show up in the Dart.LUCI pool after chrome-infra does all its work?

If these don't get Chrome installed, through the puppet configuration that should add Chrome to all machines in the pool, then I will add it manually.

Did they tell you the 120 IP addresses you got? I tried to list them but it's not working:
$ CLOUDSDK_CORE_PROJECT=dart-ci gcloud compute addresses list
Listed 0 items.

I need to know the IP addresses to whitelist the VMs for access to the Puppet master, which will install the Machine Provider agent, which will create /b and connect to Swarming.

I believe that this issue is blocked on https://buganizer.corp.google.com/issues/67646308
which asks for a /22 block of IPs which will be used by machine provider to allocate the requested GCE instances.

I think that machine provider will be creating the instances, not GCE's instance groups mechanism.

GCE creates the instances from the managed instance group, not MP. MP only creates the template and the managed instance group. Anyways, yeah we need to wait for that bug to provide a block of IP addresses that GCE can automatically assign to new instances.

The following revision refers to this bug:
  https://chrome-internal.googlesource.com/infradata/config/+/ede5eb111e83efe7e515c85b0fa178b16200e4c9

commit ede5eb111e83efe7e515c85b0fa178b16200e4c9
Author: smut <smut@google.com>
Date: Thu Feb 15 01:26:23 2018