OpenStack continuous integration generates a fair amount of logs. Devstack, tempest, and testr are quite verbose in what they have done. This information is very valuable, but is also very dense. Let's discuss which logs are important, how long they should be archived for, how they should be archived, and how they should be indexed.
Current ideas are that we might try to archive six months of logs and use logstash/elasticsearch to index a smaller subset of the logs. However, this is not set in stone and feedback will be very useful hence this design session.
This session is motivated by the recent issues with the log server running out of disk space and disagreement on which logs we actually need. The infra team wants to make sure we are logging information that is useful.