Closed Bug 1311681 Opened 9 years ago Closed 8 years ago

Use spot instances for scheduled clusters

Categories

(Data Platform and Tools Graveyard :: Telemetry Analysis Service (ATMO), defect, P2)

Product:
Data Platform and Tools Graveyard
Component:
Telemetry Analysis Service (ATMO)
Type:
defect

Tracking

(Not tracked)

Status:
RESOLVED WONTFIX

People

(Reporter: rvitillo, Unassigned)

References

Details

User Story

Let's use spot instances for our scheduled jobs similarly to how we have done for ad-hoc clusters in atmo v2.
No description provided.
Blocks: 1269754
Points: --- → 1
Priority: -- → P2
Priority: P2 → --
OPs would like to resolve this quarter (budget concerns)
Points: 1 → 2
Priority: -- → P2
Component: Metrics: Pipeline → Telemetry Analysis Service (ATMO)
Product: Cloud Services → Data Platform and Tools
I seem to recall that we tried this and ended up hitting issues w jobs getting backed up when spot instances weren't available? Is this true? Anyway, I've opened a github issue (https://github.com/mozilla/telemetry-analysis-service/issues/555) for further tracking.
My earlier comment was correct... we tried this for a while, but had issues with the scheduled jobs not running at times when spot instances weren't available. It might be possible to dynamically decide whether or not to use a spot instance based on assessing the current AWS load, but this is a much larger project and we've determined that (for now) the budget savings aren't worth the effort.
Status: NEW → RESOLVED
Closed: 8 years ago
Resolution: --- → WONTFIX
Product: Data Platform and Tools → Data Platform and Tools Graveyard
You need to log in before you can comment on or make changes to this bug.