Closed ryandotsmith closed 4 months ago
Another idea is to explicitly provide a "backup" url. I like this approach less because it is common for backup resources to be flawed since they are rarely used.
Shovel should expose metrics and logs with data about QoS for each url
Added in #259
If an RPC is failing for whatever reason it would be nice for Shovel to use a backup RPC URL.
I am thinking we can do something like this:
As Shovel downloads data, it will evenly balance requests across the URLs. If a request fails from non-200 or a timeout, the request will be retried with the next url in the list.