Elena,

I disabled SeLinux and ran into yet another problem:

create table t1 ( rowkey varchar(36) primary key, data1 varchar(60), data2 bigint ) engine=cassandra thrift_host='localhost' keyspace='xxx' column_family='xxx';

When in fact I created column family xxx under xxx keyspace.

Please note that I'm using cassandra 1.2 - I've heard that there issues with this release of Cassandra 1.2. Any ETA when it will be fixed?

Hi Alex,

What problem did you run into with the DDL above? I suppose you forgot to paste the error message.

Here we go:

MariaDB [xxx]> create table t1 ( rowkey varchar(36) primary key, data1 varchar(60), data2 bigint ) engine=cassandra thrift_host='localhost' keyspace='xxx' column_family='xxx';

ERROR 1429 (HY000): Unable to connect to foreign data source: Column family xxx not found in keyspace xxx

Also pls check http://www.slideshare.net/SergeyPetrunya/mysqlconf2013-mariadbcassandrainteroperability slide 11 out 34 there seems to be a know problem with Cassandra 1.2 integration

Right. I'll reassign it to Sergei Petrunia so that you could get firsth-hand information regarding plans to fix it.

Thx a lot, please keep us posted, we have a meetup scheduled specifically to discuss and demo MariaDB and Cassandra integration http://www.meetup.com/mysql-silicon-valley/events/119357352/ fixing this issue will greatly help to promote MariaDB.

Thx in advance

Alex -

Hello,

This seems to be the problem with data model change in Cassandra 1.2.

Cassandra 1.2 seems to have a split in the data model. One can have either:

1. Old-style column family (that one creates with cassandra-cli
or in CQL with "WITH COMPACT STORAGE" attribute)

2. New-style column family (created with CREATE TABLE in CQL-3)

Cassandra itself seems to be poor at interoperability: Thrift clients cannot
access new-style column families. On the other hand, CQL-3 requires that all
columns are pre-defined, which means you cannot use CQL-3 to read data that
was inserted through Thrift.

Cassandra SE uses Thrift and so is a victim of this mess.

I have looked at what is required to update Cassandra SE. It seems that

• We will still need to support Thrift (because some data can only be accessed through Thrift API)
• We will need to use CQL to access new-style CQL column families.

Thrift API has a execute_cql3() method, but the problem with it is that it is not possible to stream the resultset. The entire resultset will be materialized on a node in Cassandra cluster, then sent to MariaDB node, where thrift library will also materialize it before returning control to Cassandra SE. This means that attempt to run "SELECT * FROM cassandra_table" (without the implied "LIMIT 10K") will kill the node. (pure Thrift API supports streaming, current Cassandra SE will run this SELECT * query without consuming all of the memory).

Cassandra 1.2 has a new network protocol for CQL. The protocol supports streaming. Datastax does not provide a C++ client library, but it seems there is a 3rd-party client library: https://github.com/mstump/libcql. I've just found it by googling so I don't have an idea whether this library is sufficient for Cassandra SE.

Interesting read... It seems, DataStax intends to make a C++ client library: https://github.com/mstump/libcql/issues/16#issuecomment-19923916 ... and it is not clear how that will relate to libcql.

Sergei,

Any updates on this bug - I have meetup coming up at Silicon Valley MySQL user group discussing this feature - MariaDB and Cassandra integration. Do we have an ETA on this by any chance?

Unfortunately, I cannot get any resources to continue to work on this. Changing the status to "Wont fix", because currently there are no plans for further development of Cassandra SE.