nodetool repair stalled

[Paolo Crosato]

Hi,

I have two nodes with Cassandra 2.0.3, where repair sessions hang for an 
undefinite time. I'm running nodetool repair once a week on every node, 
on different days. Currently I have like 4 repair sessions running on 
each node, one since 3 weeks and none has finished.
Reading the logs I didn't find any exception, apparently one of the 
repair session got stuck at this command:

 INFO [AntiEntropySessions:10] 2014-01-05 01:00:02,804 RepairJob.java 
(line 116) [repair #5385ea40-759c-11e3-93dc-a1357a0d9222] requesting 
merkle trees for events (to [/10.255.235.19, /10.255.235.18])

Has anybody any suggestion on why a nodetool repair might be stuck and 
how to debug it?

Regards,

Paolo Crosato