How to delete segments from co-ordinator

The coordinator rules page shows that only a Kill task can be sent for an interval.
However, the kill task deletes data from permanent storage too which i dont want.

How do we just ‘delete’ data from the coordinator (Delete Task, i guess) ?

Hi Prashant,

If i am getting it correctly, what you want to achieve is to be able to drop segments from historical nodes and still keep them in deep storage.

you can do so by configuring dropRule in coordinator console,

e.g If you configure a load rule for the current month followed by a drop rule for everything else, then the current month of data is kept, and all older data is dropped from historical nodes.

more details on rules can be found here -

so i want to keep last 120 days of data and drop the rest.

if i make a period rule like:


“type” : “dropByPeriod”,

“period” : “P120D”


will it instead drop the last 120 days or will it keep the last 120 days and drop everything before?

which rule do i use for my use case?

Prashant, you should use the coordinator console for configuring rules.

Please see:

Ordering of rules is important. Segments will follow the behavior of the first rule they match.


If you have a load rule for P120D, and a dropRule for dropForever

segments with intervals in the most recent 120 days will be loaded, everything outside that window matches the drop rule and is dropped

What does ‘tier’ in the rule mean?
In the console ui i see options as null and ‘_default_tier’. which one should i choose?

there is also a button to add additional tiers but i dont understand the tier thing at all.

Also in the json examples on the RuleConfiguration page, I see:


“type” : “loadForever”,

“tieredReplicants”: {

“hot”: 1,

“_default_tier” : 1



What does ‘hot’ mean here?

Hi Prashant:!searchin/druid-development/tiers

still a bit confused.

the reply here suggests:

" Load 2 replicas by period P2M into “r3”!searchin/druid-development/tiers/druid-development/WKyD03j7-NM/CybWXNwKacsJ

Why 2? In fact why even 1 replica.

I assume replica is replicated data right?

If i just want to have one copy of data loaded into my historicals, do i still need to specify 1 replica or do i not specify a replica at all?

One replica means only one copy will be loaded.