Problem attaching volume: Connector doesn't have required information: wwpns

asked 2018-03-14 04:43:59 -0500

Vaggelis Kapoulas gravatar image

updated 2018-05-03 09:02:27 -0500

I have an ocata setup with 3 controller nodes (for HA), but I have stoped two of them while tracking my problem described below (so that I have only a set of log files to check). Cinder backend is an IBM storwiz family storage system, accessed using the fc driver. My setup mostly works ok except for the following problem:

When I configure glance to use cinder as backend, and I try to create an image, a volume is successfully created in the backend storage, but then it cannot be attached (I gues to the controller node, so that the image can be written to it). The respective error message (in the cinder volume log) is:

ERROR cinder.volume.drivers.ibm.storwize_svc.storwize_svc_fc [req-7bceaa65-07a9-47f9-b2ee-9865e144f1b5 1ad16950fda84785ab1e52747eedcfd1 287d4f4befaa4cea8a1cf77e7f2ff301 - default default] The connector does not contain the required information.

and a few lines below a slightly more informative

ERROR oslo_messaging.rpc.server InvalidInput: Invalid input received: Connector doesn't have required information: wwpns

After these errors, several similar errors are reported, the process obviously fails, and then the volume is (successfully) deleted. The relevant parts of the cinder and glance logs are appended at the end of this question.

A similar problems occurs when I try to create an instance with a permanent boot volume. The process fails as it seems that the volume cannot be attached to the controller node (to write the boot image to it). Similar errors appera in the logs.

However, I can create and attach additional volumes to instances (created with ephemeral boot disks), and the instances can use these volumes just fine (after mounting and growing a filesystem in them). That means that volumes can be attached to compute nodes.

I have looked at this and other sites for similar problems, but I have found no solution to my issue.

I have made sure that I have installed the necessary packages (sysfsutils, sg3_utils).

I have made sure that I have access to the backend starage from my controller over fc:

# systool -c fc_host -v
Class = "fc_host"

  Class Device = "host1"
  Class Device path = "/sys/devices/LNXSYSTM:00/device:00/ACPI0004:00/VMBUS:00/ebcad967-3e6d-40a7-b13a-5f76d189ee08/host1/fc_host/host1"
    dev_loss_tmo        = "60"
    node_name           = "0xc003ff0000ffff00"
    port_name           = "0xc003ff27871e0004"
    tgtid_bind_type     = "wwpn (World Wide Port Name)"
    uevent              =

    Device = "host1"
    Device path = "/sys/devices/LNXSYSTM:00/device:00/ACPI0004:00/VMBUS:00/ebcad967-3e6d-40a7-b13a-5f76d189ee08/host1"
      uevent              = "DEVTYPE=scsi_host"

I have also created a volume in the backend storage and mounted it to the controller node, so there seems to have no problem in that part.

Enabling debug output in cinder logs provides no further insight. Nothing of interest is reported after the successful creation of the volume and the first error message.

It just seems that the service responsible for preparing the connector information cannot find or include the wwpn in the connector. As I don't know the relevant workflow, I don't know what else to try or how to further troubleshoot my problem.

The only difference between the compute nodes (which volumes seem to ... (more)

edit retag flag offensive close merge delete

Comments

@fattony666 thanks for the suggestion. I have already sysfsutils installed. It seems the problem is due to the FC driver not fully populating the fc_host class in the VMs. I avoided my problems by using the file backend for glance and by installing cinder on real machines.

Vaggelis Kapoulas gravatar imageVaggelis Kapoulas ( 2018-05-03 09:23:56 -0500 )edit