Millisecond log timestamp precision lost in fluentd config

[Doug Szumski]

Issue
-----

We have truncated the timestamp so that we have lost millisecond resolution in the output from fluentd.

Example raw log message:
------------------------

==> nova/nova-placement-api.log <==
2017-10-17 14:54:37.374 21 INFO nova.api.openstack.placement.requestlog [req-112f3ff0-a877-4b16-bd89-ac009e81a704 ac342c8d47c8416580ec6f3affcd287f 4970f0b152ca41dc968b4473bb8a48d9 - default default] X.X.X.X "GET /resource_providers/6b5e1d8e-0174-4ba6-8b5b-6c72a213ca6b/inventories" status: 200 len: 520 microversion: 1.0

Example output from fluentd:
----------------------------

2017-10-17T14:54:37+00:00 openstack_python {"Pid":"21","log_level":"INFO","python_module":"nova.api.openstack.placement.requestlog","request_id":"112f3ff0-a877-4b16-bd89-ac009e81a704","user_id":"ac342c8d47c8416580ec6f3affcd287f","tenant_id":"4970f0b152ca41dc968b4473bb8a48d9","domain_id":"-","user_domain":"default","project_domain":"default","Payload":"X.X.X.X \"GET /resource_providers/6b5e1d8e-0174-4ba6-8b5b-6c72a213ca6b/inventories\" status: 200 len: 520 microversion: 1.0","Hostname":node_0001","Logger":"openstack.nova","programname":"nova-placement-api"}

[Doug Szumski (dszumski)]

This may require upgrading fluentd to v0.14: https://github.com/fluent/fluentd/issues/461

[Stig Telfer (stigtelfer)]

Millisecond precision on log timestamps is now fixed, at least it is on this system (Wallaby)