tmpfs - Using ChronicleMap in Apache Spark -


we want use chroniclemap in apache spark on workers store shared cache within vm. currently, using hazelcast share cache obviously, being able use off-heap memory free , avoiding tcp-ip (even on 127.0.0.1) things chroniclemap why want able use it.

the idea executors land within same physical machine share cache , cache persisted tmpfs stays valid till vm alive (we dont need values persisted across instance lifecycles) , worker joining start using it. ideally need lru eviction on keys stored , although chroniclemap not support it, given usage of off-heap memory, guess live restriction more concerning requirement recover crashes during access exclusively problem in our setup.

for example, lets have 5 executors running on same vm sharing cache , 1 of vms crashes during access corrupting cache. there no way other executors know this, have 1 of them recover while wait. if @ least guaranteed concurrent recovery or access during recovery or on corrupted cache throw exception, because then, failing on access can try acquiring file lock , begin recovery process , else accessing during time might receive exception indicating recovery ongoing , finally, cache fixed , things normal docs, looks behavior undefined leaves many doubts.

is possible chroniclemap working in our setup? has else had success getting work in apache spark, given these requirements? advice help.


Comments

Popular posts from this blog

node.js - Node js - Trying to send POST request, but it is not loading javascript content -

javascript - Replicate keyboard event with html button -

javascript - Web audio api 5.1 surround example not working in firefox -