When I created the schema, the KS were created as directory in the .../data/ directory
Can you provide the logs from start up ? 


I just noticed that when the new cluster nodes start up, auto_bootstrap is set to true. Given that it's no longer in the YAML file, I didn't set it to false; would that explain it ?
That will not cause any problems. 

Cheers

-----------------
Aaron Morton
Freelance Cassandra Consultant
New Zealand

@aaronmorton

On 1/05/2013, at 9:08 AM, Philippe <watcherfr@gmail.com> wrote:

Hi Aaron,
thanks for the response.

Permissions are correct : owner is cassandra (ubuntu) and permissions are drwxr-xr-x

When I created the schema, the KS were created as directory in the .../data/ directory
When I use cassandra-cli, set the CL to QUORUM, ensure two instances are up (nodetool ring) and set a bogus column, I get an Unavailable exception which is weird.

I just noticed that when the new cluster nodes start up, auto_bootstrap is set to true. Given that it's no longer in the YAML file, I didn't set it to false; would that explain it ?

Once I removed it and set logging to TRACE, I get these relevant log lines

SliceQueryFilter.java (line 123) collecting 1 of 2147483647: XXX:5123@1367333147034
a bunch of Schema.java (line 382) Adding org.apache.cassandra.config.CFMetaData@45f8acdc[cfId=1009,ksName=XXX,cfName=YYY...
Removing compacted SSTable files from XXX
No bootstrapping, leaving or moving nodes -> empty pending ranges for XXX (my KS)
Table.java (line 317) Initializing XXX.YYY (KS and CF)

Any ideas ?


2013/4/30 aaron morton <aaron@thelastpickle.com>
Double check the file permissions ?

Write some data (using cqlsh or cassandra-cli) and flush to make sure the new files are created where you expect them to be.

Cheers

-----------------
Aaron Morton
Freelance Cassandra Consultant
New Zealand

@aaronmorton
http://www.thelastpickle.com

On 1/05/2013, at 4:15 AM, Philippe <watcherfr@gmail.com> wrote:

> Hello,
>
> I'm trying to bring up a copy of an existing 3-node cluster running 1.0.8 into a 3-node cluster running 1.0.11.
> The new cluster has been configured to have the same tokens and the same partitioner.
>
> Initially, I copied the files in the data directory of each node into their corresponding node on the new cluster. When starting the new cluster, it didn't pick up the KS & CF.
> So I remove the directories, created the schema, stopped cassandra and copied the data back. None of the SSTables are opened at startup.
>
> When I nodetool refresh them, it says "No new SSTables where found for  XXX/YYY" where X is my KS and Y is my CF
>
> I'm totally stumped: any ideas as to why this is happening ? I've checked that there are actual files there and that the permissions are correct.
>
> Thanks.
>