albertogpz commented on a change in pull request #6052:
URL: https://github.com/apache/geode/pull/6052#discussion_r598469813
##########
File path:
geode-core/src/main/java/org/apache/geode/cache/wan/GatewaySender.java
##########
@@ -174,7 +174,32 @@
*/
int GET_TRANSACTION_EVENTS_FROM_QUEUE_RETRIES =
Integer.getInteger(GeodeGlossary.GEMFIRE_PREFIX +
"get-transaction-events-from-queue-retries",
- 10);
+ 2);
Review comment:
> But 2 is still too less. How about 3~5?
Given that prior to this change we had 10 retries and no timeout, and these
retries were done really fast (in less than 1 ms) I did not want to add a much
higher waiting time because it does not seem necessary if the events will
eventually reach the queue so I thought it would be better to have some small
waiting time (2ms) and have a smaller number of retries. These numbers seem
reasonable and have not caused flakyneess in the test cases.
If we put a higher waiting time and more retries as you are suggesting and,
for some reason, the configuration of the gateway senders or the traffic
received contains transactions but they cannot be completed (even with infinite
time), then you are going to be delaying replication for a long time: each
batch for waiting time*number of retries which could be a lot.
That's why I prefer to leave these values low and, if in some case it is
necessary, they can be changed with the system properties.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
