Stephen Hemminger <step...@networkplumber.org> writes: > On Tue, 5 Jul 2016 16:51:18 +0200 > Phil Sutter <p...@nwl.cc> wrote: > >> Stress-testing OpenStack Neutron revealed poor performance of 'ip netns' >> when dealing with a high amount of namespaces. The cause of this lies in >> the combination of how iproute2 mounts NETNS_RUN_DIR and the netns files >> therein and the fact that systemd makes all mount points of the system >> shared. >> >> Changes since v1: >> - Added Suggested-by tag to patches. >> >> Phil Sutter (2): >> ipnetns: Move NETNS_RUN_DIR into it's own propagation group >> ipnetns: Make netns mount points private >> >> ip/ipnetns.c | 12 +++++++++++- >> 1 file changed, 11 insertions(+), 1 deletion(-) >> > > I want an ack for Eric on this, it seems a little risky
If I have read the tea leaves properly the first patch is fighting a system configuration of "mount --rbind /run /var/run" which seems inappropriate. A symlink is likely a better fix. The effect of the second patch is probably better had by performing the optimization in the kernel, so everyone gets the benefit. I am persuadable to change this behavior. But if all we are doing is optimization and not intentionally changing the semantics there appear to be better places for the optimizations to happen. Eric