Revert "Set temporary single CPU affinity before cgroup cpuset transition"#4283
Conversation
|
Now, this needs to be reverted before the 1.2 release, as otherwise we'll have to support this. |
|
I see the runtime- spec PR(opencontainers/runtime-spec#1253) has been approved, so it maybe merged recently? How about let the revert commit, bump runtime-spec commit and the new implementation commit completed in one PR. |
Makes sense (although upper-level runtime change is also needed for the whole thing to work again). Marking this one a draft for the time being. |
|
Because #3923 has not been in any release, and the related PR(opencontainers/runtime-spec#1253) has not been merged yet, so we need to merge this revert PR first if we want to draft 1.2.0-rc2. |
rata
left a comment
rata
left a comment
There was a problem hiding this comment.
LGTM
The runtime-spec PR this was implementing was closed (not merged), the new PR is still open. Reverting this seems the right thing to do here.
@kolyshkin do you agree on marking this ready for review and merging?
|
Yes, we need to merge it as we can't have this in a released version (and it blocks some other work). Once opencontainers/runtime-spec#1253 is merged I'll work on implementing it to replace this one. |
There's too much logic here figuring out which CPUs to use. Runc is a low level tool and is not supposed to be that "smart". What's worse, this logic is executed on every exec, making it slower. Some of the logic in (*setnsProcess).start is executed even if no annotation is set, thus making ALL execs slow. Also, this should be a property of a process, rather than annotation. The plan is to rework this. This reverts commit afc23e3. Signed-off-by: Kir Kolyshkin <kolyshkin@gmail.com>
|
We can prepare release 1.2.0-rc2 now. |
4 participants
This reverts #3923, for reasons outlined in #3923 (comment). A short version below.
There's too much logic here figuring out which CPUs to use. Runc is a low level tool and is not supposed to be that "smart". What's worse, this logic is executed on every exec, making it slower. Some of the logic in (*setnsProcess).start is executed even if no annotation is set, thus making ALL execs slow.
Also, this should be a property of a process, rather than annotation.
The plan is to rework this.
This reverts commit afc23e3.