Closed tianyin closed 2 years ago
For the faults onto leader, can you inject the faults during the workload, so that the results account for the leader re-election?
@tianyin This support has been implemented and I've run the experiment using it. The results are exactly the same. 15% of the requests fail due to the leader crash while there is leader re-election. But the rest of the throughput and latencies are almost similar to previously reported results.
I can't access the figures in Q10 -- not sure why.
Updated the links, sorry.
In https://github.com/xlab-uiuc/slooo_internal/issues/37, you said "Not exactly sure what the reason might be. Have to investigate more." -- can you do this?
I think for this we needs deeper understanding of rethinks code, I'll try to look at the logs if I can find something.
I still really understand the choice of using 50MB -- why not smaller?
From my observation anything less than 50MB limit causes the leader to crash. The leader OOMs.
Sorry that I just had time to read over your answers again.
Yes, could you finish the following: