Ceph

Well, we could probably log each time a new client authenticates with the mon?

That would be good enough, IMHO.

- One more enhancement would be great here.
- As we recommend running both ceph server side daemons and clients to be in same version and if they are not in same version log a warning in ceph.log(cluster logs).

- It will help a lot in case of NOVA/KVM client instances upgrade procedure. For example if nova instances(qemu-kvm) processes are running firefly and ceph cluster daemons are running hammer.

- Now we upgrade our nova-compute nodes to hammer but did not stop and start nova instances(qemu-kvm)processes (needs a down time) or did not live-migrate(do not need downtime)to other nova-computes.

- This causes nova qemu-kvm in memory process still running firefly and server code running hammer code. After that if we will change the tunables to hammer (and bucket algorithm to straw2) all instacnes which are still running firefly code in memory will crash in instance logs:

terminate called after throwing an instance of 'ceph::buffer::malformed_input'
  what():  buffer::malformed_input: *unsupported bucket algorithm: 5*

- This feature will help to avoid such mistakes.
- As if we see version mismatch warning after upgrading the clients also in cluster logs(ceph.log) we will come to know we still needs to either stop and start instances or live migrate them.

- While discussing with Kefu and Brad we have one more idea of adding a log if tunables are not compatible between ceph cluster and clients.

- Also these all logs can be handled by 'config_opts.h' with addition of new options and by default true.

Why not log all clients' versions when they connect, regardless if they're older/newer/equal to the server? (That was the original request).

(You mentioned one specific problem about version incompatibilities, but there are other use cases for logging the version, e.g. monitoring client statistics).

Sorry Dan if I have made things confusing.

My new note is for one additional request in same path of enhancement suggested by you and your original request is still valid and it will not be a warning message it will be a INFO message like Sam had added Log each time a new client authenticates with the mon.

With my suggestion it would be one more addition to the log if client and server are not in same version and then log a WRN message in ceph.log(cluster logs) and ERR if tunables are not compatible.

I should have not changed the title as title covers all the three requests!

Cool, sounds good.

I think the ceph-mgr PR that John put in has some of the rudiments to enable connected-client version tracking. You may want to coordinate with that, Kefu.

I can't imagine we'll add a persistent warning about old clients, since we work hard to preserve cross-version compatibility. Warnings about that should be the responsibility of things layered on top of Ceph, like whatever GUIs get built that talk to ceph-mgr. (also, kernel clients don't have anything like the same versions that userspace does ;)

My 2 cents re ceph-mgr: it would be nice if this feature was simple enough to be easily backported to jewel, otherwise OP (me) won't be able to enable the vary_r tunable until the L release :-/

Really, all we need is to log a client's IP when it connects with feature flags less than a configured value. You can even log it at level 3 or something so normal clusters aren't spammed.