HTTP module default changes in k6 v0.50 impacting script compatibility

Direct answer

To replicate v0.47 HTTP behavior in v0.50, explicitly declare timeout, maxRedirects, and noConnectionReuse in your k6 script. These are the controls for request timeout, redirect following, and TCP connection reuse. Whether v0.50 actually changed their defaults relative to v0.47 is version-sensitive and not confirmable from general model knowledge; verify against the exact installed version’s documentation or source before assuming a default shift.

Why these three options matter

• timeout — sets the per-request wait limit. A lower timeout can truncate slow responses and raise http_req_failed; a higher one can mask latency regressions.
• maxRedirects — caps automatic redirect following. Following a redirect adds another request and its timing to http_req_duration and can change the final status code. Lowering the cap can convert a redirect loop into a terminal 3xx, altering failure semantics.
• noConnectionReuse — controls keep-alive reuse. Reuse skips TCP/TLS setup on later requests, lowering http_req_connecting and http_req_tls_handshaking. Disabling reuse moves that time into setup and can expose DNS/TCP/port limits under high concurrency.

Connection reuse and metric accuracy under high concurrency

Connection reuse does not change total server work, but it changes where time is attributed. With reuse, setup cost is paid once per connection and amortized across requests, so http_req_duration looks lower and http_req_connecting stays near zero. Without reuse, every request pays DNS, TCP, and TLS setup, inflating http_req_blocked and http_req_connecting. If a v0.47 baseline reused connections and a v0.50 run does not, the delta can look like a server regression when it is only a client-side connection strategy difference. Set noConnectionReuse to match your baseline; do not enable it as a general tuning change, because it can reduce throughput and increase resource use.

Steps to pin the baseline

1. Set the three options explicitly to the values your v0.47 baseline used.

import http from 'k6/http';
export const options = {
  timeout: '60s',           // match v0.47 baseline
  maxRedirects: 10,         // match v0.47 baseline; 0 disables redirects
  noConnectionReuse: false, // match v0.47 reuse behavior
};

1. Run the same minimal script under both versions against a local endpoint with a known redirect and a slow response. Compare res.status, res.timings, and the summary metrics http_req_blocked, http_req_connecting, http_req_tls_handshaking, http_req_duration, and http_req_failed.
2. Keep scenario, tags, and thresholds identical. If thresholds fail only in v0.50, capture the exact metric and scenario name; if requests fail, capture response.error_code and whether the URL redirects.

One diagnostic detail that changes the recommendation

Are the discrepancies you are seeing threshold-only failures, or request-level errors? If they are threshold-only, focus on connection-reuse attribution and timing deltas. If they are request-level, capture the error code and redirect chain to determine whether maxRedirects or timeout is the active variable.

Confirm the installed version with k6 version and check the matching options reference or source for exact defaults before changing your script.