2024-11-20T06:30:52.622Z

<Yingxin Cheng> The graph looks clear with socket and ht-sibling information available!

>  The real difference would most likely be once there are enough reactor thread and some might run on sibling cores.
Recalling my tests to evaluate how crimson messenger performance scales with cores, the pattern is linear when add more physical cores. After all the physical cores are involved, I started to include the sibling cores, and the performance stops to increase.

2024-11-20T17:17:31.280Z

<Matan Breizman> Hey all,
Seastore seems to be stable enough for our supported workloads for a while now (🥳) [1].
I think that T-release can be a good opportunity to replace our default object store to Seastore.
We could still test with both object stores for future comparisons.
@Yingxin Cheng, are you aware of any reasons not to go forward with this?
Thanks!

[1] <https://pulpito.ceph.com/teuthology-2024-11-09_20:56:02-crimson-rados-main-distro-crimson-smithi/>

2024-11-20T17:58:22.846Z

<Brett Niver> With Bluestore, we spent a year upstream randomly picking BS vs FS in teuthology, any thoughts to do something similar?

2024-11-20T18:04:14.567Z

<Matan Breizman> Yes, currently we run with both object stores for the entire suite

2024-11-20T18:04:38.389Z

<Matan Breizman> Yes, currently we run with both object stores for the entire suite (for each job)

2024-11-20T18:05:03.736Z

<Matan Breizman> Yes, currently we run on teu with both object stores for the entire test suite (for each job)