Updated numad packages that fix several bugs are now available for Red Hat Enterprise Linux 6.
The numad packages provide a daemon for NUMA (Non-Uniform Memory Architecture) systems, monitors NUMA characteristics and manages placement of processes and memory to minimize memory latency. The packages also provide an interface that can be used to query the numad daemon for the best manual placement of an application.
- Previously, running the numad daemon on a system executing a process with very large resident memory, such as a Windows Server 2012 guest, could cause memory swapping. As a consequence, significant latencies under some circumstances occurred on the system, which could in turn lead to other processes, such as qemu-kvm, becoming unresponsive. With this update, numad no longer causes memory swapping in the above scenario, and the consequent latencies and hangs no longer occur.
- Prior to this update, when a process bound to a set of NUMA nodes depleted the system memory, the system started memory swapping instead of using other NUMA nodes for memory allocation. Consequently, the system experienced significant latencies or became unresponsive. With this update, numad unbinds memory nodes after moving the process memory, which allows for other nodes to be used for memory allocation, and thus prevents the described latencies and hangs.
- Previously, the numad daemon ignored existing control groups when localizing QEMU threads, and as a consequence, it incorrectly consolidated all running threads into a single control group. This update introduces numad support for multiple control groups, and numad therefore no longer moves QEMU threads into undesired control groups.
Users of numad are advised to upgrade to these updated packages, which fix these bugs.