OpenStack log timestamps are stored into Elasticsearch without millisecs

Bug #1646601 reported by Rakesh Patnaik
12
This bug affects 2 people
Affects Status Importance Assigned to Milestone
Fuel Plugins
Confirmed
Low
Unassigned

Bug Description

LMA collector drops the millisecond fraction of the log line (.876) as listed below

2016-12-01 07:35:34.867 21503 INFO nova.osapi_compute.wsgi.server [req-uuid usrid uuid - - -] xxx.xxx.xxx.xxx "GET /v2/uuid/os-services HTTP/1.1" status: 200 len: 2792 time: 0.0519361

it shows up on elastic search as follows missing the .867 value

       2016-12-01T07:35:34

I attempted modifying the elasticsearch encoder to state timestamp as follows but it stops sending to elasticsearch as some of the other logstreamers do not have the millisecond information as above.

/etc/log_collector/encoder-elasticsearch.toml
timestamp= "%Y-%m-%dT%H:%M:%S.%f"
and
timestamp= "%Y-%m-%dT%H:%M:%S.%L"

as documented at https://github.com/cactus/gostrftime/blob/master/gostrftime.go

Changed in fuel-plugins:
status: New → Confirmed
summary: - LMA collector drops millisecond information from openstack log lines
- before posting to elasticsearch
+ OpenStack log timestamps are stored into Elasticsearch without millisecs
Changed in fuel-plugins:
importance: Undecided → Low
Revision history for this message
Simon Pasquier (simon-pasquier) wrote :

This would require a new release of Heka because v0.10.0 doesn't support the %f flag.

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.