Closed juvenzhu closed 3 years ago
@khaf Can you help us take a look! Thanks
Of course. I need a bit of time to be able to reproduce the issue. I'll have to ping our server team to see if this is in fact unexpected behavior as well.
Thanks @khaf. This issue is rather important to us, it's been impacting our service SLA everytime there's a server event causing migrations. Do you prefer us to create a corresponding ticket on the support time to expedite the process?
Creating a support ticket would help, since I think we'll need help from other teams to diagnose the issue. At this point while I have some idea of why this can happen, I'm not completely positive if that's the case and what can be done to mitigate the problem.
Can you also report the values you use for the policy?
Thanks
I have opened a support ticket. This is the policy we are using:
The values for the batchPolicy are: -ConcurrentNodes: 0 -AllowInline: true -AllowPartialResults: true -basePolicy: - TotalTimeout: 500ms - SocketTimeout:250ms - MaxRetries: 3 - SleepBetweenRetries: 100ms - SleepMultiplier: 2 - SendKey: true
Many errors returned to the client also generate a warning message in the server logs. Could you check if such a warning appears and provide the message here?
@khaf @kportertx this turned out to be a server issue and the new release fixed this.
Thanks a lot for helping.
No problem, we were working on the issue behind the scene and helped the server team in debugging the issue. Feel free to open new issues any time.
We are using BatchGetComplex API to perform batchGet. Recently, we found when there is migration going on on the server side, the BatchGetComplex API will return error (PARAMETER_ERROR). The error will go away once the migration has been finished.
The client version we are using is 2.12.0 and 3.0.3