Hacker News new | past | comments | ask | show | jobs | submit login

Antirez responded to the original article citing that Sentinel, which was used for this test, was never designed to accomplish what was tested. So the Redis portion of that article is rather misleading: http://antirez.com/news/55



I am confident from theoretical argument [1] and experiment [2] that any system for failover between asynchronously replicated Redis nodes (in the absence of a strong external coordinator for requests) is subject to this kind of data loss and inconsistency.

If you read the post carefully, and Antirez's followup [3], I think you'll find Salvatore's claim is not that Redis is robust to partition, but that he doesn't think the risk is significant enough to address. I disagree: partitions are a real challenge for distributed systems at all scales. [4]

[1] http://aphyr.com/posts/287-asynchronous-replication-with-fai...

[2] http://aphyr.com/posts/283-call-me-maybe-redis

[3] http://antirez.com/news/57

[4] http://aphyr.com/posts/288-the-network-is-reliable




Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: