Lossless Upgrade of BOSH Deployment
summary
Demonstrates a lossless upgrade for BOSH deployments by checkpointing live process memory and restoring it on a recreated VM, preserving caches, pipes, domain sockets and TCP sequence state. Uses CRIU-style checkpoint/restore integrated into BOSH drain scripts, stores binary images on distributed storage, and restores state so applications see minimal disruption; demo moves ~10,000 RabbitMQ messages through a recreate/restore cycle and returns the full message set and cached state.
AI-written summary · about
chapters (11)
- 0:00Introduction to lossless upgrade
- 1:29Problem with stateful upgrades
- 4:43Introducing CRIU for stateful migration
- 9:51How CRIU works under the hood
- 11:37Stateful migration in BOSH deployments
- 15:08Live demo of CRIU in action
- 21:04Q&A on CRIU and migration
- 26:54Optimizing state transfer for databases
- 29:03CRIU's independence from services
- 30:36Alternatives to CRIU and future work
- 31:24Closing remarks and contact info
related talks
30mBurning Down the House: How to Plan for and Deal with Disaster Recovery in CF
25mBOSH: The Simple Solution for Complex Systems
28mBurning Down The House: How to Plan for and Deal with Disaster Recovery in CF
30mStreamlining the BOSH Experience
08mYou CAN Teach an Old BOSH New Tricks!
19mBOSH – Restore from an Availability Zone Outage
28mHow to Deploy CF and Other Bosh Releases on Bare-Metal Environment
25mBOSH Bootstrapping