incubator-cassandra-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jehan Bing <je...@orb.com>
Subject 0.6.13->1.0 upgrade + node move -> resurrected data
Date Thu, 08 Dec 2011 23:30:50 GMT
Hi,

Here is what I do, starting from 4 0.6.13 nodes with RF 2:
- repair all the nodes
- compact all the nodes
- upgrade to 1.0
- scrub all the nodes
- repair all the nodes
- compact all the nodes
- move one node to a new token

For some requests, I don't get the same result before and after the 
move. Some deleted columns are undeleted. Why are they back? Some of the 
data was deleted in September so any tombstone should have expired a 
long time ago (default GCGraceSeconds) so where does the old data come 
from? Shouldn't the repair+compact made sure that it's all gone?

What can I do to fix this?


	Jehan


Mime
View raw message