Hi, There’s previous discussion about this issue. https://lists.ceph.io/hyperkitty/list/ceph-users@xxxxxxx/thread/ZZZGEVD4L4SEMEBMTZYDW2OOVWQFXGOA/ Can you check if there is any host with _admin label on it? Removing the label works for me. Regards, Gene Kuo > On Apr 27, 2022, at 18:21, Ulrich Klein <Ulrich.Klein@xxxxxxxxxxxxxx> wrote: > > Hi, > > Yesterday I upgraded my smallest test system, 4 Raspberries 4B, from Pacific 16.2.7 (cephadm/containerized) to 17.2.0 using > ceph orch upgrade start --ceph-version 17.2.0 > > It mostly worked ok, but wouldn't have finished without manual intervention. > Apparently each time a mgr is upgraded the process creates new /etc/ceph/ceph.conf and /etc/ceph/ceph.client.admin.keyring files on all nodes. > To do that it looks like it first copies the files to /tmp/etc/ceph/ceph.conf on the node, then changes owwner and permission and then tries to move the file into place. Unfortunately it changes owner/permission in way so that it doesn't have permission to write to and move the file resulting in somethig like this in an infinite (?) loop: > > 2022-04-27T09:03:45.032808+0000 mgr.ceph00.lpaijp (mgr.2314108) 605 : cephadm [ERR] executing refresh((['ceph00', 'ceph01', 'ceph02', 'ceph03'],)) failed. > Traceback (most recent call last): > File "/usr/share/ceph/mgr/cephadm/ssh.py", line 221, in _write_remote_file > await asyncssh.scp(f.name, (conn, tmp_path)) > File "/lib/python3.6/site-packages/asyncssh/scp.py", line 922, in scp > await source.run(srcpath) > File "/lib/python3.6/site-packages/asyncssh/scp.py", line 458, in run > self.handle_error(exc) > File "/lib/python3.6/site-packages/asyncssh/scp.py", line 307, in handle_error > raise exc from None > File "/lib/python3.6/site-packages/asyncssh/scp.py", line 456, in run > await self._send_files(path, b'') > File "/lib/python3.6/site-packages/asyncssh/scp.py", line 438, in _send_files > self.handle_error(exc) > File "/lib/python3.6/site-packages/asyncssh/scp.py", line 307, in handle_error > raise exc from None > File "/lib/python3.6/site-packages/asyncssh/scp.py", line 434, in _send_files > await self._send_file(srcpath, dstpath, attrs) > File "/lib/python3.6/site-packages/asyncssh/scp.py", line 365, in _send_file > await self._make_cd_request(b'C', attrs, size, srcpath) > File "/lib/python3.6/site-packages/asyncssh/scp.py", line 343, in _make_cd_request > self._fs.basename(path)) > File "/lib/python3.6/site-packages/asyncssh/scp.py", line 224, in make_request > raise exc > asyncssh.sftp.SFTPFailure: scp: /tmp/etc/ceph/ceph.conf.new: Permission denied > > During handling of the above exception, another exception occurred: > > Traceback (most recent call last): > File "/usr/share/ceph/mgr/cephadm/utils.py", line 76, in do_work > return f(*arg) > File "/usr/share/ceph/mgr/cephadm/serve.py", line 265, in refresh > self._write_client_files(client_files, host) > File "/usr/share/ceph/mgr/cephadm/serve.py", line 1052, in _write_client_files > self.mgr.ssh.write_remote_file(host, path, content, mode, uid, gid) > File "/usr/share/ceph/mgr/cephadm/ssh.py", line 238, in write_remote_file > host, path, content, mode, uid, gid, addr)) > File "/usr/share/ceph/mgr/cephadm/module.py", line 569, in wait_async > return self.event_loop.get_result(coro) > File "/usr/share/ceph/mgr/cephadm/ssh.py", line 48, in get_result > return asyncio.run_coroutine_threadsafe(coro, self._loop).result() > File "/lib64/python3.6/concurrent/futures/_base.py", line 432, in result > return self.__get_result() > File "/lib64/python3.6/concurrent/futures/_base.py", line 384, in __get_result > raise self._exception > File "/usr/share/ceph/mgr/cephadm/ssh.py", line 226, in _write_remote_file > raise OrchestratorError(msg) > orchestrator._interface.OrchestratorError: Unable to write ceph02:/etc/ceph/ceph.conf: scp: /tmp/etc/ceph/ceph.conf.new: Permission denied > > > On each node I had to do > cd /usr/bin > mv chmod chmod_real ; ln -s true chmod > mv chown chown_real ; ln -s true chown > > And then whenever the file(s) appeared: > chmod_real 666 /tmp/etc/ceph/ceph.conf.new > > to make it get over that hurdle. And once finished restore the chown/chmod binaries and permissions. > I wonder if anyone else has seen that on Intel/AMD machines? Looks like a pretty obvious problem with the process shooting itself in the permission foot, and on bigger clusters that process would be a time consuming pain. > > Ciao, Uli > > _______________________________________________ > ceph-users mailing list -- ceph-users@xxxxxxx > To unsubscribe send an email to ceph-users-leave@xxxxxxx _______________________________________________ ceph-users mailing list -- ceph-users@xxxxxxx To unsubscribe send an email to ceph-users-leave@xxxxxxx