SocketTimeoutException should be translated as HectorTransportException

[sbridges]

In ExceptionTranslator, org.apache.cassandra.thrift.TimedOutException and java.net.SocketTimeoutException are both translated as HTimedoutException.

I don't think it makes sense to translate these the same. TimedOutException is thrown by cassandra when an rpc call takes longer than the configurable timeout, while SocketTimeoutException is a networking issue.

We had an issue where hector would hang when one of our cass machines was powered down, as by default it looks like we retry HTimedOutExceptions forever.

I think it makes more sense to translate SocketTimeoutException as a HectorTransportException, which would correctly mark the host as down.

If you agree, I can submit a patch.

[patricioe]

+1 from my side. Thanks.

[sbridges]

What is the protocol. I committed,

sbridges@fb76ae6

which shows up in this bug above, do I issue a pull request as well?

[patricioe]

Sorry, I don't follow this last part.

Are you trying to say that that is still an issue in Cassandra ?

the removed lines seem ok to me.

On Wed, Mar 7, 2012 at 4:36 PM, sbridges <
reply@reply.github.com

wrote:

What is the protocol. I committed,

sbridges@fb76ae6

which shows up in this bug above, do I issue a pull request as well?

---

Reply to this email directly or view it on GitHub:
https://github.com/rantav/hector/issues/434#issuecomment-4382267

[sbridges]

just a git hub workflow issue. I pushed the changes to sbridges/hector, and github puts a note on this issue. Do I issue a pull request for rantav/hector now?

[patricioe]

sure