Comment 3 for bug 1680774

Previously, I was using kernel 3.19 and I don't remember having the issue.
I don't really want to test an unstable kernel on production infrastructure. I was hoping for some concrete tips to troubleshoot scsi layer problems.
Anyways, if I deployed it now, I could confirm that it does NOT happen in no less than half a year. The particular machine that failed now had an uptime of 119 days, another one with the same software has 135 days and counting. But taken together, I'm experiencing the same problem about every month in the cloud cluster.