tursodatabase/turso · error

--seconds requires --benchmark throughput

Error message

--seconds requires --benchmark throughput

What it means

The FTS benchmark runner only accepts --seconds (benchmark duration) when running the throughput benchmark. In the Search benchmark branch of run(), ensure! rejects any invocation where --seconds was supplied, because search benchmarks are query-count driven, not time driven. This is CLI argument validation to prevent ambiguous benchmark configurations.

Solutions

1. Remove --seconds from the command line when using the search benchmark
2. Switch to --benchmark throughput if a time-limited run is actually desired
3. If a wrapper script always passes --seconds, make it conditional on the benchmark mode

Example fix

// before
fts-bench --benchmark search --seconds 10 --queries 1000
// after
fts-bench --benchmark search --queries 1000

if benchmark == Search && seconds.is_some() { panic!("--seconds only applies to --benchmark throughput"); }

When it happens

Trigger: Running the perf/fts binary with both --benchmark search (or default search) and a --seconds value, e.g. `fts-bench --benchmark search --seconds 10`. The ensure! at perf/fts/src/main.rs:71 fires inside run() whenever args.benchmark is Benchmark::Search and args.seconds.is_some().

Common situations: Developers copying a throughput-benchmark command line and switching only the benchmark name; scripting harnesses that always pass --seconds regardless of benchmark mode; misunderstanding that search mode measures fixed query counts rather than a time window.

Understand the failure class

Background: Conflicting config options: "cannot be used together" — configuration validation errors across open-source libraries — this error's family across 162 libraries.

Related errors

• first-query latency requires one query per connection
• need at least one query per connection
• --queries requires --benchmark search
• seconds must be finite and positive
• throughput requires --state warm

AI-assisted analysis of tursodatabase/turso@8d4a589f8d (2026-09-20). Data as JSON: /api/errors/8111f0cbcb287b2d. Report an issue: GitHub.

Appendix: source

Thrown at perf/fts/src/main.rs:71

    ensure!(args.connections > 0, "connections must be positive");
    eprintln!("bundled SQLite version: {}", rusqlite::version());
    let runtime = match args.benchmark {
        Benchmark::Search => tokio::runtime::Builder::new_current_thread()
            .enable_all()
            .build()?,
        Benchmark::Throughput => tokio::runtime::Builder::new_multi_thread()
            .worker_threads(args.connections)
            .max_blocking_threads(args.connections)
            .enable_all()
            .build()?,
    };
    runtime.block_on(run(args))
}

async fn run(mut args: Args) -> Result<()> {
    let (query_limit, min_seconds) = match args.benchmark {
        Benchmark::Search => {
            ensure!(
                args.seconds.is_none(),
                "--seconds requires --benchmark throughput"
            );
            let queries = args
                .queries
                .unwrap_or(if matches!(args.state, QueryState::First) {
                    args.connections
                } else {
                    10_000
                });
            ensure!(
                queries >= args.connections,
                "need at least one query per connection"
            );
            ensure!(
                !matches!(args.state, QueryState::First) || queries == args.connections,
                "first-query latency requires one query per connection"
            );

View on GitHub (pinned to 8d4a589f8d)