Mastering Service Mesh

Retrying failed transactions

As we explained in Chapter 14, Understanding the Linkerd Service Mesh, a blind retry on failed transactions can lead to retry storms. At the time of writing, Linkerd attempts to address this issue by introducing the retry and retry budget configurations.

Before we attempt to retry the failed transactions, it is important to decide and know which request should be re-tried and how many times. To do this, follow these steps:

1. Run the linkerd routes from books to authors and view the metrics:

$ linkerd -n linkerd-lab routes deploy/books --to svc/authorsROUTE                       SERVICE   SUCCESS      RPS  ---DELETE /authors/{id}.json   authors     0.00%   0.0rps  ---GET /authors.json           authors     0.00%   0.0rps  ---GET /authors/{id}.json authors 0.00% 0.0rps ...