I relaunched my cluster from the scratch (due to another reason). After the
relaunch I could ran nodetool repair -par -inc -pr on the nodes without
issue, but pretty match the moment when I started pushing production load
to the cluster I ran into the same problem again. I opened a ticket first
On Fri, Oct 31, 2014 at 8:55 AM, Juho Mäkinen juho.maki...@gmail.com
wrote:
I can't yet call this conclusive, but it seems that I can't run
incremental repairs on the current 2.1.1 and I'm still wondering if anybody
else is experiencing the same problem.
You have repro steps, if I were you I
It appears to come from the ActiveRepairService.prepareForRepair portion of the
Code.
Are you sure all nodes are reachable from the node you are initiating repair
on, at the same time?
Any Node up/down/died messages?
Rahul Neelakantan
On Oct 30, 2014, at 6:37 AM, Juho Mäkinen
No, the cluster seems to be performing just fine. It seems that the
prepareForRepair callback() could be easily modified to print which node(s)
are unable to respond, so that the debugging effort could be focused
better. This of course doesn't help this case as it's not trivial to add
the log