The clients won't care about the split_realms and the kclient will
treat this as a corrupted snaptrace.
Fixes commit
93e7267757508520dfc22cff1ab20558bd4a44d4 ("mds: send snap
related messages centrally during mds recovery")
Fixes: https://tracker.ceph.com/issues/61217
Signed-off-by: Xiubo Li <xiubli@redhat.com>
<< realm->get_newest_seq() << " on " << *realm << dendl;
auto snap = make_message<MClientSnap>(CEPH_SNAP_OP_UPDATE);
snap->bl = mds->server->get_snap_trace(client, realm);
- for (const auto& child : realm->open_children)
- snap->split_realms.push_back(child->inode->ino());
updates.emplace(std::piecewise_construct, std::forward_as_tuple(client), std::forward_as_tuple(snap));
} else {
dout(10) << "finish_snaprealm_reconnect client." << client << " up to date"