Hello Nicola, In my world, any cluster that has a backend network different from the public network is misdesigned and broken. The reason is clearly spelled out here https://docs.ceph.com/en/latest/rados/troubleshooting/troubleshooting-osd/#flapping-osds When a private network (or even a single host link) fails or degrades while > the public network continues operating normally, OSDs may not handle this > situation well. In such situations, OSDs use the public network to report > each other down to the monitors, while marking themselves up. The > monitors then send out-- again on the public network--an updated cluster > map with the affected OSDs marked down. These OSDs reply to the monitors > “I’m not dead yet!”, and the cycle repeats. We call this scenario > ‘flapping`, and it can be difficult to isolate and remediate. Without a > private network, this irksome dynamic is avoided: OSDs are generally either > up or down without flapping. And when such flapping happens, the cluster essentially halts, from the client perspective, because of the uninterrupted stream of pauses for peering. The only situations where I would just growl instead of complaining as loudly as possible are: 1. The user has an LACP bond of two interfaces for the backend network, connected to two different switches in a MC-LAG configuration, plus a monitoring system that raises even a single link down as a critical incident. 2. The user proves that the backend network cannot fail without the public network also failing. This includes the case where both are implemented using VLANs on the same NIC, plus possibly some QoS settings on the switch. On Sat, Feb 15, 2025 at 8:58 AM Nicola Mori <nicolamori@xxxxxxx> wrote: > Dear Ceph users, > > I have a question about the benefits of creating a back network for my > cluster. Currently, my machines are all connected to a gigabit switch, > each with 2 bonded interfaces. The cluster just have a front network. > Recently I got a new gigabit switch, so now I have 2 and I was wondering > if there would be any benefit in setting up a back network on one switch > and a front network on the other, using one interface per network. In > this way I'll loose the redundancy, but other benefits and drawbacks are > not clear to me. > Would some expert help me in better understanding this issue, please? > Thanks in advance, > > Nicola > _______________________________________________ > ceph-users mailing list -- ceph-users@xxxxxxx > To unsubscribe send an email to ceph-users-leave@xxxxxxx > -- Alexander Patrakov _______________________________________________ ceph-users mailing list -- ceph-users@xxxxxxx To unsubscribe send an email to ceph-users-leave@xxxxxxx
 |