CentOS fails to deploy with LVM storage layout

[Lee Trager]

When trying to deploy CentOS using the default LVM layout the deployment fails with the following error

Traceback (most recent call last):
  File "/tmp/tmpipnJj9/target/curtin/curtin-hooks", line 333, in <module>
    main()
  File "/tmp/tmpipnJj9/target/curtin/curtin-hooks", line 308, in main
    devices = get_block_devices(target)
  File "/tmp/tmpipnJj9/target/curtin/curtin-hooks", line 63, in get_block_devices
    (blockdev, part) = block.get_blockdev_for_partition(maybepart)
  File "/curtin/curtin/block/__init__.py", line 149, in get_blockdev_for_partition
    raise ValueError("%s had no syspath (%s)" % (devpath, syspath))
ValueError: /dev/vgroot-lvroot (dm-0) had no syspath (/sys/class/block/vgroot-lvroot (dm-0))
Unexpected error while running command.
Command: ['/tmp/tmpipnJj9/target/curtin/curtin-hooks']
Exit code: 1
Reason: -
Stdout: ''
Stderr: ''
Installation failed with exception: Unexpected error while running command.
Command: ['curtin', 'curthooks']
Exit code: 3
Reason: -
Stdout: 'Traceback (most recent call last):\n File "/tmp/tmpipnJj9/target/curtin/curtin-hooks", line 333, in <module>\n main()\n File "/tmp/tmpipnJj9/target/curtin/curtin-hooks", line 308, in main\n devices = get_block_devices(target)\n File "/tmp/tmpipnJj9/target/curtin/curtin-hooks", line 63, in get_block_devices\n (blockdev, part) = block.get_blockdev_for_partition(maybepart)\n File "/curtin/curtin/block/__init__.py", line 149, in get_blockdev_for_partition\n raise ValueError("%s had no syspath (%s)" % (devpath, syspath))\nValueError: /dev/vgroot-lvroot (dm-0) had no syspath (/sys/class/block/vgroot-lvroot (dm-0))\nUnexpected error while running command.\nCommand: [\'/tmp/tmpipnJj9/target/curtin/curtin-hooks\']\nExit code: 1\nReason: -\nStdout: \'\'\nStderr: \'\'\n'
Stderr: ''

[Lee Trager (ltrager)]

Deploying with a flat storage layout allows CentOS to install with no issue.

[Lee Trager (ltrager)]

The bug is due to the CentOS Curtin script not being able to understand all of the storage options we now have. Since most of the storage code we use for Ubuntu could be used for CentOS we may want to move the Curtin hooks out of maas-images and add CentOS support to Curtin itself.

[Andres Rodriguez (andreserl)]

To use the flat storage layout:

maas <user> node set-storage-layout <node-id> storage_layout=flat

[Sagar Shukla (sa-shukla)]

When I try to run above mentioned command in my environment, I get following error message.

# maas root node set-storage-layout node-262e1660-67a1-XXXX-YYYY-ZZZZZZZZZ storage_layout=flat
usage: /usr/lib/python2.7/dist-packages/maascli/__main__.py root node
       [-h] COMMAND ...
/usr/lib/python2.7/dist-packages/maascli/__main__.py root node: error: argument COMMAND: invalid choice: u'set-storage-layout' (choose from 'power-parameters', 'commission', 'read', 'stop', 'update', 'claim-sticky-ip-address', 'start', 'mark-broken', 'abort-operation', 'details', 'mark-fixed', 'release-sticky-ip-address', 'release', 'query-power-state', 'delete')

Am I missing something?

[Blake Rouse (blake-rouse)]

You need to be using MAAS 1.9, also make sure that your client definition of the API endpoint is up-to-date. You can do that with a "maas refresh" command.

[Sagar Shukla (sa-shukla)]

is MAAS 1.9 in stable state for deployment? I am managing 70 servers through MAAS, so just wanted to ensure that I do not loose them.

[Sagar Shukla (sa-shukla)]

Looking at 1.8.2 and 1.9 differences, 1.8.2 has default storage layout of flat which is in accordance with the requirement for CentOS deployment, so not sure if this is really required with MAAS 1.8.2 . Issue with MAAS 1.8.2 for CentOS deployment seems to be very much different.

Has someone tried CentOS deployment with MAAS 1.8.2

[Blake Rouse (blake-rouse)]

All version of 1.8.X use flat layout. Only way to use LVM is to upgrade to 1.9. I would not recommend an upgrade at this time as 1.9 is still in alpha.

[Sagar Shukla (sa-shukla)]

I am running into issues with CentOS deployment and on IRC chat, I was recommended to refer to this bug. But this issue appears to be different.

Do you have any suggestions for the issue that I am facing? I have logged bug given below to track my issue:
https://bugs.launchpad.net/maas/+bug/1502839

[Sagar Shukla (sa-shukla)]

okay ... no worries, I found that issue is with function pick_cluster_controller_address() in /usr/lib/python2.7/dist-packages/maasserver/preseed.py file. I have not yet been able to identify exact fix, but found a workaround in the interim.