Details
-
Improvement
-
Status: Open
-
Major
-
Resolution: Unresolved
-
None
-
None
Description
Unhealthy nodes are usual In large clusters.
There are some unhealthy cases as follows:
- there is no enough storage capacity.
- frequent task errors or failures in certain nodes
- other reasons reported by workers
TajoMaster and ResourceTracker should keep those nodes in the black list, and than resource allocations of TesourceTracker should consider the black list.
Attachments
Issue Links
- is related to
-
TAJO-1214 (Umbrella) Improve task and node failure handling
- Open