3 ms·
Every system fail prompts people to exclaim "why aren't there safeguards?". Every time. Well guess what, if we try to do new stuff, we will run into new problem
by namaria 4y ago
Every system fail prompts people to exclaim "why aren't there safeguards?". Every time. Well guess what, if we try to do new stuff, we will run into new problems.
- wouldbecouldbe 4y agoThere is nothing new about using redis for cache, or returning a list for a user.
- namaria 4y agoAre you trying to say cache invalidation in a distributed system is a trivial problem?
- chatmasta 4y agoI'm not disagreeing with you, and I'm not the commenter you're replying to, but it's worth noting that cache leakage and cache invalidation are two different problems.
- namaria 4y agoYou're right. Thanks for pointing that out. My original point still stands, distributed systems are hard and people demanding zero failures are setting an impossible standard.
- oulu2006 4y agoIt's non-trivial but it's also not that hard, there are well known strategies for achieving it; especially if you relax guarantees and only promise eventual consistency then it becomes fairly trivial - we do this for example and have little problems with it.
- skywhopper 4y agoThis wasn’t a cache invalidation problem. It was a cache corruption error.
- wouldbecouldbe 4y agoIm saying there is nothing new about it.