Unclear real-time use case

I have a realtime spec:

  • hour segment granularity
  • 1 minute intermediate persist
  • 5 minutes window period.
    18:06 - push 1000 messages with 18:05 in timestamp. Can see them in druid.

… waiting the realtime to persist intermediate data …

… shut the real-time down …

18:15 - start real-time node

Even if window period is 5 minutes, after the start i expected to see the 18:05 o’clock messages because they had been persisted to disk before i shut down the node. They were not present. Probably i misunderstood the settings meaning. Any idea?

Thanks, Vadim.

The question would be: if the data is persisted, server goes down and back, will the persisted data be available if it’s out of windowPeriod ?

Hi Vadim, yes, Druid tries to reload any data it has persisted and will try to hand things off periodically.