Race condition : a member can be both declared as succeeded and failed for a
query when shutting down
-----------------------------------------------------------------------------------------------------
Key: HEDERA-19
URL: https://forge.continuent.org/jira/browse/HEDERA-19
Project: Hedera
Type: Bug
Versions: Hedera 1.5
Reporter: Stephane Giron
Assigned to: Stephane Giron
Fix For: Hedera 1.5.4
When using WAIT_ALL, under load and when a controller is shutting down, it is
possible that it sends a success notification for a query to the other
controller, but this last one will in parallel declare that the first is
failing.
As a result, the leaving controller may be added to the failed members group
and to the succeeded members group at the same time (for queries that are
notified). This will prevent the remaining controller from accepting later
responses (dropped as the number of "presumed" reponses is equal to the number
of expected responses), and one of these responses may be it self response.
As it is dropped, the client will receive an exception : request failed on all
controllers (even if it succeeded on the remaining one).
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
https://forge.continuent.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see:
http://www.atlassian.com/software/jira
_______________________________________________
Hedera mailing list
[email protected]
https://forge.continuent.org/mailman/listinfo/hedera
