things_you_need

inspired by https://goncalo.mendescabrita.com/notes/things-you-should-definitely-do-in-a-greenfield-application/

• dns caching on your cluster's nodes; maybe in your app
• compression on your load balancer
• handle ssrf (see https://github.com/arkadiyt/ssrf_filter)
• solid_cache as a scratchpad on your DB is very useful
• sentry is cool but if you're not self hosting it needs to be heavily limited
• logs/traces/metrics are all useful, so is having redis, kafka, debezium, dagster, different analytics DB, ...

background jobs

• alerts on sidekiq poison pill
• queues and workers based on job latency SLA. HPA on that same metric.
• you'll eventually need transactional push with bulk support (sorry sidekiq)

sql

• low lock_timeout on migrations (1s) with a retry loop
• querytags in all queries
• postgres-lock-logger: something to periodically inspect lock trees
• pganalyze is sweet
• you'll need a connection pooler eventually
• something to reindex old indexes
• maybe sometimes a good ole pg_squeeze
• low statement_timeout, idle_in_transaction_session_timeout, lock_timeout, transaction_timeout, idle_session_timeout