Ceph

• cascading failures
• single OSD failure
• failure to complete peering/recovery
• unfound objects after recovery
• full
• slow
• fails to respond to some request

• failure

• failure

• stops forwarding requests

• RGW request timeouts
• OSD file system timeouts
• OSD that is "down" but still "in"
• degraded placement groups

Mark Kampe wrote:

Additional issues from Carl's list:

• RGW request timeouts

That's a symptom, not a cause...

• OSD file system timeouts

What timeouts? We have a few that cause suicides but I suspect he just means OSDs being slow in the filesystem.

• OSD that is "down" but still "in"
• degraded placement groups

I'm not sure what either of these are about. Both are revealed with "ceph -s" (more detail under "ceph osd dump" and "ceph pg dump"), and neither are problems in and of themselves.