Statistic cache elements should overflow to disk. They should be wiped upon app startup, and last as long as the server is running. Ensure that there is a way of clearing the stats cache individually. Cache generation is currently happening on-demand.
Maybe we should make the cache eternal, as long as there is a way of clearing the cache. Testing: - Performance (lacking base line for performance testing) - Some level of stats regression testing
We need to make sure the cache isn't flushed to disk synchronously on every single update, or translation updates will probably become too expensive. Some options which might help: 1. Enable async writes: http://ehcache.org/documentation/2.5/apis/write-through-caching 2. Only flush to disk: http://ehcache.org/documentation/2.5/faq#how-do-i-get-a-memory-only-cache-to-persist-to-disk-between-vm-restarts
Pull request: https://github.com/zanata/zanata-server/pull/460
A few more issues: * Property ehcache.disk.store.dir should be given. * Installation guide should be updated. * zanata-setup.sh should be updated. * The cache will not be wiped out during server restart. * The cache will updated on events on * Translator translates a new string. * Translator clear a string. * Translator change state from FUZZY -> Translated * Review change state from Translated -> Rejected.
VERIFIED with Zanata 3.5.0-SNAPSHOT (git-server-3.4.1-52-g47cdeb5)