)]}'
{"/PATCHSET_LEVEL":[{"author":{"_account_id":13671,"name":"Vladislav Belogrudov","email":"v.belogrudov@yadro.com","username":"vb"},"change_message_id":"b147de7078246437df779ceab77a06d18ced9ca3","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":1,"id":"c9d8988f_1e169ba6","updated":"2021-11-30 08:32:55.000000000","message":"Hi Maciej,\n\nfrom nvme-disconnect man page:\n       Disconnects and removes one or more existing NVMe over Fabrics controllers. If the --nqn option is specified all controllers connecting to the\n       Subsystem identified by subnqn will be removed. If the --device option is specified the controller specified by the --device option will be\n       removed.\n\nWe cannot disconnect an individual disk with the command - only whole nqn subsystems or controllers. Also the device should just disappear once storage system \"unmaps\"/\"unbinds\" the disk from the host. Still I believe such situations with ghost volumes can happen (broken session?). It\u0027s only safe to delete controller or subsystem when there are no volumes managed by them for the host. \n","commit_id":"8d3c7638e511f25f66eea40c9eae3f51595b7845"},{"author":{"_account_id":28050,"name":"Maciej Szwed","email":"maciej.szwed@intel.com","username":"maciejszwed"},"change_message_id":"dbf15baa89abc57f8d4b6fc972b9dc0afce60ab8","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":1,"id":"a82c3ff5_13fd11f7","updated":"2021-11-30 08:20:26.000000000","message":"Vladislav I just noticed your patch, that removes disconnect command. We should still have disconnect command to avoid ghost drivers on initiator side. I think we could disconnect using -d \u003cdevice\u003e instead using nqn - let me know what do you think about it.","commit_id":"8d3c7638e511f25f66eea40c9eae3f51595b7845"},{"author":{"_account_id":28050,"name":"Maciej Szwed","email":"maciej.szwed@intel.com","username":"maciejszwed"},"change_message_id":"f88c0033f33f52937188a6b665456bf8f58880c8","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":1,"id":"0648ddda_94ef8225","in_reply_to":"c9d8988f_1e169ba6","updated":"2021-11-30 11:11:02.000000000","message":"You are right. We should probably track the number of volumes within a controller and disconnect controller if all volumes were removed, but controller still exists in the system (broken session case). I will prepare patch for that if you agree that this would be sufficient solution.","commit_id":"8d3c7638e511f25f66eea40c9eae3f51595b7845"}]}