Embracing resiliency in GraphQL-Drop the cache-aside pattern and overcome the...
summary
GraphQL gateway at Joyn experiences catastrophic traffic when a shared Redis cache fails — with cache hit ratios of ~90% causing roughly a 10x spike and 95% around 20x — so the team replaces cache-aside by shifting heavy work into the write path, adding change data capture into a messaging platform, running a worker inside GraphQL to populate a local datastore, and rolling out via feature flags (silent sampling then 10%, 20%, 50% up to 100%). The approach works under controlled tests, trading storage/compute and eventual consistency for resiliency.
AI-written summary · about
chapters (12)
- 0:00GraphQL resiliency challenges
- 4:25Cache avalanche effect
- 9:21Cascading failures mitigation
- 11:50Drop cache-aside in GraphQL
- 14:44Write and read path explained
- 17:12EPG data access patterns
- 21:07GraphQL to local storage
- 22:55Safe incremental rollout
- 25:49When to use or avoid
- 27:56Write path trade-offs
- 29:45Separate data generation
- 30:05Summary and Q&A
related talks
24mA GraphQL Survival Kit
21mCaching your GraphQL data graph
52mState management in a GraphQL era
10mEvolution of Application Networking Patterns with GraphQL
37mAll in client GraphQL
36mThe State Management Rollercoaster
37mGraphQL and Microservices or How we learned to be big and small at the same time
22m8 Years of GraphQL: Unraveling the Trade-Offs