Restarting Druid realtime without data loss

Hi Druid Experts,
I have Druid realtime (v. 0.8) that I want to restart without loss of data. The current instance started to write another daily segment to
/tmp/1468407735578-0/{dataSource}/2016-11-17T00:00:00.000Z_2016-11-18T00:00:00.000ZWhen I kill the realtime and start it over it starts to write to :
/tmp/1468411111111-0/{dataSource}/2016-11-17T00:00:00.000Z_2016-11-18T00:00:00.000ZIs there any way to reuse, inject, embarce partially constructed segment in /tmp/1468407735578-0 by new Druid realtime instance?
Or in another words, what is the procedure to restart realtime without data loss?
I have read that it should be possible but wheneer I restart realtime the nonfinished segment is constructed from scratch with the cost of data loss.
Maybe something with configs?

Thanks in advance for the help.
Best,
Pawel

Or another scenario.
I have a 3 not-yet-pushed completed segments in /tmp corresponding to just stopped realtime. Is it possible that I copy them in the respective /tmp of just started realtime and it will push it to deep-storage.
And general question. What is the trigger to push segments from /tmp , just presence of them?

Best,
Pawel

Hey Pawel,

Have you read this - http://druid.io/docs/latest/operations/rolling-updates.html

The doc describes two ways to redeploy middle managers without incurring any data loss. “Rolling restart (restore-based)” is I think the easiest (and best IMO) and is available since 0.8.3

  • Parag

Or another scenario.
I have a 3 not-yet-pushed completed segments in /tmp corresponding to just stopped realtime. Is it possible that I copy them in the respective /tmp of just started realtime and it will push it to deep-storage.
And general question. What is the trigger to push segments from /tmp , just presence of them?

Best,
Pawel