Hi, in our case the problem was on the client side. When you write "logs from a host", do you mean an OSD host or a host where client connections come from? Its not clear from your problem description *who* is requesting the wrong peer. The reason for this message is that something tries to talk to an older instance of a daemon. The number after the / is a nonce that is assigned after each restart to be able to distinguish different instances of the same service (say, OSD or MDS). Usually, these get updated on peering. If its clients that are stuck, you probably need to reboot the client hosts. Please specify what is trying to reach the outdated OSD instances. Then a relevant developer is more likely to look at it. Since its not MDS-kclient interaction it might be useful to open a new case. Best regards, ================= Frank Schilder AIT Risø Campus Bygning 109, rum S14 ________________________________________ From: U S <ultrasagenexus@xxxxxxxxx> Sent: Tuesday, September 19, 2023 5:35 PM To: ceph-users@xxxxxxx Subject: Re: libceph: mds1 IP+PORT wrong peer at address Hi, We are unable to resolve these issues and OSD restarts have made the ceph cluster unusable. We are wondering if downgrading ceph version from 18.2.0 to 17.2.6. Please let us know if this is supported and if so, please point me to the procedure to do the same. Thanks. _______________________________________________ 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
 |