On Wed, 4 Sep 2024 12:22:49 GMT, Matthias Baesken <mbaes...@openjdk.org> wrote:
> Hi , my guess is too that it is the second call. In case we want to go for a > loop, should we have a maximum number of retry operations like 10 or 100 ? > Another option might be to always make the buffer a little bigger (how much > can be discussed) and use the larger buffer to account for some increase in > the number of processes. It may have to loop indefinitely to allow for cases where something else is forking 100+ processes. In practise I assume it will only need to retry once or twice. You could combine the approaches so that you allocate a bit more than is required to take account of a few processes created in the window between the two calls to sysctl. So I think we should just do this. We still need to figure out the exception when sys call fails for some other reason. ------------- PR Comment: https://git.openjdk.org/jdk/pull/20839#issuecomment-2329251128
