[CASSANDRA-6439] get_range_slice doesn't support wrapping (start_key, end_token) range

• Type: Bug
• Status: Resolved
• Priority: Low
• Resolution: Won't Fix
• Severity: Low
• Since Version: 1.1.9
• Assignee: Tom Hobbs
• Reporter: Francesco Piccinno

Authors

Tom Hobbs

• Created: 03/Dec/13 17:01
• Updated: 16/Apr/19 09:31
• Resolved: 23/Feb/15 23:55

Description

I am trying to achieve a linear scan on the data contained in a cassandra node by exploiting tokens. The idea behind my approach is to request through the pycassa SystemManager a list of tokens that the cluster is responsible of, and then for each token, issue a key_range command specifying start and end interval. The problem is that apparently some tokens returned by the server does not respect the property start < end. I think that the bug is due to the fact that Murmur3RandomPartitioner is used, but I am just guessing here.

Anyway here are the steps to reproduce the bug:

Triggering the bug

$ pycassaShell -k KEYSPACE
In [1]: tokens = map(lambda x: (int(x.start_token), int(x.end_token)), SYSTEM_MANAGER.describe_ring('KEYSPACE'))

In [2]: len(filter(lambda x: x[0] < x[1], tokens))
Out[2]: 255

In [3]: len(filter(lambda x: x[0] > x[1], tokens))
Out[3]: 1

In [4]: filter(lambda x: x[0] > x[1], tokens)
Out[4]: [(9207458196362321348, -9182599474778206823)]

In [5]: for i in CF.get_range(start_token="9207458196362321348", finish_token="-9182599474778206823"): print i
# ...
# after some objects are printed
# ...
InvalidRequestException: InvalidRequestException(why="Start key's token sorts after end token")