Error creating ceph.conf

[verdurin]

I'm seeing this error when I try to create a luminous cluster:

fatal: [ceph0]: FAILED! => {
    "msg": "'dict object' has no attribute u'ipv4'"
}

This is presumably because that interface has no IP address yet.

If I change the monitor_interface to refer to an interface that already has an IP, I see this error:

fatal: [ceph0]: FAILED! => {
    "msg": "No first item, sequence was empty."
}

Here are the values I've added to group_vars/all.yml:

osd_scenario: lvm
osd_objectstore: bluestore
lvm_volumes:
  - data: /dev/sda
    db: db-sda-lv
    db_vg: pilot-ceph0
    wal: wal-sda-lv
    wal_vg: pilot-ceph0
  - data: /dev/sdb
    db: db-sdb-lv
    db_vg: pilot-ceph0
    wal: wal-sdb-lv
    wal_vg: pilot-ceph0
  - data: /dev/sdc
    db: db-sdc-lv
    db_vg: pilot-ceph0
    wal: wal-sdc-lv
    wal_vg: pilot-ceph0
  - data: /dev/sdd
    db: db-sdd-lv
    db_vg: pilot-ceph0
    wal: wal-sdd-lv
    wal_vg: pilot-ceph0
  - data: /dev/sde
    db: db-sde-lv
    db_vg: pilot-ceph0
    wal: wal-sde-lv
    wal_vg: pilot-ceph0
  - data: /dev/sdf
    db: db-sdf-lv
    db_vg: pilot-ceph0
    wal: wal-sdf-lv
    wal_vg: pilot-ceph0
  - data: /dev/sdg
    db: db-sdg-lv
    db_vg: pilot-ceph0
    wal: wal-sdg-lv
    wal_vg: pilot-ceph0
  - data: /dev/sdh
    db: db-sdh-lv
    db_vg: pilot-ceph0
    wal: wal-sdh-lv
    wal_vg: pilot-ceph0
  - data: /dev/sdi
    db: db-sdi-lv
    db_vg: pilot-ceph0
    db: db-sdi-lv
    db_vg: pilot-ceph0
    wal: wal-sdi-lv
    wal_vg: pilot-ceph0
  - data: /dev/sdj
    db: db-sdj-lv
    db_vg: pilot-ceph0
    wal: wal-sdj-lv
    wal_vg: pilot-ceph0
  - data: /dev/sdk
    db: db-sdk-lv
    db_vg: pilot-ceph0
    wal: wal-sdk-lv
    wal_vg: pilot-ceph0
  - data: /dev/sdl
    db: db-sdl-lv
    db_vg: pilot-ceph0
    wal: wal-sdl-lv
    wal_vg: pilot-ceph0

ceph_origin: "repository"

ceph_repository:  "community"

ceph_stable_release: "luminous"

radosgw_address_block: "172.16.16.0/24"

monitor_interface: "eno4"

public_network: "172.16.48.0/24"

cluster_network: "172.16.64.0/24"

ip_version: "ipv4"

I've seen the same error or very similar errors with stable-3.0, stable-3.1 and master.

[guits]

@verdurin Are you using ansible 2.6 ? We currently support only 2.4 and 2.5

[verdurin]

@guits no, it's 2.4.2.0.

[guits]

please, share the full playbook run log in that case

[andymcc]

@verdurin it'd be useful to see what facts Ansible is seeing too - using "ansible all -m setup -i inventory_file" or something similar. The error seems to suggest the network CIDR doesn't match the ip addresses on the interface - or something similar.

[verdurin]

Thanks for looking. The context here is that I'd been trying to do this via TripleO, and it was failing owing to device naming issues across the nodes, so I thought I'd just use plain ceph-ansible. That background meant an implicit assumption that it would setup interfaces for me, which is obviously nonsense.

[verdurin]

Right. Today I was using ceph-deploy, so I configured the interfaces on the nodes. Now ceph-ansible is proceeding. I do wonder if the error shown here could be improved?

Do you still want to see that information?

[guits]

@verdurin any chance you can provide the full log ?

[verdurin]

@guits you mean of the last time it failed? (it worked with properly configured interfaces)

[guits]

@verdurin If you have an actual issue, please, provide the log. Otherwise, I'll assume I can close this issue.

[verdurin]

@guits It's working for me now, so I'll close it. Thanks.