This message was deleted.

Yes it assumes that the local segment cache is per historical.

@Adam Peck that means using local storage as deep storage is incompatible with more than one historical - unless every historical has its own storage?

local deep storage (

druid.storage.storageDirectory

) can be on a shared volume (like NFS or something like that) but the server caches (

druid.segmentCache.locations

) must be on dedicated per-server volumes

Ok, thanks for the explanation @Gian Merlino

np!

@Gian Merlino can I disable the cache? Does not really seem to make sense in my scenario (all historicals in the same cluster accessing the same shared volume)?

@Heiner Segment cache would be a local disk per historical and would be much much faster than NFS/Shared volume

👍 thanks

As druid is build for 🚤 each historical basically runs q's on the local cached segment which is super fast