stress

README

Stress Example

This example provides a repeatable stress harness for Flows on Citus.

It supports two modes:

• --mode=worker: run a worker process (intended to be run as multiple containers)
• --mode=generator: enqueue a fixed number of workflow runs for A/B/C/D and wait for completion, then print DB-derived completion stats.

Quick start (Citus + 5 worker containers)

From repo root:

1. Start Postgres (Citus) and 5 worker containers:

cd examples/stress
export FLOWS_TEST_ID=$(date +%s)
docker compose up --build -d --scale worker=5

1. Run the generator on the host (recommended):

go run ./examples/stress \
  --mode=generator \
  --database-url "postgres://test:test@localhost:5432/flows_test?sslmode=disable" \
  --schema flows \
  --shards 32 \
  --test-id "$FLOWS_TEST_ID" \
  --total-a 100000 \
  --total-b 10000 \
  --total-c 1000 \
  --total-d 100

This writes JSON outputs under examples/stress/results/.

Optional: sample docker CPU/memory

If you have Docker CLI on the host:

go run ./examples/stress \
  --mode=generator \
  --database-url "postgres://test:test@localhost:5432/flows_test?sslmode=disable" \
  --test-id "$FLOWS_TEST_ID" \
  --docker-stats-interval 2s \
  --docker-stats-filter "citus,worker"

The docker samples are written to examples/stress/results/<test_id>_docker_stats.jsonl.

Notes

• Each workflow type runs in its own goroutine pool. The --workers flag controls the concurrency per workflow type.
• For Citus, the worker queries with an equality predicate on workflow_name to enable proper shard routing.
• Completion stats are derived from runs.created_at and runs.updated_at in the Flows schema.