Support of haproxy log aggregation capabilites

[Santhosh Fernandes]

We need to collect haproxy logs of amphora instance to a centralized location for our monitoring and analysis. Can we add hooks to achieve log aggregation, so that we can integrate with splunk, graylog,logstash/kibana and others.

[Ganpat Agarwal (gans-developer)]

steps did:

Added variable "log_server" in "haproxy_amphora" section in octavia.conf
Passed it as parameter while creating instance for JinjaTemplater
Added log_server to common/config.py

haproxy does not allow to configure FQDN as log address.
 ( http://cbonte.github.io/haproxy-dconv/1.8/configuration.html#log )

When provided FQDN, config file check is failing.

Will have to pass on data in IP(:PORT) format.

[Michael Johnson (johnsom)]

Yes, we also disable DNS inside the amp for performance reasons.

[Michael Johnson (johnsom)]

Also note, there are two sets of logs with HAProxy, the administrative logs and then the connection tracking logs. This bug is tracking the connection tracking logs and not the administrative logs.

They should be handled separately.

[Adam Harwell (adam-harwell)]

I believe in our meeting topic about this, there was general consensus around using a syslog related method to capture logs, for a variety of reasons.
http://eavesdrop.openstack.org/meetings/octavia/2017/octavia.2017-06-14-20.00.log.html#l-181

[OpenStack Infra (hudson-openstack)]

Change abandoned by Michael Johnson (<email address hidden>) on branch: master
Review: https://review.opendev.org/471682
Reason: This is no longer needed as log offloading has replaced this idea.

[Gregory Thiemonge (gthiemonge)]

Abandoned after re-enabling the Octavia launchpad.