Re-trying: after timeout, POST requests

Hi all,

this is a re-post from slack.

To reduce long-tail latency of a service, we’d like to configure linkerd to retry requests after a timeout. The docs seem to focus on retries after 50x responses, so I was wondering if this is possible?

When using linkerd in front of an Elasticsearch cluster, it’d be useful to be able to retry all failed requests, not just idempotent ones. Elasticsearch search requests are usually submitted as POST requests, despite having GET semantics, since the query is transmitted in the body. Is this possible?

Thanks,
–Chris

@ccmtaylor Linkerd currently does not allow POSTs to be retryable since by per RFC2616 they are not idempotent. Not to worry though, you could create a custom response classifier that includes classifies POSTs as retryable The code for Linkerd’s response classifiers can be found here and provides a great start to writing a custom classifier.

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.