Ceph

Joao had this to say:

personally, I don't think it would be too hard to get updates on deleted pools being outputted to the cluster log (which is basically what 'ceph -w' shows). I wonder, however, whether other people would be reasonably interested in that info.

fwiw, I'm under the impression that the audit log entries should be sent to the subscriber (e.g., 'ceph -w' or whatever). and there should be at least two audit log entries referring the pool deletion - one at receipt of message, another at acknowledgement of success/failure. These log messages are not amazingly readable by a human though, but should be there.

If anyone though wants to give a try at getting proper log messages on this, this should happen in OSDMonitor::prepare_pool_op_delete().

Please bear in mind that the log message should only be written to mon->clog() once the command has successfully finished. If I had to guess, it would probably take creating a new Context that would take the message to be written to clog, as well as the current C_Command() or something, and then issuing the write to clog() and calling C_Command's finish() upon finishing the context successfully. And now that I wrote this, this may not be as trivial as I initially thought. Not too complicated though.

I see a few commands executed via the ceph cli getting logged in the cluster log, for eg. osd pool create/delete, injectargs etc, eg.:

from='client.? 127.0.0.1:0/2039727119' entity='client.admin' cmd=[{"pool2": "foo", "prefix": "osd pool delete", "sure": "--yes-i-really-really-mean-it", "pool": "foo"}]: dispatch

Though if we go via the rados tool, or say the admin socket or via librados pybind for eg., these ops do not seem to appear in the log anymore.

what do we want?
may be we want the ceph -w can display the results of rados tool, admin socket or via librados pybind?