
There was a sea turtle at an aquarium that I visited with a 3d printed shell, so why not this?
I’d prefer to use the confiscated tusks to beat the poachers with. After that, they should give them back.
There was a sea turtle at an aquarium that I visited with a 3d printed shell, so why not this?
I’d prefer to use the confiscated tusks to beat the poachers with. After that, they should give them back.
I am interested in compression. I may give it a try when I swap out my desktop system. I did try btrfs in it’s early, post alpha stage, but found that the support was not ready yet. I think I had a VM system that complained. It is older now and more mature and maybe it’s worth another look.
Those are some good points. I guess I was thinking about the hardware. At least where I do RAID, it’s on the controller, so that offloads much of the parity checking and such to the controller and not the CPU. It’s all probably negligible for the apps that I run, but my hardware is quite old, so maybe trying to squeeze all the performance I can is a worthwhile activity.
Generally, if a lower level can do a thing, I prefer to have the lower level do it. It’s not really a reason, just a rule of thumb. I like to think that the lower level is more efficient to do the thing.
I use LVM snapshots to do my backups. I don’t have any other reason for it.
That all being said, I’m using btrfs on one system and if I really like it, I may migrate to it. It does seem a whole lot simpler to have one thing to learn than all the layers.
I’ve got raid 6 at the base level and LVM for partitioning and ext4 filesystem for a k8s setup. Based on this, btrfs doesn’t provide me with any advantages that I don’t already have at a lower level.
Additionaly, for my system, btrfs uses more bits per file or something such that I was running out of disk space vs ext4. Yeah, I can go buy more disks, but I like to think that I’m running at peak efficiency, using all the bits, with no waste.
That is the plan. Imagine an app that can provide personalized pricing to extract just less than the amount that would cause you to go elsewhere?
It knows when you get paid and can splurge. It knows when you are drunk or high and have less self control. It’s the digital pricing tags at the grocery store, but personalized to you (and not with your best interests in mind).
Perhaps I’ve been naieve.
I have local incremental backups and rsync to the remote. Doesn’t syncthing have incremental also? You have a good point about syncing a destroyed disk to your offsite backup. I know S3 has some sort of protection, but haven’t played with it.
I have tailscale mostly set up. What’s the issue with USB drives? I’ve got a raspberry pi on the other end with a RO SD card so it won’t go bad.
This reminds me that I need alerts monitoring set up. ; -)
I’ll have to check this out.
I attended some LUGs before covid and could see something like this being facilitated there. It also reminds me of the Reddit meetups that I never partook in.
That’s something that I hadn’t considered!
I wasn’t aware of the untrusted setting. That sounds like a good option.
Yes. It’s the “put a copy somewhere else” that I’m trying to solve for without a lot of cost and effort. So far, having a remote copy at a relative’s is good for being off site and cost, but the amount of time to support it has been less than ideal since the Pi will sometimes become unresponsive for unknown reasons and getting the family member to reboot it “is too hard”.
Take some time and really analyze your threat model. There are different solutions for each of them. For example, protecting against a friend swiping the drives may be as simple as LUKS on the drive and a USB key with the unlock keys. Another poster suggested leaving the backup computer wide open but encrypting the files that you back up with symmetric or asymmetric, based on your needs. If you’re hiding it from the government, check your local laws. You may be guilty until proven innocent in which case you need “plausible deniability” of what’s on the drive. That’s a different solution. Are you dealing with a well funded nation-state adversary? Maybe keying in the password isn’t such a bad idea.
I’m using LUKS with mandos on a raspberry PI. I back up to a Pi at a friend’s house over TailScale where the disk is wide open, but Duplicity will encrypt the backup file. My threat model is a run of the mill thief swiping the computers and script kiddies hacking in.
You’re doing God’s work!
Over my career, it’s sad to see how the technical communications groups are the first to get cut because “developers should document their own code”. No, most can’t. Also, the lack of good documentation leads to churn in other areas. It’s difficult to measure it, but for those in the know, it’s painfully obvious.
I feel like everyone's missing the point. Even 20 minutes a week is almost a day a year of your life sitting at a charger. I fill up my gas tank once a week and it takes maybe 5 minutes which is 4 hours a year that I spend feeding my car, staring at the stupid advertisements for a bacon-egg-and-cheese cinnamon roll covered in maple syrup or whatever other impulse items lie within the gas station. 5 minutes isn't enough to do anything whereas if I plan for 20 minutes, I'm going to go get a tea or something.
On the other hand, something we can all agree is a waste of time is, "how many hours of your life have been/will be spent sitting at a traffic light?"
I had one from Sony a long time ago. It even had a cable you could attach between two of 'em (600 CDs!) so that it could seamlessly start playing another track while loading the next song. I dropped it during a move and the next time I opened the door, it spit gears at me. I had intended to fix it some day, but started watching Hoarders and decided it wasn't worth it.
I’ve been trying to learn K8s and more recently the Gateway API. The struggles are that most Helm charts don’t know Gateway (most are barely Ingressroute) and I’m trying to find a solution to one service affecting the other gateways.when a service cannot find a pod, the httproute fails and when one route fails, the ingress fails. It’s a weird cascading problem.
Right now, I’m considering adding a secondary service to each gateway that resolves to a static error page. I haven’t looked into it yet; it cane to me in the brief moment of clarity before I fell asleep last night.
Also, I may be doing everything wrong, but I am learning and learning is fun.