Skip to content

Latest commit

 

History

History
7 lines (4 loc) · 1.42 KB

README.md

File metadata and controls

7 lines (4 loc) · 1.42 KB

PkgServerLogAnalysis.jl

First off, to use these processing routines, you'll need access to the logs. If you've already gotten the requisite SSH access to the PkgServers correctly, then sync_logs.jl will connect to them just fine. If not, you can test new analysis passes on the test data contained within the test suite, stored within an artifact in this repository. See bin/load_test_data.jl for an example oof how to load that test data. The test data is generated through applying bin/anonymize_log.jl to a real logfile.

The logs are stored in separate log files per day, per server, collated by logrotate on each individual server. sync_logs.jl transfers the compressed gzip archives of each day's logs, and parse_logfiles() reads them in with a giant regex to parse out relevant pieces of information. Because parsing is fairly expensive, we cache the parse results in compressed .csvz files, stored in a scratch space.

Right now, there is a single main analysis script stored in bin, package_hits_by_day.jl. It's a Pluto.jl notebook, I suggest running it inside of a Pluto session with multiple threads, to improve parsing efficiency via Pluto.run(;threads="8") or so. Future decisions will solidify the contract of analysis passes, and determine the best way to present it to the world.