Setup new osgeo8 server with LXD

[robe]

As mentioned here: ​https://lists.osgeo.org/pipermail/sac/2020-November/012798.html ​https://lists.osgeo.org/pipermail/sac/2020-November/012810.html ​https://lists.osgeo.org/pipermail/sac/2020-November/012808.html

OSMF donated their two HP Proliant servers to us. OSUOSL is holding them for us until we are ready to rebuild.

The plan is to put Ubuntu 20.04 on them and LXD. For these we will use them as extra CI bots. I am hoping to also use them to house LXD VMS -- VMS would be useful for testing OS that don't really support containerization or emulating architectures.

I'm thinking things like FreeBSD and ARM architecture. LXD uses QEMU internally for hosting the VMS but you use the LXD interface for managing them just like the lighter weight LXC containers.

Also thinking these would be useful for while we are upgrading OSGeo7 and OSGeo3/4 for moving mission critical things or hosting something like OpenNebula to allow projects to provision their own servers for testing etc.

The specs as I understand it are as follows:

Stormy 01 - seems the biggest of the 2
https://web.archive.org/web/20191112203036/https://hardware.openstreetmap.org/servers/stormfly-01.openstreetmap.org/


    HP ProLiant DL360 G6
    71 GB RAM - 18 x 4GB 1333 MT/s DDR3 DIMM
    Disk Controller: HP Smart Array P410i
    Disks: 8 x 300 GB HP EG0300FAWHV
    1.64 TB RAID6 array over 8 disks
    BIOS HP version P64

Stormy 02 - Much less disk space we may want to consider replacing the disks on this one before going too far
https://web.archive.org/web/20191112203042/https://hardware.openstreetmap.org/servers/stormfly-02.openstreetmap.org/

    HP ProLiant DL360 G6
    71 GB RAM - 18 x 4GB 1333 MT/s DDR3 DIMM
    Disk Controllers: HP Smart Array P410i
    Disks: 8 x 72 GB HP DH0072FAQRD
    410.01 GB RAID6 array over 8 disks
    BIOS HP version P64

Changing to oseo8 since osgeo9 is already set up and in use

[robe]

Milestone renamed

[robe]

Move these to next contract milestone

[robe]

osgeo9 is setup and currently oonly used for weblate (has a hop and nginx container too)

[strk]

Ticket retargeted after milestone closed

[robe]

Done with setup and all done using ansible including the lvm slice for lxd - ​https://git.osgeo.org/gitea/sac/ansible-deployment/src/branch/master/deployment/roles/osgeo8

One frustration I had is the lxc remote add doesn't work with --password. I think that might be a bug as all I have read suggests it should work. But only way I could get it to work was manually add it, and when prompted for password type it in. Then delete it. At that point the client key of the server is registered on the remote server.

You can see that with:

lxc remote switch osgeo9 #gets me into osgeo9
lxc config trust list

Ideally we should be using client tokens detailed here: ​https://linuxcontainers.org/lxd/docs/master/authentication/#authentication-token

but not sure we can automate that with ansible. If we do the prompt thing, not sure how to get ansible to answer the prompt. Anyway I have a TODO there saying it doesn't work -

​https://git.osgeo.org/gitea/sac/ansible-deployment/src/branch/master/deployment/roles/lxd-host/tasks/prod_lxd_remote.yml#L15

I also have a hop setup using a new osgeo-debian-11 image I built using ansible and stored on osgeo9. I'm sure strk will be horrified with my ansible code (like the fact I just copied and pasted stuff from his ssh-server role instead of defining a playbook to build hop that had an ssh-server role), but oh well. Which is defined in this script - ​https://git.osgeo.org/gitea/sac/ansible-deployment/src/branch/master/deployment/roles/images/tasks/build-osgeo-debian-11.yml