[
https://issues.apache.org/jira/browse/TINKERPOP-1453?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15504382#comment-15504382
]
Carlos Cheung commented on TINKERPOP-1453:
------------------------------------------
How much of an issue would it be to remove most of the Tornado specifics and
implement a threaded solution?
I can't seem to see a way to implement this utilizing Tornado's future as it's
built in the way that it's assumed 1 result = 1 response. However, in this case
Gremlin Server could very well produce 2 responses for every result in the case
of an error.
> Allow Gremlin-Python to handle asynchronous failure
> ---------------------------------------------------
>
> Key: TINKERPOP-1453
> URL: https://issues.apache.org/jira/browse/TINKERPOP-1453
> Project: TinkerPop
> Issue Type: Improvement
> Components: driver, language-variant
> Affects Versions: 3.2.2
> Environment: Titan 1.1
> Gremlin 3.2.2
> Reporter: Carlos Cheung
>
> Titan occasionally returns transaction errors for a transaction that was sent
> a few requests back. Since Gremlin-Python assumes a request sent will get
> only its own answer back, this can cause the request-answer to no longer be
> in sync and there is no mechanism to recover.
> It would be nice to have a way for the receiver to check the request-id to
> see if it corresponds to the original request. Additionally there should be a
> rolling transaction log in the case of a server error, so a recovery/try
> might be possible.
> By implementing this, it would be possible to send Gremlin queries
> asynchronously as well.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
