Closed Bug 1078972 Opened 10 years ago Closed 7 years ago

Stand up releng infra in us-west-1

Categories

(Release Engineering :: General, defect)

x86_64
Linux
defect
Not set
normal

Tracking

(Not tracked)

RESOLVED WONTFIX

People

(Reporter: rail, Unassigned)

Details

(Whiteboard: [kanban:engops:https://mozilla.kanbanize.com/ctrl_board/6/2501] )

Looks like spot prices in us-west-1 are comparable with prices in other US regions. Average prices for September for instance types we use are below.

Standing up would require setting up VPC, subnets, security groups, puppet masters, buildbot masters, etc.

Sorted by average price:

===== m1.medium =====
us-west-1 0.00808717948718
us-east-1 0.009308033241
us-west-2 0.0117338296112
===== m3.2xlarge =====
us-west-2 0.106082876165
us-east-1 0.243992095536
us-west-1 0.249989967638
===== c3.xlarge =====
us-west-2 0.0372971819927
us-west-1 0.0396091441112
us-east-1 0.212801941748
===== m3.medium =====
us-west-1 0.0280343497758
us-east-1 0.119250991501
us-west-2 0.1417
===== r3.xlarge =====
us-west-1 0.0321209150327
us-east-1 0.0327683232811
us-west-2 0.0363334153984
===== r3.2xlarge =====
us-west-1 0.0654128983308
us-west-2 0.0767122617854
us-east-1 0.0792172057133
===== c3.2xlarge =====
us-west-1 0.08246069375
us-west-2 0.0857382912311
us-east-1 0.0864698258851
===== m3.xlarge =====
us-west-2 0.077998854233
us-east-1 0.147909088359
us-west-1 0.266855487805
Whiteboard: [kanban:engops:https://mozilla.kanbanize.com/ctrl_board/6/2501]
We already have a VPC to us-west-1 (N. California) and the prices for g2.2xlarge (ie g-w732) are consistently less than our $0.70 bid. Prices are also a lot less spiky than us-east-1 & us-west-2, so there may be less demand in us-west-1 and it may be worth working on this bug.

Initial investigations
* check prices to see if it'd be useful to enable more than g-w732 in us-west-1
* check progress of move to taskcluster, no point doing this work if won't be used for long

We'd need
* a puppet master
* a pair of log aggregator instances
* a proxxy instance ? Need to verify if we're still using that
* some buildbot masters
* some subnets, route tables etc
* new workers in slavealloc and buildbot config

There is already a VPC, nagios, and smokeping/canary there.

catlee, what do you think of this ?
Flags: needinfo?(catlee)
I think we could probably get away without a puppet master, initially at least.
We are still using proxxy, so will need one stood up.

I think we should definitely try it out. There are other new regions as well that have come online, so this could be good practice :)
Flags: needinfo?(catlee)
is this bug still required? If so, I'd like to reassign to the buildduty folks.
Flags: needinfo?(rail)
I'm not quite sure now. With more and more stuff moving to TC, we can probably WONTFIX this bug.

Chris, what you say?
Flags: needinfo?(rail) → needinfo?(catlee)
Agreed.
Status: NEW → RESOLVED
Closed: 7 years ago
Flags: needinfo?(catlee)
Resolution: --- → WONTFIX
Component: General Automation → General
You need to log in before you can comment on or make changes to this bug.