sudo/rootwrap calls consuming large amounts of CPU

[James Page]

I've only seen this on zesty deployments (things appear to be OK on xenial); the net result of these slow calls is that network setup ops for instances take along time to complete; this results in failed instance launches (and as a result failed tempest tests).

I'm going to recheck without proposed so see if I see the same problems.

ProblemType: Bug
DistroRelease: Ubuntu 17.04
Package: neutron-openvswitch-agent 2:10.0.3-0ubuntu1
ProcVersionSignature: Ubuntu 4.10.0-37.41-generic 4.10.17
Uname: Linux 4.10.0-37-generic x86_64
ApportVersion: 2.20.4-0ubuntu4.5
Architecture: amd64
Date: Wed Nov 1 11:23:00 2017
Ec2AMI: ami-00000302
Ec2AMIManifest: FIXME
Ec2AvailabilityZone: nova
Ec2InstanceType: m1.medium
Ec2Kernel: unavailable
Ec2Ramdisk: unavailable
PackageArchitecture: all
ProcEnviron:
 TERM=screen-256color-bce
 PATH=(custom, no user)
 LANG=en_US.UTF-8
 SHELL=/bin/bash
SourcePackage: neutron
UpgradeStatus: No upgrade log present (probably fresh install)
modified.conffile..etc.neutron.plugins.ml2.openvswitch_agent.ini: [modified]
mtime.conffile..etc.neutron.plugins.ml2.openvswitch_agent.ini: 2017-11-01T10:27:38.248334

[James Page (james-page)]

I think I see the same issue without proposed so this appears to be a general zesty type issue.

[James Page (james-page)]

$ time sudo /usr/bin/python /usr/bin/neutron-rootwrap /etc/neutron/rootwrap.conf ip6tables-save

real 0m0.590s
user 0m0.496s
sys 0m0.064s

$ time sudo ip6tables-save

real 0m0.050s
user 0m0.004s
sys 0m0.016s

neutron-rootwrap is adding an overhead of 540ms - which when aggregated appears to be causing issues.

[James Page (james-page)]

Overhead of rootwrap on zesty does appear to be heavier; on bionic the overhead is 100ms or less.

As zesty is approaching EOL marking this as a Wont Fix.