Ok that makes some sense. What kind of traffic is passed between the hosts in this rebalancing?
On Thursday, January 15, 2015 at 1:19:15 PM UTC+11, LesMikesell wrote: > > On Wed, Jan 14, 2015 at 8:01 PM, gunna <[email protected] <javascript:>> > wrote: > > I'm doing a large amount of reading on the subject but had a question > about > > something. Firstly I believe while you can have an array or memcached > > server they do not replicate. If they don't replicate then what is the > > reason you would have an array? Is it purely for redundancy? What > happens > > if a query is run on server 1 and loaded into memcahed but on server 2 > that > > query is not as current as server 1. If server 1 fails and you go to > server > > 2 the query will be in cache but be stale? > > > > You use multiple servers so you still have some running in case one or > a few fail. You can either use a client hashing strategy that simply > fails and pulls from the persistent database for the percentage of > servers that are down, or you can use one that rebalances across the > remaining servers. When you cache something you set how long that > value is allowed to be used. Even in the rare case of rebalancing and > servers going in and out of the cluster such that a client queries a > server that does not have the latest value, it still won't return > something older than the time you set as ok to reuse it. > > -- > Les Mikesell > [email protected] <javascript:> > -- --- You received this message because you are subscribed to the Google Groups "memcached" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
