Cassandra data migration from 1.9.8 to 2.0.2

[Lakshmi Kanth <lk...@gmail.com>]

Hi

I am trying to migrate data from a 3 node cluster to a new 5 node cluster.
 After completion of the migration I am seeing few additional records in
new system.  After analysis, I found that these records were already
deleted in the source system just few days before the migration.   It seems
like the tomstones from source system are getting migrated to target system.

Please let me know if there is any way to avoid these tombstone rows.

*Source System :*

Cassandra 1.9.8
3 nodes
Replication factor 1.

*Target System :*
Cassandra 2.0.2
5 nodes
Replication factor 3.

*Approach Followed for migration :*

On each of the node from source cluster, flush data and generate snapshot.
Move all the snapshots for all column families to new cluster.
Run sstable loader for all the column families from one of the node from
target cluster.
Re-build indexes.


Regards,
LK

Re: Cassandra data migration from 1.9.8 to 2.0.2

[Lakshmi Kanth <lk...@gmail.com>]

Thanks Robert.

I wrongly mentioned the source Cassandra version.  Actual source version is
1.2.9.


Regards,
LakshmiKanth


On Tue, Mar 18, 2014 at 12:06 PM, Robert Coli <rc...@eventbrite.com> wrote:

> On Mon, Mar 17, 2014 at 6:19 PM, Lakshmi Kanth <lk...@gmail.com> wrote:
>
>> Cassandra 1.9.8
>>
>
> No? This version does not exist.
>
>
>> Cassandra 2.0.2
>>
>
> Cassandra 2.0.x versions up to and including 2.0.5 have serious issues.
> Including this one, which randomly tombstones your data when you use
> SSTableloader.
>
> https://issues.apache.org/jira/browse/CASSANDRA-6527 (fixed in 2.0.4)
>
> This is the reason why I do not currently recommend running Cassandra
> 2.0.x in production.
>
> https://engineering.eventbrite.com/what-version-of-cassandra-should-i-run/
>
> If I were you, with such small cluster sizes, I would consider the
> "copy-all-data-to-all-nodes-and-run-cleanup" method.
>
> http://www.palominodb.com/blog/2012/09/25/bulk-loading-options-cassandra
>
> =Rob
>
>

Re: Cassandra data migration from 1.9.8 to 2.0.2

[Robert Coli <rc...@eventbrite.com>]

On Mon, Mar 17, 2014 at 6:19 PM, Lakshmi Kanth <lk...@gmail.com> wrote:

> Cassandra 1.9.8
>

No? This version does not exist.


> Cassandra 2.0.2
>

Cassandra 2.0.x versions up to and including 2.0.5 have serious issues.
Including this one, which randomly tombstones your data when you use
SSTableloader.

https://issues.apache.org/jira/browse/CASSANDRA-6527 (fixed in 2.0.4)

This is the reason why I do not currently recommend running Cassandra 2.0.x
in production.

https://engineering.eventbrite.com/what-version-of-cassandra-should-i-run/

If I were you, with such small cluster sizes, I would consider the
"copy-all-data-to-all-nodes-and-run-cleanup" method.

http://www.palominodb.com/blog/2012/09/25/bulk-loading-options-cassandra

=Rob