Re: [libvirt] [PATCH 2/2] numa: Rewrite virNumaGetNodeCPUs() to query CPUs dynamically.

2015-04-30 Thread Andrea Bolognani
On Wed, 2015-04-29 at 12:58 +0100, Daniel P. Berrange wrote: It seems to me that every application that uses libnuma is going to potentially suffer from the same problem. As such I don't think it is a good use of effort to workaround it in libvirt - it should be fixed in libnuma itself, so

[libvirt] [PATCH 2/2] numa: Rewrite virNumaGetNodeCPUs() to query CPUs dynamically.

2015-04-29 Thread Andrea Bolognani
numa_node_to_cpus() uses an internal cache to store the mapping between NUMA nodes and CPUs (see Bug #1213835), which means long-running processes such as libvirtd will get stale data if CPU hotplugging is used during their lifetime. The function has been rewritten to collect the information

Re: [libvirt] [PATCH 2/2] numa: Rewrite virNumaGetNodeCPUs() to query CPUs dynamically.

2015-04-29 Thread Daniel P. Berrange
On Wed, Apr 29, 2015 at 09:53:31AM +0200, Andrea Bolognani wrote: numa_node_to_cpus() uses an internal cache to store the mapping between NUMA nodes and CPUs (see Bug #1213835), which means long-running processes such as libvirtd will get stale data if CPU hotplugging is used during their

Re: [libvirt] [PATCH 2/2] numa: Rewrite virNumaGetNodeCPUs() to query CPUs dynamically.

2015-04-29 Thread Martin Kletzander
On Wed, Apr 29, 2015 at 09:53:31AM +0200, Andrea Bolognani wrote: numa_node_to_cpus() uses an internal cache to store the mapping between NUMA nodes and CPUs (see Bug #1213835), which means long-running processes such as libvirtd will get stale data if CPU hotplugging is used during their