It’s now the second time that I’m having this weird issue, the first time it was on a disposable environment (with a more recent version of Juju) so I just recreated it from scratch, but now it’s in a production environment so I don’t have as much luck.
I have 3 EC2 instances as machines created by Juju with 6 LXD containers (2 per servers).
When trying to connect through TCP to some AWS services (Like ElasticSearch or ElastiCache) I get no content replies from the servers, but the connection gets established successfully and it started happening after a reboot, it was working perfectly before that reboot. I do not have this issue with any other machines on the network or online. Now, 1 of those machine haven’t been rebooted yet, but is not having the issue so it makes me wonder if a recent update could have brought the issue.
Everything works fine if the connection is established directly from the machine rather than from inside the containers.
Juju version: 2.4.3
Ubuntu version on all machines and containers: 18.04
The only weird thing I’ve noticed is how they all have 30+ veth interfaces, but I fail to see a link since all 3 machines have those. Could I delete all veth interfaces safely / would LXD recreate the ones it needs?
Anyone had any similar experiences?