~/talk

Flink on Karmada: Building Resilient Data Pipelines on Multi-Cluster K8s

links:none on file
name:Wang Li
links:none on file
slides:none on file

Bloomberg runs 1,000 Flink jobs across multiple Kubernetes clusters and builds a Karmada extension that preserves Flink state for automated cross‑cluster failover. The implementation adds state preservation via JSONPath in the propagation policy to copy jobID into labels, and a Kyverno mutating webhook injects the initial savepoint path so a redeployed job resumes from its checkpoint. It reduces manual redeploys and eases maintenance windows, but requires tuning tolerationSeconds and has some cluster‑failover gaps still being addressed.

AI-written summary · about

  1. 0:00Flink on Karmada intro
  2. 1:07Speakers and agenda
  3. 2:09Bloomberg’s streaming platform
  4. 3:14Flink fundamentals and state
  5. 4:39Flink on K8s with Operator
  6. 6:04Multi-cluster Flink challenges
  7. 8:30Control plane for resilient pipelines
  8. 9:12Karmada for multi-cluster
  9. 10:14Karmada health tracking
  10. 11:15First Flink-Karmada integration
  11. 12:22Initial Karmada-Flink limits
  12. 13:03Flink job state for health
  13. 14:50Unhealthy state detection
  14. 15:54Automated failover in Karmada
  15. 16:36State preservation on failover
  16. 18:21Failover workflow with state
  17. 19:26Karmada benefits for Flink
  18. 20:48Generalizing failover solution
  19. 21:50Failover sensitivity config
  20. 22:31Limitations and improvements
  21. 23:53Future of failover and state
  22. 24:14Community and next steps
~/search
type to search 6,249 conferences, 138,987 speakers and 185,172 talks
~/help
cmdctrl k
search every conference, speaker and series
/
focus the filter search
h
home
c
conferences
s
speakers
e
series
t
light / dark — the status bar button does the same
?
this help
esc
close this, then clear every filter — or just the search, while you are typing in it

Ignored while you are typing. Pages without a filter rail have no search to focus.