Ask Your Question

darragh-oreilly's profile - activity

2017-12-28 03:27:17 -0500 received badge  Necromancer (source)
2017-05-27 13:18:25 -0500 received badge  Nice Answer (source)
2017-01-10 06:04:42 -0500 answered a question DPDK: Inter VM communication of iperf3 TCP throughput is very low on same host compare to non DPDK throughput

I see the about the same numbers. Ethtool -k on the non-DPDK VM nics show a lot of offloads that are not available on the DPDK VM nics. One of these is tcp segmentation offload, and with this I see packets of avg size ~60,000 bytes between VMs on the same host. Turn tso off (ethtool -K ens3 tso off), the packet size is ~1500 and the rate drops from about 17 to 2 Gbps.

It seems DPDK/vhostuser does not provide this offload, and the packet size is ~1500. But the packets per second is better.

2016-10-18 12:06:57 -0500 commented question How to isolate tenant networks connected to a shared router?

maybe fwaas with provider rules that only allow north-south traffic. Or maybe the new address-scopes feature can be used.

2016-10-18 10:39:57 -0500 commented question How how does icmp packet travel across two compute nodes without br-int and br-tun running?

that question was already asked and answered

2016-10-17 13:43:35 -0500 commented question How how does icmp packet travel across two compute nodes without br-int and br-tun running?

the any with tcpdump does not always work for sone reason. Anyway you are filtering for icmp, but the icmp packets will be encapsulated in udp if you are using vxlan - so you need to filter for udp. If using gre you need to tcpdump -ni ethX proto gre

2016-10-17 13:29:28 -0500 commented question Devstack and OpenvSwitch v2.3+

A grep -r openvswitch devstack/ gives this devstack/lib/neutron_plugins/ovs_base: install_package $(get_packages "openvswitch") - I would try commenting that out

2016-10-12 12:33:23 -0500 answered a question br-tun and br-int are not UP but still VMs are pinging

This is normal. Ifconfig is showing interfaces br-int and br-tun - not bridges. These are OVS ports that get created when the bridges were created. But they are not used by the neutron-openvswitch-agent, so it is best that you leave them down. They are trunk ports, but tcpdumping them won't show much except broadcasts and floods, because the bridge learns where the macs are - ovs-appctl fdb/show br-int

Note OVS is a little different than Linux bridges in this regard. With linux bridge, the interface with the same name as the bridge needs to be up, or the bridge will not forward.

2016-09-19 03:13:51 -0500 answered a question how do I disable a nova compute node?

To remove a host from the scheduler, this works Kilo and KVM:

nova service-disable <host> nova-compute

Use nova service-list or nova hypervisor-list to see the status.

2016-09-18 02:16:06 -0500 commented question Help me install devstack in linux 16.04

why not use the steps here ?

2016-09-17 11:18:13 -0500 commented question High amount of swap space being used

forgot about rpc_workers - that would be other 32 processes.

2016-09-17 06:08:24 -0500 commented question High amount of swap space being used

the screenshot shows multiple neutron-server processes - probably 32 as 1 per cpu is the default. You should reduce this - I think it is api_workers in one of the neutron config files. The same goes for other services like nova-api etc. That should reduce you memory usage.

2016-09-17 02:27:30 -0500 commented answer odd message - Too Many External Networks

unfortunately he didn't bother to say what he was trying to do. Maybe create a floating ip.

2016-09-16 15:44:48 -0500 commented question Afetr a successful Live migration, instance is still in migration status

I'm looking at a similiar problem in with kilo. But there is a ERROR and stacktrace in the destination's compute log. It failed to PUT on the neutron port. On the neutron server there was a DBdeadlock so I need to troubleshoot the percona db next.

2016-09-16 14:21:23 -0500 commented question Afetr a successful Live migration, instance is still in migration status

Are there any errors in the nova-compute logs from the time of the migration?

2016-09-16 10:00:49 -0500 commented question Afetr a successful Live migration, instance is still in migration status

can you try nova reset-state --active <server>

2016-09-15 12:44:10 -0500 commented question Launchin instance on version 2.3.0 with Linuxbridging failed

don't see your ml2_conf.ini Also check the linuxbridge agent logs on the compute

2016-09-15 12:34:06 -0500 commented question Afetr a successful Live migration, instance is still in migration status

what version?

2016-09-14 04:19:27 -0500 commented question MTU size of VXLAN Tunnel in OpenStack Mitaka

where did you find this patch_mtu option? I don't see it in the source code.

2016-09-12 04:09:29 -0500 commented answer python openstacksdk, how to authorite?
2016-09-12 04:06:22 -0500 edited question python openstacksdk, how to authorite?

hey, i trying to connect via the openstacksdk python lib, following this guide:

here is my code:

from openstack import connection

if __name__ == '__main__':

    auth_args = {
        'auth_url': '',
        'project_name': 'admin',
        'username': 'admin',
        'password': 'PASSWORD',
    conn = connection.Connection(**auth_args)

but i only get this exception:

openstack.exceptions.HttpException: HttpException: Expecting to find domain in project - the server could not comply with the request since it is either malformed or otherwise incorrect. The client is assumed to be in error.

the command line client is working fine, with the very same credentials.

this is a installation based on this documentation:

what am i doing wrong?

2016-08-11 01:41:06 -0500 commented answer Bridge network interface for VM networks not working in openstack ubuntu

the question seems to be about nova-network and not neutron

2016-08-11 01:31:50 -0500 edited question BGP Configuration - what is the agent node?

I am trying to set up openstack environment in a LAB and am playing wanted to deploy BGP routing as part Mitaka release, as the configuration guide it says to change some configuration on Agent node, but i am confused as to what is Agent node, as per the installation for this release is concerned, it only talks about compute and controller node but not Agent node, is this a separate machine with SDN controller?

2016-08-11 01:30:40 -0500 answered a question BGP Configuration - what is the agent node?


Agents can reside on hosts with or without other Networking service agents.

Many people run network nodes that are dedicated to network services like l3 routing, dhcp etc.

In the doc example, it is shown running on the controller node

2016-08-11 01:19:05 -0500 answered a question Any scripts that enable VirtualMachine to restart all the Openstack services would help

Git-revert this commit in your local devstack repo to get back. It doesn't restart all services (like some cinder service), but it does all I need.

2016-08-09 12:01:51 -0500 commented question Kilo port_security_enabled=false for network

probably this one - but not backported to kilo -

2016-08-04 08:21:38 -0500 received badge  Good Answer (source)
2016-07-29 10:09:42 -0500 commented question DVR, Floating IPs are not working. Failed sending gratuitous ARP

this problem is being worked on here

2016-07-29 06:30:07 -0500 received badge  Necromancer (source)
2016-07-28 14:40:06 -0500 received badge  Nice Answer (source)
2016-07-28 13:55:11 -0500 answered a question Can we have multiple subnets on a neutron network?

Yes you can. The common use case is dual-stack IPv4+IPv6.

2016-07-28 13:03:31 -0500 commented question DVR, Floating IPs are not working. Failed sending gratuitous ARP

What distro/verions? What happens when you do ip netns exec fip-4f2774d1-dfb8-4833-8374-806e1fc40827 arping -A -I fg-86481da8-4c -c 3 -w 4.5

2016-07-20 12:45:39 -0500 commented answer should bridges(br-int,tun,ex) UP ?

this is incorrect

2016-07-19 03:39:23 -0500 commented question Running a router inside an OpenStack VM
2016-07-13 10:54:36 -0500 commented question Do I need a router, can't I do Bridging.

you can use neutron vlan or flat provider networks

2016-07-13 09:47:09 -0500 commented answer Unable to delete port

maybe your version does not have them

2016-07-13 09:36:04 -0500 commented question Linux bridge RTNETLINK exists

I don't know - maybe something to do with existing routes. Maybe you could try configuring the bridge manually and use bridge_mappings instead.

2016-07-08 09:03:32 -0500 commented question Linux bridge RTNETLINK exists

maybe you are not running at DEBUG level. Do you see any lines with DEBUG in them?

2016-07-08 08:06:16 -0500 commented question Linux bridge RTNETLINK exists

yep - that's what I mean. Unfortunately the exception does not include the actual command that failed. Can you look up the log before the exception to find it. Should begin with "Running command"

2016-07-07 13:39:36 -0500 commented question Linux bridge RTNETLINK exists

can you find a log message with "Running command" just before the error. You seem to be using a flat network, and the interface in the interface_mappings has an IP. I would always use unnumbered interfaces for these.

2016-06-25 03:36:43 -0500 commented question routing between LAN and WAN interface is not working openstack VNF instance

execute the sysctl command on the node running nova-compute

2016-06-17 04:54:31 -0500 commented question Set router gateway for non-external networks

maybe if NEXT_HOP_IP is reachable from some VLAN provider net that is attached as an internal interface to the router, then neutron router-update $ROUTER_ID --routes type=dict list=true destination=,nexthop=$NEXT_HOP_IP

2016-06-17 04:47:07 -0500 commented answer How Openstack is adding tap in OVS when mechanism driver is ODL.

donno - I'm not vey familiar with that part of the code

2016-06-15 09:54:54 -0500 answered a question How Openstack is adding tap in OVS when mechanism driver is ODL.

nova-compute creates the tap. It communicates with the neutron api to create the logical port, and uses the details it gets back.