Migrating 15k Redis servers to Valkey
summary
Migration replaces a Redis VM by spawning a new Valkey VM, issuing replicaof to stream the RDB, pausing clients, flipping replication, and retiring the old node so migration occurs with minimal downtime. Platform exposes a one-click console upgrade and also runs automatic moves in maintenance windows. Outcome: about 10,760 automatic and 3,800 customer‑triggered migrations (~14,560 total). Hit points: a Valkey ACL load crash was patched upstream, Redis 7.4/8 RDB incompatibilities remain, and Terraform users need a state remove/import workaround.
AI-written summary · about
chapters (10)
- 0:00Introduction and motivation
- 2:20Migration challenges and goals
- 3:03Technical migration process
- 5:54Customer experience and one-click solution
- 6:40Automatic migration during maintenance
- 7:23Terraform state mismatch issue
- 8:52Bugs and complications encountered
- 9:13Customer and partner buy-in
- 9:55Migration results and numbers
- 10:18Q&A session
related talks
33mDiscover Valkey: The Path to Now
41mValkey 8 — Release of a Redis Fork Focused on Performance
44mValkey: Latency in Cloud Environments
48mValkey clusters unboxed
27mAn Introduction to Messaging in Valkey
25mMigrating Datastores
10mThe unreasonable effectiveness of key value databases
50mImmutable Values in the Big Wide World