Open argahsuknesib opened 6 months ago
I checked the code, and it is possible that the /advanceAndPushObservationPointerToTheEnd
still uses the original timestamp implementation with the naive algorithm.
Hmm, but if I just ignore the /advancePushPointerToTheEnd
and just use /startAutoReplay
this works fine. I think we can specify in README the steps to do both real-time and original-timestamp replay.
However, even when using the /startAutoReplay
the latest inbox location is not patched when a new container is created.
Hi Kush, I believe this is intended behaviour. /advancePushPointerToTheEnd
tries to submit all observations as fast as possible into the pod. If we were to adapt the timestamp, either all observation would be more or less at the same time, or we would be sending observations with a timestamp in the future towards the pod.
Hello @svrstich ,
Yes I agree with /advancesPushPointerToTheEnd
's intention. However, the inbox is still not updated for which I am working on in my fork.
Hello @svrstich , I have been testing the recent patch made by you for the real-time-replayer.
When replaying the observations one by one, the events get added to the latest inbox of the LDES in LDP, which works fine. However, after creation of the new container, the timestamps still use the original timestamp instead of the realtime ones. when triggering the
/advancePushPointerToTheEnd
.My config file looks like this,