• Use newer ScyllaDB versions #3768 : 2 dependent PRs (#3752 , #3771 )
• Replication factor 3 on ScyllaDB #3769
• Ask confirmation to start benchmark #3744
• Use enum for BuildMode #3753
• Increase upper bound for inbox size metric #3742
• Make RocksDB config dependant of current hardware #3740
• All tasks wait before starting #3676 👈 (View in Graphite)
• Add max blocking threads option to client/server/proxy #3660 : 1 other dependent PR (#3637 )
• Print total tasks started on benchmark #3659
• More dashboard changes #3658
• Add --build-mode to net up --kubernetes #3741
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

Would an alternative approach be to prepare benchmarks on a single thread while running benchmark tasks on separate threads? That way the preparation part is not being starved.

@deuszx Depending on how many benchmark tasks are running, since they're blocking tasks, even if the prepare benchmarks is a single thread, that thread might still struggle to get scheduled to run in the same way :/
And doing the preparation single threaded will also make it slower

Merge activity

• Apr 11, 4:39 PM EDT: A user started a stack merge that includes this pull request via Graphite.
• Apr 11, 4:41 PM EDT: Graphite rebased this pull request as part of a merge.
• Apr 11, 4:42 PM EDT: A user merged this pull request with Graphite.