Old-School Infra Engineers on PB-Scale AI Logs: Hadoop Did This on a Tuesday

basedjensen · x · 2026-09-28

An infra veteran mocks AI safety folks treating petabyte-scale model logs as "an unprecedented observability problem" needing fleets of agents — telecom CDRs, signaling traces and network logs at PB scale were routine Hadoop work a decade ago.

The recipe: dump into HDFS, partition by day/site, Hive compiling into a massive MapReduce job, YARN distributing tasks, then fight data skew (one reducer owning 40% of keys) and rerun. Node dies? Hadoop retries elsewhere. "Our agentic observability platform was Grafana, grep, awk, a 2012 shell script, and one senior engineer staring at reducer 317 stuck at 99%. PB of logs is not scary. We had standards."

Original post →

More from Fun

Fun channel →