Hi, CacheStore semantic suppose you will warm-up a cache via calling cache.loadCache(). When you use read-through and forget load cache from back store then it is ok if no backups are updated on read operation.
Backups and replicated cache in case of read-through just add some tolerance to node-loss and allow you to not fallback to read-through after cache was warned-up. If you don't care about warm-up then most likely you are ok with read-through happens even if some other node already cached a value. AFAIK, read-through caches values localy on backup nodes to prevent further read from remote primary, that are expected in case of replicated cache. On Wed, Aug 22, 2018 at 2:20 PM Prasad Bhalerao < [email protected]> wrote: > I am just wondering would you really need backup partitions in case of > replicated mode? The whole data is replicated, so I think each node itself > is kind of backup for other. > > > On Wed, Aug 22, 2018, 2:41 PM yonggu.lee <[email protected]> wrote: > >> I have a read-through cache backed by a hbase table. I this case, is >> there no >> difference between Replicated and Partitioned mode? >> >> When I tested, although I set up a Replicated mode, the cache only had >> Primary, and not Backup part (in web console, monitoring tab) >> >> >> >> >> >> -- >> Sent from: http://apache-ignite-users.70518.x6.nabble.com/ >> > -- Best regards, Andrey V. Mashenkov
