Huhhh.. It seems worthwhile to point out two inconsistencies, then. 1. the "old way", of ceph config set global rbd_cache false doesnt require this odd redundant "global set global" syntax. It is confusing to users to have to specify "global" twice. may I suggest that the syntax for rbd config be adjusted, so that people can simply use rbd config global ... or rbd config client ... not rbd config global set (global/client) 2. The fact that "ceph config set global rbd_cache false" works, seems to imply that putting [global] rbd cache = false in /etc/ceph/ceph.conf should work also. Except it doesnt. Even after fully shutting down every node in the ceph cluster and doing a cold startup. is that a bug? ----- Original Message ----- From: "Jason Dillaman" <jdillama@redhat.com> To: "Philip Brown" <pbrown@medata.com> Cc: "dillaman" <dillaman@redhat.com>, "ceph-users" <ceph-users@ceph.io> Sent: Thursday, December 17, 2020 8:24:59 AM Subject: Re: [ceph-users] Re: bug? cant turn off rbd cache? On Thu, Dec 17, 2020 at 11:21 AM Philip Brown <pbrown@medata.com> wrote:
I guess I left out in my examples, where I tried rbd_cache as well, and failed
# rbd config global set rbd_cache false rbd: invalid config entity: rbd_cache (must be global, client or client.<id>)
But that's not a valid command -- you attempted to provide the configuration key name for the entity (as the error message is telling you). $ rbd config global set global rbd_cache false ... or ... $ rbd config global set client rbd_cache false ... or ... $ rbd config global set client.host_a rbd_cache false