Miscelanious Linux and OpenStack Notes

Linux Networking Notes

Handy options for tcpdump:

tcpdump -nlei $interface

Check on the status of HAProxy:

Gather a variety of infornation from HAProxy:

echo "show info;show stat;show table" | socat /var/lib/haproxy/stats stdio

Cleanly print statistics for a few columns:

echo "show stat" | socat /var/lib/haproxy/stats stdio | awk 'BEGIN{FS=","} {sub(/^#\ */, "", $0); print $1,$2,$9,$10}' | column -t

List out all available columns for easy reference:

echo "show stat" | socat /var/lib/haproxy/stats stdio | awk 'BEGIN {FS=","} NR == 1 {sub(/^#\ */, "", $0); for (i=1; i < NF; i++) { print "$"i" = "$i }}'

Additional details for each column can be found in the HAProxy documentation http://www.haproxy.org/download/1.5/doc/configuration.txt

Pretty ip link details and statistics output:

ip -s -d -o link | sed G | tr '\\' '\n'

Check what Interface pairs up with a veth interface:

ethtool -S $vethDevice

This ID will match up with the entry number in the output of ip link or ip addr

An example of using this to match up a veth interface in a namespace to its corresponding veth interface in another namespace:

root@node-1:~# ip netns exec haproxy ip -d link show dev b_public
36: b_public: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP mode DEFAULT group default qlen 1000
    link/ether 32:cc:0c:f3:ea:34 brd ff:ff:ff:ff:ff:ff promiscuity 0
    veth
root@node-1:~# ip netns exec haproxy ethtool -S  b_public
NIC statistics:
     peer_ifindex: 37
root@node-1:~# ip -o -d link | grep "^37"
37: v_public: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast master br-ex state UP mode DEFAULT group default qlen 1000\    link/ether d2:06:96:a4:39:18 brd ff:ff:ff:ff:ff:ff promiscuity 1 \    veth
root@node-1:~#

List all Network Namespaces without using the ip command:

ls /var/run/netns

Find the Inode of a process’s Network Namespace:

readlink   /proc/{0..9}*/ns/net  | sort | uniq

Find the Inode of a Network Namespace:

stat /var/run/netns/*

OpenStack Networking Notes

Quickly find what router is associated with a Network from the DB

I have yet to find a quick way to identify which Neutron Router is associated with a given Neutron Network ID, so I’ve assembled the following query to speed up this process:

SELECT router_id,port_id,network_id  FROM ports t1 INNER JOIN routerports t2 ON t1.id = t2.port_id where network_id="NETWORK_UUID";

This query can easily be called in bash with the following function:

network_to_router() { mysql neutron -e "SELECT router_id,port_id,network_id  FROM ports t1 INNER JOIN routerports t2 ON t1.id = t2.port_id where network_id=\"$1\";" ; }

OpenStack Client Notes

Easier Python Bindings

The standard OpenStack Python bindings are painful to use. Python Shade is so much easier. Python Shade

Using clouds.yaml

An easier way to define configuration options for the CLI and Python Shade bindings. No more exporting environment variables! clouds.yaml An example clouds.yaml featuring forcing a v3 endpoint (without v2 identity endpoints in the service catalog) as well as forcing the cli to use the public endpoint.

clouds.yaml can be placed in ~/config/openstack/ /etc/openstack/ or in your current working directory.

Example clouds.yaml

clouds:
  mos8:
    region_name: RegionOne
    auth:
      username: admin
      password: admin
      project_name: admin
      auth_url: https://172.16.104.35:5000/v3
      user_domain_name: default
    identity_api_version: 3
    interface: public
    verify: False

Using a clouds.yaml

$ openstack  --os-cloud mos8 token issue

Using a socks proxy with the openstack clients

The latest version of python requests supports utilizing a socks proxy, thus allowing the Openstack clients to utilize a socks proxy directly

$ sudo pip install --upgrade requests
$ sudo pip install requests[socks]

You will can now export the following environment variables in your shell session to utilize a socks proxy.

$ export http_proxy="socks5://localhost:2081"
$ export https_proxy="socks5://localhost:2081"

You will now be able to utilize the standard openstack cli tools normally.