RFR: 8334357: Use NonInterleavingLogStream for report_metadata_oome

Axel Boldt-Christmas aboldtch at openjdk.org
Mon Jun 24 12:15:11 UTC 2024


On Mon, 24 Jun 2024 08:18:15 GMT, Johan Sjölen <jsjolen at openjdk.org> wrote:

> The NonInterleavingLogStream works by buffering its output before flushing to output, so is there a risk that the logging itself will fail since this is an OOME?

I tried to reason about this as well. The common reason to get here should be that we are hitting JVM limits on metaspace, not because of native memory exhaustion. It seems unlikely that we would even manage to get to this point if the system resources are so constrained that a buffer cannot be allocated. (We run a whole GC cycle before throwing the OOME).

-------------

PR Comment: https://git.openjdk.org/jdk/pull/19770#issuecomment-2186434000


More information about the hotspot-runtime-dev mailing list