PXaaS integration with Monitoring

Two types of metrics are supported:

• Generic metrics using collectd
• VNF specific metrics using a Python script and the monitoring API

Generic metrics

Requirements

• Collectd version 5.5.1 (Debian comes with collectd v 5.1 so I installed v 5.5.1 manually. I followed this)
• Make sure that the nameserver 10.30.0.11 is defined under /etc/resolv.conf

Steps

• Install Collectd v 5.5.1
• Edit collectd.conf under /etc/collectd/collectd.conf
• Change 'Hostname' (VNF's OpenStack instance UUID) with the output of

curl -s http://169.254.169.254/openstack/latest/meta_data.json | sed -e 's/.*"uuid": "//; s/".*//'

• Alternative run once the script init_uuid in order to get the instance UUID automatically at startup
• Define the ip and port of the monitoring server. Change 'server' field under 'Plugin network'
• Define the interface you need to monitor. Change 'Interface' field under 'Plugin interface'
• Restart collectd

sudo service collectd restart

VNF specific metrics

Metrics sent to the Monitoring Server

• Number of HTTP requests received. The number of HTTP requests received by Squid since the last measurement.
• Cache hits percentage of all requests. The percentage of HTTP requests that result in a cache hit for the last 5 minutes. It also includes cases in which Squid validates a cached response and receives a 304 (Not Modified) reply.
• Cache hits percentage of bytes sent. The percentage of all data that is transferred directly from the cache rather than from the origin server.
• Memory hits percentage. The percentage of all cache hits that were served from memory (hits that are logged as TCP_MEM_HIT in Squid’s logs).
• Disk hits percentage. The percentage of all cache hits that were served from disk (hits that are logged as TCP_HIT in Squid’s logs)
• Cache disk utilization. The amount of disk currently being used by the cached objects divided by the total amount of disk that can be allocated for caching.
• Cache memory utilization. The amount of memory (RAM) currently being used by the cached objects divided by the maximum amount of memory that can be allocated for caching.
• Number of users accessing the proxy. Squid assumes that each user has a unique IP address.

Requirements

• squidclient v 3.5.5. Collects statistics from Squid proxy
• Python 2.7
• scripts
• Make sure that the nameserver 10.30.0.11 is defined under /etc/resolv.conf

Steps

• Make sure that squidclient v3.5.5 and Python 2.7 are installed
• Edit monitoring.py. Change 'host' and 'port' to the corresponding monitoring server's host and port. Change 'instance_uuid' (see above)
• Run the script. It sends once the metrics to the monitoring server

python monitoring.py

• Create a cron job to send the metrics on time intervals

crontab -l > mycron
echo "*/1 * * * * python /home/proxyvnf/monitoring/monitoring.py" >> mycron # executes once a minute
crontab mycron
rm mycron