> Maybe, you are right.... maybe not so. > Please forgive my liberty, of course there are alternatives like > redius and memcachedb which dedicate for persistent storage, of course > there are data-disaster-tolerate designing which > realized by libmemcached or proxy. but if we continue to work, there > are always things that can be improved. if we can, memcached restore > data after updating, without too much data lost, is a rather fantastic > thing in lots of memcached fans eyes....
I take this as a good sign; we've been releasing often enough that people are complaining about features again! I'll reiterate a point here, though: With redis/memcachedb, you typically have one pair of replicated instances. Clusters are much more rare. With memcached, it is designed to be a non-replicated cluster. It should affect you the same if your machine catches on fire or if you restart memcached. If you can't handle a restart, you can't handle failure. If you're not designing for failure, you suck. I'm not saying such a feature will never be added; but it won't be to solve the issue in the way you frame it. -Dormando
