Details
-
Improvement
-
Status: Resolved
-
Major
-
Resolution: Duplicate
-
Impala 2.6.0, Impala 2.7.0, Impala 2.8.0
Description
During initial startup when first query is run on cluster or at end of --kerberos_reinit_interval, each impala daemon sends out N-1 number of TGS_REQ to KDC where N is the number of impala daemons. This amounts to N(N-1) TGS_REQ at once going from impala service to KDC.
This tends to overload KDC server.
The request here is to stagger these TGS_REQ requests so all of them do not hit KDC at the same time.
Attachments
Issue Links
- is duplicated by
-
IMPALA-3189 Address scalability issue with N^2 KDC requests on cluster startup
- Open