)]}'
{"specs/keystone/backlog/fernet-etcd.rst":[{"author":{"_account_id":6488,"name":"Clint Byrum","email":"clint@fewbar.com","username":"clint-fewbar"},"change_message_id":"65ce4d829ff80dbd14a5956229748075a920eb38","unresolved":false,"context_lines":[{"line_number":65,"context_line":"thoroughly examined and controlled. The current Fernet implementation limits"},{"line_number":66,"context_line":"recommends limited access to the `/etc/keystone/fernet-keys` directory. This"},{"line_number":67,"context_line":"same practice should be recommended for deployments wishing to use etcd for"},{"line_number":68,"context_line":"Fernet key storage."},{"line_number":69,"context_line":""},{"line_number":70,"context_line":"Notifications Impact"},{"line_number":71,"context_line":"--------------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7f231b9d_dc37a5e3","line":68,"updated":"2017-06-09 16:13:10.000000000","message":"This understates the impact quite a bit.\n\nWith files on disk, you have a _push_ distribution model. Which means that you choose exactly when and where to send your key material. Of course, you can screw up the permissions, and a local user on those machines could read them, but the only compromise opportunity is by a local process on machines where you have chosen to put key material.\n\nBy contrast, etcd3 supports strong cryptographic auth via client cert:\n\nhttps://coreos.com/etcd/docs/latest/op-guide/security.html\n\nThis means that you now have _new_ key material to generate, rotate, and distribute to each keystone node to authenticate to etcd3. So now you have all the hardship of local file key storage, but none of the benefits. Add to this that there\u0027s no audit logging in etcd, and you have a recipe for key compromise.","commit_id":"98a08cb259a5ec49b44fd4b055833d4e2849584a"},{"author":{"_account_id":5046,"name":"Lance Bragstad","email":"lbragstad@redhat.com","username":"ldbragst"},"change_message_id":"3edbaba69c1ccabac841df0896cb2236fc4d78d2","unresolved":false,"context_lines":[{"line_number":65,"context_line":"thoroughly examined and controlled. The current Fernet implementation limits"},{"line_number":66,"context_line":"recommends limited access to the `/etc/keystone/fernet-keys` directory. This"},{"line_number":67,"context_line":"same practice should be recommended for deployments wishing to use etcd for"},{"line_number":68,"context_line":"Fernet key storage."},{"line_number":69,"context_line":""},{"line_number":70,"context_line":"Notifications Impact"},{"line_number":71,"context_line":"--------------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7f231b9d_dccb2570","line":68,"in_reply_to":"7f231b9d_dc37a5e3","updated":"2017-06-09 16:34:59.000000000","message":"Yeah - that\u0027s a valid point. I was mainly looking at the operator overhead of another etcd deployment specific for fernet keys, but the case you\u0027ve brought up here makes a lot of sense.\n\nIf I spin another revision, I\u0027ll certainly add it. With those points though, I\u0027m not sure how useful this is anymore.\n\nThanks for the review!","commit_id":"98a08cb259a5ec49b44fd4b055833d4e2849584a"}]}
