CODEWITHSUNDEEP
cassandradatastax-java-driver
Micah Zoltu
Micah Zoltu

Reputation: 7432

Retry all hosts forever Retry Policy?

I have some data that must be written to Cassandra. Initially I thought this would be simple by just setting up a RetryPolicy for ReadTimeout, WriteTimeout, Unavailable and RequestError. However, the only two retry decisions that I see are tryNextHost and retry.

retry appears to try forever, but it tries the same host over and over again. Even if the node cleanly leaves the cluster and announces its departure (handing off ownership of its range to another node), this retry policy appears to continue to try to send to the same host.

tryNextHost solves that problem, but it appears to only try each host once and then gives up (lets the exception bubble up).

I need a retry policy that rotates through hosts (or at least moves hosts when one cleanly leaves the cluster) but I can set to retry forever. I investigated writing my own RetryDecision but there is no public constructor and even if I got around that, I don't see a way to get around this since its constructor is pretty limited.

How can I retry forever AND rotate hosts?

Upvotes: 0

Views: 549

Answers (1)

Olivier Michallat
Olivier Michallat

Reputation: 2312

RetryPolicy alone will not allow you to do that. The other component involved is LoadBalancingPolicy, which determines which hosts will be tried (in other words, it generates the sequence over which repeated tryNextHost calls would iterate).

So I guess you could write a custom load balancing policy that returns infinite iterators. But frankly that's quite unusual and you should consider the implications of requests potentially looping forever.

Upvotes: 2

Related Questions