Hi All, After struggling for some time with Cassandra clustering the conclusion is that we cannot deliver Cassandra clustering with 0.1.0 (if we stick to the deadline of feb 1). After intensive testing there just doesn't seem to be a working workaround for this issue:
https://issues.apache.org/jira/browse/CASSANDRA-1992 Effectively this means that after adding a node to the cluster, data is missing, rows and/or columns are moved to different CF's or duplicated. As a consequence, Cassandra storage providers might or might not start, will throw all kind of exceptions and (hopefully this is indeed triggered by this issue) ConcurrentModificationExceptions are thrown in the Felix whiteboard service. Restarting nodes, performing nodetool repairs, cleanups, compactions, etc. etc. do not fix those issues, in fact in most cases they make it even worse. This morning Brandon attached a patch to issue 1992. I tried to apply the patch but with this patch applied on the trunk (which I doubt is the 'real' version as the version number is still 0.7.0-rc4) the Cassandra daemon doesn't start at all (it throws a java.lang.NoSuchFieldError: instanceinstance). So we will have to wait until this patch is approved, merged to the trunk and preferably for the 0.7.1 version in which this is planned to be fixed. Until then, we cannot setup a Cassandra cluster with more than 1 node. I will however finish the remaining todo's so that you can setup a Cassandra cluster in 0.1.0, it just doesn't work ;-> Regards, Ivo GX | Ivo Ladage-van Doorn | Product Architect | Wijchenseweg 111 | 6538 SW Nijmegen | The Netherlands | T +31(0)24 - 388 82 61 | F +31(0)24 - 388 86 21 | ivo.ladage-vandoorn at gxsoftware.com<mailto:ivo.ladage-vandoorn at gxsoftware.com> | www.gxsoftware.com<http://www.gxsoftware.com> | twitter.com/GXSoftware<http://twitter.com/GXSoftware> -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.amdatu.org/pipermail/amdatu-developers/attachments/20110119/08e60f0c/attachment-0001.html
