Hi Benedikt,
in my opinion it seems best to open a JIRA ticket and adding a SSCCE* to it,
so it doesn't get lost and someone can take care about it. Just in case you
already have a patch proposal at hand, we of course appreciate that too.
Regards,
JensG
* see http://sscce.org
-----Ursprungliche Nachricht-----
From: Benedikt, Neuenfeldt (NED)
Sent: Thursday, November 28, 2013 1:56 AM
To: [email protected]
Subject: Re: TFramedTransport sending empty buffer
Hi Jens,
thank you for your prompt reply.
The cassandra version is 1.2.11 and on the client side I'm using Hector
with Thrift 0.7.0.
The error only seems to occur when a read timeout on the client side
occurs though.
I have tracked it to the following steps:
1) Read timeout occurs in native code
2) Hector closes the connection
3) Before the Thrift closes the connection, it flushes and sends a 0
byte frame
The easiest way to replicate it is to set the client read timeout to a
very low number (e.g. 100ms)
With best regards,
Ben
(2013/11/28 6:27), Jens Geyer wrote:
Hi Ben,
from my experiences with both Cassandra and Thrift this is somewhat
unusual.
Maybe it helps to know, which language you are using and what Cassandra
client (if any)?
Have fun,
Jens
-----Ursprungliche Nachricht-----
From: Benedikt, Neuenfeldt (NED)
Sent: Wednesday, November 27, 2013 8:53 AM
To: [email protected]
Subject: TFramedTransport sending empty buffer
While using Cassandra I noticed that TFramedTransport flushes before
closing the connection.
However, it will still flush the buffer even if it is empty, basically
sending a 0 sized packet.
This results in Cassandra reporting: Read an invalid frame size of 0.
Are you using TFramedTransport?
I am not sure whether this is part of the protocol or whether it is
Cassandra being overly verbose.
Any ideas on this? The issue still persists when using the latest
version (0.9.1).
With best regards,
Ben
