All posts / tag
#platform
-
Catching the secret before the commit, not after the audit
A leaked credential is cheapest to stop before git history. This is how leakferret scans staged changes, filters fixtures, verifies keys, and rewrites code.
-
What concurrent Ruby looks like, twelve years after the book
A cache benchmark stayed serial as thread count rose. I traced the Ruby runtime, then tested where mutexes, Ractors, fibers, and snapshots help.
-
Debugging Redis::CannotConnectError in Ruby
Redis timeouts arrived in short bursts while infrastructure stayed healthy. One Ruby client path rebuilt the connection and dropped its configured timeout.
-
Dynamic per-user Rails debug logs, scoped to the request
One user could reproduce a production bug. A request-scoped Rails override captured their debug logs, expired on its own, and left other traffic alone.
-
Tuning Rails log levels per class, without a redeploy
One Rails class needed debug logs in production. A temporary per-class override changed at runtime and expired automatically, without enabling debug across the app.
-
Rolling PgBouncer without dropped queries
PgBouncer rollouts reset live client sockets after Kubernetes drained the pod. Shorter Rails connection lifetimes closed the proxy shutdown timing gap.
-
When config edits start feeling like deploys
A saved setting took 30 seconds to reach every worker. The database kept the value while a ConfigMap watch became the fleet-wide change signal.