Reputation: 41
Varnish 4 is retrying the request every 120 seconds continuously if the corresponding backend endpoint is taking more than 120 seconds. While the backend is retrying the request, the original varnish request is waiting. The original varnish request is not returning 503. All the other backend endpoints are responding within 1 minute. Default backend timeout and max-retries settings are used. No explicit request restarts are used in the vcl. The varnish version is 4.1.4
Is it the expected behaviour or is it a bug? If it is a bug, how can it be mitigated?
Upvotes: 1
Views: 701
Reputation: 1322
This is probably caused by this bug: https://github.com/varnishcache/varnish-cache/pull/2135
Varnish is supposed to retry once when using keepalive connections. But there was a bug in the logic and in some edge cases it retried infinite times.
Upvotes: 1