cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Andy Tolbert (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (CASSANDRA-12686) Communicate timeouts and other driver relevant options in SUPPORTED response or some other mechanism
Date Thu, 22 Sep 2016 15:11:20 GMT

     [ https://issues.apache.org/jira/browse/CASSANDRA-12686?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Andy Tolbert updated CASSANDRA-12686:
-------------------------------------
    Labels: client-impacting  (was: )

> Communicate timeouts and other driver relevant options in SUPPORTED response or some
other mechanism
> ----------------------------------------------------------------------------------------------------
>
>                 Key: CASSANDRA-12686
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-12686
>             Project: Cassandra
>          Issue Type: Improvement
>            Reporter: Andy Tolbert
>              Labels: client-impacting
>
> It would be really useful if driver clients had a mechanism to understand what the configured
timeouts on the C* side are.
> Ideally a driver should be configured in such a way that it's client timeout is greater
than the C* timeouts ({{write_request_timeout_in_ms}}, {{read_request_timeout_in_ms}}, etc.)
so its retry policy may make the appropriate decision based on the kind of timeout received
from cassandra.   This is why most driver clients have a client timeout of 12 seconds.   If
the client knew the server timeouts, it could adjust its client timeout accordingly.
> At the moment, the only place I think where this could be communicated is through a {{SUPPORTED}}
message when the client sends an {{OPTIONS}} message, but that could be viewed as awkward.
 Also consider that some clients use the {{OPTIONS}} message as a form of heartbeat, so adding
more to a {{SUPPORTED}} message could add some (likely trivial) data on the wire between server
and client.
> Alternatively, it could also be interesting if the client could configure the timeout
on the server end (with some ceiling set by C*).



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message