incubator-cassandra-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Stefan Kaufmann <sta...@gmail.com>
Subject uncomplete RangeSlices
Date Wed, 18 Aug 2010 09:52:29 GMT
During my tests, I found some strange bootstrap + get Range Slice behavior.

In my testing environment im creating Datasets which look like those:
sampleDPUIDforTestingIssues1/00000000000
 -> value = 123456789
sampleDPUIDforTestingIssues1/00000999999
 -> value = 123456789

i create 1 Million of those Keys (increasing number at the end) and then I
have a query-script which uses get RangeSlices (10 000 Keys per Slices) in a
loop to query all keys. The script checks if all key's could be queried.

When I use the Bootstrap option to add a new node, mye query script shows me
a large amount of missing keys. (looks like the same key-range from one node
in the ring). However, when I try to query one single key out of the range
(not using slices), I can read the value without any problem.


My Script says:
Following errors for sampleDPUIDforTestingIssues1 collection found:
471886 - 736000 are missing.

This is the ring configuration:
Address       Status     Load
Range                                      Ring

sampleDPUIDforTestingIssues1/00000736000
10.20.10.112  Up         252.67 MB
qaDOmXnF48Q4sZks                                      |<--|
10.20.10.111  Up         252.67 MB
sampleDPUIDforTestingIssues1/00000471885   |   |
10.20.10.110  Up         252.66 MB
sampleDPUIDforTestingIssues1/00000736000   |-->|


When I add new nodes, by coping the storage/Keyspace data instead of
bootstrapping - everything works fine.

I searched the Issues-DB, and I found those two old ones..
(I'm not sure, if they really have to do something with this)

https://issues.apache.org/jira/browse/CASSANDRA-682
https://issues.apache.org/jira/browse/CASSANDRA-902

Mime
View raw message