ceph version 0.80.5 the reason that primary osd drop the secondary osd's subop reply message is because of a bug that the message sequence number is truncated to unsigned 32 bit from unsigned 64 bit.
the bug is already reported in the upstream: https://tracker.ceph.com/issues/16122 and fixed at https://github.com/ceph/ceph/pull/9456
it's already backported to Jewel, but not to harmer
ceph version 0.80.5
the reason that primary osd drop the secondary osd's subop reply message is because of
a bug that the message sequence number is truncated to unsigned 32 bit from unsigned 64 bit.
the bug is already reported in the upstream: https:/ /tracker. ceph.com/ issues/ 16122 /github. com/ceph/ ceph/pull/ 9456
and fixed at https:/
it's already backported to Jewel, but not to harmer