Releases: m-lab/etl-gardener
Add TCPInfo instance, various other fixes.
This release has many changes from April 1 to June 12.
Primary changes are:
- Change table detail query to count distinct rows and tasks. This should improve sanity check reliability.
- Add tcpinfo config, and queries for dedup and sanity check.
- Update table detail query to support tcpinfo, which does not have test_id or task_filename fields.
The tcpinfo instance has been working in sandbox. Not clear whether it is WAI in staging. It successfully processed one date to completion, but then the queues had to be cleared and instance restarted.
Travis best practices, sidestream dedup query
Release: Gardener with updated de-dup query
#140 update dedup query and table detail query
Should reduce lost rows and work around duplicate historical archives.
Increase recheck delay to 60 minutes
#136 Increase delay to 60 minutes
Don't trust StreamingBuffer status
This release includes critical change to recheck StreamingBuffer status, which is unreliable.
Advancing minor version number, as this should correct a major bug and allow correct behavior.
#133 Use histogram for time-in-state
#131 wait 5 minutes for stable table (actually 10 minutes)
#132 fix restart with dateSkip
Retain templated table on error
#130 delay deleting dedup source
This allows examination of the dedup source when sanitycheck fails.
Prod - preserve error states in datastore
Fix task queue false "empty queue" triggers
The major update is to fix false IsEmpty checks due to occasional TQ bug.
Also includes improvements to logging, and a new metric to track suspicious TQ empty stats.
Change START_TIME to include complete history
Merge pull request #116 from m-lab/sandbox-all Set correct starting dates for all 4 data types
Production release - add traceroute dataset
This release adds the traceroute gardener to production.
As well, this release changes the source bucket for all gardeners to archive-measurement-lab.