Details
-
Improvement
-
Status: Open
-
Major
-
Resolution: Unresolved
-
1.4.3
-
None
-
None
Description
Currently if Accumulo is down, getConnector will sit indefinitely. This is designed to be tolerant in several processes which wait for Accumulo to come up into a stable state.
However, for writing client applications, it's horrendous. It provides no ability for the developer to provide feedback to the client. There should be some method to allow ServerClient.executeRaw() to eventually kick out of the infinite loop it currently has. If the developer wants the code to continue infinitely, they can either disregard the flag or wrap the getConnector() code in a loop themselves.
More generally it would be nice if the user could configure a timeout that would be honored by all client API calls.
Attachments
Issue Links
- is duplicated by
-
ACCUMULO-1278 ZooKeeperInstance sessionTimeout > 63 blocks forever
- Resolved
- is related to
-
ACCUMULO-2589 Create new client API
- Open
-
ACCUMULO-1726 Typed configuration framework
- Resolved
-
ACCUMULO-2241 Make retry in the face of network problems configurable
- Resolved
- relates to
-
ACCUMULO-1635 Support external configuration in client API
- Resolved
1.
|
Connector/ZooCache code enters infinite loop when Zookeeper connection lost. | Resolved | Unassigned |