It sounds like you're trying to read entire rows at once. Past a certain point (depending on your heap size) you won't be able to do that, you need to "page" through them N columns at a time.
On Sat, Jun 4, 2011 at 12:27 PM, Mario Micklisch <mario.mickli...@hpm-kommunikation.de> wrote: > Hello there! > I have ran into a strange problem several times now and I wonder if someone > here has an solution for me: > > For some of my data I want to keep track all the ID's I have used. To do > that, I am putting the ID's as column into rows. > At first I wanted to put all ID's into one row (because the limit of 2 > billion column seemed high enough) and then it happened: > For some reason I was no longer able to remove or update existing rows or > column. Adding new rows and removing them was possible, but write access to > the older rows did not longer work. > I have restarted the cluster, didn't help either. Only truncating the column > family helped. > Because this happened several times after creating rows with 5.000+ columns > I decided to reduce the number of columns to a maximum of 1.000 per row and > everything is now working perfectly. > > I am working on version 0.8-rc1 and for testing purpose I am running only > one node with the default settings. > > My questions are: > 1) Because 0.8 is not marked as stable, is this a known problem? > 2) Is there something in the configuration I need to change to handle a > bigger number of columns? > 3) Can I check on the cassandra-cli why updates are failing? (there were no > error messages when trying deletes with the CLI manually) > 4) Is there a recommended number of columns? (I guess this only depends on > my systems memory?) > > Thanks to everyone who took some time to read! > Mario > -- Jonathan Ellis Project Chair, Apache Cassandra co-founder of DataStax, the source for professional Cassandra support http://www.datastax.com
