On 08/10/2018 05:27 AM, Fajar A. Nugraha wrote:
On Thu, Aug 9, 2018 at 7:57 PM, Pierre Couderc <pie...@couderc.eu <mailto:pie...@couderc.eu>> wrote:

    On 08/09/2018 11:30 AM, Fajar A. Nugraha wrote:

    Basically you'd just need to copy /var/lib/lxd and whatever
    storage backend you use (I use zfs), and then copy them back
    later. Since I also put /var/lib/lxd on zfs (this is a custom
    setup), I simply need to export-import my pool.

    /var/lib/lxc alone, nothing about /var/lxc ?

Are you using lxc1 (e.g. lxc-create commands) or lxd?

When lxd is installed as package (e.g. installed as apt on ubuntu), you only need /var/lib/lxd and its storage pool (which will be mounted on /var/lib/lxd/storage-pools/...).

Here's what I'm using:
- I start AWS spot instance
- I have a custom ubuntu template, with lxd installed but not started. It thus has an empty /var/lib/lxd,  with no storage pools and network. - I have a separate EBS disk, used by a zfs pool 'data'. I then have 'data/lib/lxd' which I mount as '/var/lib/lxd', and 'data/lxd' which is registered as lxd storage pool 'default'.
- I create containers (using that default pool)
- if that spot instance is terminated (thus the "root"/OS disk is lost), I can simply create a new spot instance again, and attach the 'data' pool there. I will then have access to all my containers.
ok, fine

Is that similar to what you need?
Yes, this is very similar. Thank you.

Note that lxc1 and lxd from snap uses different directories than lxd from package.

Sorry for the noise : I use lxd (from sources on debian), and I had not seen that /var/lib/lxc exits but is empty...

lxc-users mailing list

Reply via email to