Iván Valdés Castillo

Hey @ahrtr, I think there's value in enabling this with my suggestion due to the recent findings on #720. How would you suggest to proceed with this PR?

Hi @mrueg, it looks like it's failing because the benchmark job is timing out. We could either **just** increase the timeout or **also** split the benchmark into two jobs. I...

I have a branch with the functional code for this. However, `benchstat` needs at least six benchmark runs to generate a prediction with 95% confidence. Or at least two to...

Here's some discussion and progress regarding these last comments: I currently managed to do three runs using `ubuntu-latest-8-cores` runners and tunning the running parameters with `{ksize,vsize}=512`, which runs **under 30...

> It seems that you run the test in parallel. If you run them sequentially, it will be longer. [#750 (comment)](https://github.com/etcd-io/bbolt/pull/750#discussion_r1594281959) Yes, if we run them sequentially, it will take...

> I remembered that cncf provides 16 cores for us. You can try it with ubuntu-latest-16-cores. That's good to know. I'll try it out with them.

> > I remembered that cncf provides 16 cores for us. You can try it with ubuntu-latest-16-cores. > > That's good to know. I'll try it out with them. Uh,...

It takes about one hour to run the benchmarks sequentially: https://github.com/etcd-io/bbolt/actions/runs/9087376452 If this is acceptable, we can roll it out with 8-core runners. However, I still wonder if the Prow...

I'll clean up the code and mark the other PR ready for review.

@jonbretman, I think that issue may not be related. Seems like what you're using is not allocating a PTY (or TTY). In my case, the screen gets blank, as the...