Wang Shilong
Wang Shilong
I changed codes a bit from master PR to kick off mpi requirement, master did not have problem because master included DAOS-7734 build: Wrap MPI code in unified interface change...
> Do we need to run this through daily_regression to possibly catch any tests that may fail with this change? Pushed a commit test for this.
> Do we need to run this through daily_regression to possibly catch any tests that may fail with this change? @jolivier23 I included daily_regression in the latest test tag and...
> can you please explain how this domain plays into the oclass selection? See dc_set_oclass() for example number of domains affects group size when object class is unknown, and we...
@daos-stack/daos-gatekeeper pylint and Jira report failure are CI issue and not related to this PR, other tests passed.
@daos-stack/daos-gatekeeper PR modified api minor version which touched Sconcript, there are some other pylint errors triggerd there, but it is not related to this PR
@daos-stack/daos-gatekeeper PR is ready for landing now.
> The atomics side looks good but I'm trying to work through the other implications. > > This will mean the "client" hash table (I'm not sure where that's used)...
> switch spinlock to read/write lock on the client side, and using ATOMIC for addref/decref looks fine to me. For "switch one global lock to per bucket lock", one point...
> switch spinlock to read/write lock on the client side, and using ATOMIC for addref/decref looks fine to me. For "switch one global lock to per bucket lock", one point...