Ceph

Oops. Here's the rest of the report:

../../src/osd/PG.cc: 3891: FAILED assert(0 == "we got a bad state machine event")

The usage scenario here was 6-8 osds per node (5 nodes total). Two nodes were shutdown, causing the osds on those nodes to disappear from the map, and rebalancing to begin. That all seemed to go fine, until the nodes and osds restarted on the two nodes that had been shutdown. This caused many of the osds (but not all) on the other nodes to crash. They all seem to fail at the same assertion above. The end of the log from one of the crashed osds is attached, as well as a core file.