-----Original Message----- > From: Liu Yuan [mailto:[email protected]] > Sent: Wednesday, February 29, 2012 6:16 PM > To: huxinwei > Cc: [email protected]; Liu Jiang > Subject: Re: [Sheepdog] panic in get_nth_node > > On 02/29/2012 04:30 PM, huxinwei wrote: > > > Hi list: > > > > In my environment (2 sheep only), sheep always panic while recovering > from a left node returning. > > > > It turns out to be a intend behavior in get_nth_node: > > > > ========================================= > > if (idx == base) { > > panic("bug"); /* not found */ > > ========================================= > > > > While I agree this is the correct in most scenarios, it does seem to be > > too > intrusive while recovering in my trivial test. > > To be specific, find_tgt_node calls get_nth_node > > > > I don't have a lot of faith in my own workaround either. Let me know what > you think ;) > > > > Thanks. > > > How to reproduce this issue in your case?
Let's say we have 2 nodes, running 3 sheep instance Node1 # sheep /home/sheep1 -p 7000 Node2 # sheep /home/sheep1 -p 7000 Node2 # sheep /home/sheep2 -p 7001 Node1 # collie cluster format -c 2 # farm or simple doesn't matter here Node1 # collie vdi create ss1 1G ... Keep ss1 busy writing ... Then kill sheep2 on node2, will almost surely panic sheep1 on node2 too. FYI. > Thanks, > Yuan -- sheepdog mailing list [email protected] http://lists.wpkg.org/mailman/listinfo/sheepdog
