@cojojeff also, could you please make a tarball of the /var/snap/lxd/common/lxd/database directory and send it to me by email (free.ekanayaka@canonical.com). Thanks
@stgraber ok, I’m almost sure that the commit I pointed is the one introducing the regression. Newly bootstrapped nodes currently have address 1 in the raft configuration (since that happens before clustering is enabled and doesn’t get changed afterwise), but probably during the hashicorp/raft days 0 was used the address of the bootstrap node, and that’s what that commit breaks.
Pretty much confirmed that what I described is the culprit, I can basically reproduce it.
Just a quick note that I also into this (Ubuntu 18.04 with snap LXD). I haven’t touched LXD on that system in months.
I was able to restart the daemon manually, and the socket was accessible again. So it’s not all borked it seems. I’m in no hurry, so I’ll just wait until the patch hits.
emailed.
I’m currently back online, should I proceed with the update to these three files?
Even if you are back online you will likely hit this problem again until the code fix is deployed on your system. The new snap should be available in the channel shortly and you should automatically get refreshed, just keep an eye on it to make sure you get it (snap changes lxd will tell, or you can try to kick the refresh by hand with snap refresh lxd).
@freeekanayaka @stgraber @tomp Thanks for your assistance in getting LXD back online. Great project !!! Have a wonderful day.
Jeff Joseph (cojojeff)