Closed Bug 974727 Opened 11 years ago Closed 8 years ago

Fallback to slower(or pv) nodes past a certain watermark: eg m3.large, m1.large, m1.xlarge (or m1.small for tests)..or faster but older c1.xlarge nodes which are cheap on spot

Categories

(Infrastructure & Operations Graveyard :: CIDuty, task)

Product:
Infrastructure & Operations Graveyard
Component:
CIDuty
Platform:
x86_64
Linux
Type:
task
Priority:
Not set
Severity:
normal

Tracking

(Not tracked)

Status:
RESOLVED INCOMPLETE

People

(Reporter: taras.mozilla, Unassigned)

References

Details

We should prefer running on faster nodes when it makes sense. However it makes sense to use slower nodes when: * There is low spot node availability * We could place limits on the number of fancy(ie expensive) instances and spawn the rest as cheap(slower) instances. This would be handy at peak times (eg above our stable load) to still make progress at handling the queue without blowing our budget * Ondemand prices for these aren't as attractive, but it may work out that running jobs on these slower nodes may still be more cost-effective than on faster nodes...in this case it might make sense to switch our ondemand nodes to these slower ones with the expectation that these are nodes of last resort to make up for dips in spot capacity
Summary: Fallback to slower nodes past a certain watermark: eg m3.large, m1.large, m1.xlarge (or m1.small for tests) → Fallback to slower(or pv) nodes past a certain watermark: eg m3.large, m1.large, m1.xlarge (or m1.small for tests)..or faster but older c1.xlarge nodes which are cheap on spot
Status: NEW → RESOLVED
Closed: 8 years ago
Resolution: --- → INCOMPLETE
Component: Platform Support → Buildduty
Product: Release Engineering → Infrastructure & Operations
Product: Infrastructure & Operations → Infrastructure & Operations Graveyard
You need to log in before you can comment on or make changes to this bug.