Under things me and my users notice aren’t working right away, at the top of the list is email. So I notice when those alerts aren’t able to get through, because if email is down I have my phone ringing off the hook because my dad can’t get to his online auctions to see if he won that toaster for $5. So email is like, the best option.

If you use Live Migrate, realize that it doesn’t work on an LXC, only VMs. Your containers will be restarted with the LXC on the new node.

I would like to play with ceph but I don’t have a lot of spare equipment anymore, and I understand ZFS pretty well, and trust it. Maybe the next cluster upgrade if I ever do another one.

And I have an almost unhealthy paranoia after see so many shitshows in my career, so having a pile of copies just helps me sleep at night. The day I have to delve into the last layer is the day I build another layer, but that hasn’t happened recently. PBS dedup is pretty damn good so it’s not much extra to keep a lot of copies.

Ah, OK. Now I get your point.

Yes, RAID 10 ZFS with no ARC, 6GB SAS drives.

I have no idea what you have going on, I’ve never seen LXCs take that long, even if I include the time it takes to down the containers and bring them up after a reboot.

What are you using for running them? I just tested my docker LXC and it took 16 seconds from when I typed “reboot” to having a login prompt. And that’s on an ancient R410 server running proxmox.

I think you’re doing it wrong. LXCs boot almost instantaneously on a hypervisor since they hijack the host kernel, I’d be surprised if my CTs take 5 seconds.

I would agree on the live migration issue but I guess you pick your services accordingly. I have a VM that runs docker and a LXC docker host, and I pick my containers for each accordingly.

The advantages you gain with running a hypervisor on something like ZFS is immeasurable, for snapshotting, replication, snapshot backups and high availability. You don’t have to quiese machines to back them up and you can do instant COW snapshots before upgrades.

KVM doesn’t really have overhead, that’s the kernel part. Maybe a bit of RAM, but with LXCs it’s negligible.

I think Mailcow is a fair bit further along in features than this. I used this for a short bit but wasn’t overly impressed, and you are right about how running a docker stack is less hassle for updating.

I followed where it was going and it was a forgejo repo where there were some action sets but not that one. I figured they were using their own sets and hadn’t gotten around to java yet.

Proxmox servers are mirrored zpools, not that RAID is a backup. Replication between Proxmox servers every 15 minutes for HA guests, hourly for less critical guests. Full backups with PBS at 5AM and 7PM, 2 sets apiece with one set that goes off site and is rotated weekly. Differential replication every day to zfs.rent. I keep 30 dailies, 12 weeklys, 24 monthly and infinite annuals.

Periodic test restores of all backups at various granularities at least monthly or whenever I’m bored or fuck something up.

Yes, former sysadmin.

Probably don’t have to pay them, they’ll do it for free.

Stick with Traefik if you’ve figured it out. It’s much more powerful than NPM in my opinion. If you insist on using NPM, you might want to try NPMPlus, it has more bells and whistles and is more actively maintained.

actions/setup-java@v4 would fail trying to find the java setup script at Forgejo’s runner source repo, and apparently it wasn’t there when I went to look. I’ll look at it another time when maybe all the backend is put together or there’s a way I can host the actions locally so I’m not relying on outside sources that might pollute my CI output.

I like that. I tried to get Actions in Forgejo working and that was a dead-end. So I’ve been using act manually.

Appreciate the writeup.

From Jon’s podcast? Wasn’t that a horrible interview? Jon starts put saying “what’s the plan instead of just saying you have to get the message out” and Jeffries doubles down on “branding”.

Its really disheartening to see that bullshit gaslighting is still the MO of the Democrats going forward. Every word of that interview was more of the same old shit, and unfortunately Jon didn’t push back on him at all.

Forgejo. There are so many things that can use a git repo but I don’t want to have them out in the wild, so I host them myself, safe and sound behind my firewall.

I also mirror other github forks so they don’t go away whenever those services decide to rugpull them.