• 0 Posts
  • 13 Comments
Joined 2 years ago
cake
Cake day: June 14th, 2023

help-circle
  • You don’t need cards to have full bandwidth, they only time it will matter is when you’re loading the models on the card. You need a motherboard with x16 slots but even x4 connections would be good enough. Running the model doesn’t need a lot of bandwidth. Remember you only load the model once then reuse it.

    An x4 pcie gen 4 slot has ~7.8 GiB/s theoretical transfer rate (after overhead), a x16 has ~31.5GiB/s - so disk I/O is likely your limit even for a x4 slot.

    • overhead was already in calculations

  • We can’t ever stop this kind of stuff, but with something like fail2ban you can set it up to block on too many failures.

    Really though - ensuring your system is kept up to date and uses strong passwords or use a SSH keys is the best defence. Blocking doesn’t prevent them from trying a few times. Moving SSH to a non standard port will stop most of the automated attacks but it won’t stop someone who is dedicated.



  • You get easy access to their addons with a VM (aka HAOS). You can do the same thing yourself but you have to do it all (creating the containers, configuring them, figuring out how to connect them to HA/your network/etc., updating them as needed) - whereas with HAOS it generally just works. If you want that control great but go in with that understanding.


  • BTRFS has RAID built into the file system - instead of using MD you use BTRFS profiles which tell the system how to handle data.

    For instance

    • file system data (critical for the file system to function): raid1c3 which means 3 copies of core P file system data on 3 different devices
    • user data: raid1 (so duplicating all your data on two different devices)

    With this set up you could lose one device (of n, the total doesn’t matter), and not lose any data, and still be able to boot to recover with too much hassle.

    BTRFS does block checksums, can scan for bit rot and recover from it, and generally tries to make your data safe. It technically supports raid5/6 for user data, the issue is around unclean shutdowns and a potential write hole where you could lose data, but if your system has a UPS backup and is on a relatively recent kernel it’s not any more dangerous than MD raid5/6 as I understand it.






  • From a Linux command line it would be the command called arp, you need to add a static arp entry. I don’t know how that works on sense, but on Linux it would be something like arp -s IP MAC

    Maybe there’s a module in opnsense to help. The way I’ve done this before is using a machine connected to the same network at my target to wake up by logging into that machine and issuing the wake command.


  • WoL packets are usually sent to the ip broadcast address for the network as they’re not ip based. I don’t know if this would ever work well across networks. Can you do send the wol packet from the opnsense router instead? Does it work then?

    If you’re sending it to the IP of the server, it likely works soon after your turn the machine off because the ARP entry hasn’t timed out yet, but once it times out it won’t work anymore. The router doesn’t know how to get to the machine. You may be able to add a static arp mapping to get it to work long term.


  • All that yes. The Wooting One (original that uses IR light) let you use buttons to simulate controller axes, change how hard you need to press to activate, and add second functions to keys. It was an interesting idea but I found the gaming part the original keyboard to be only usable in a limited set of games as it’s not as sensitive as a controller stick, and as a keyboard it wasn’t great either. Hopefully V1 problems, I know they had through another version of the IR keyboard, and then came out with the Hall effect keyboard. I like the idea but never could get used to it, and when the spacebar was loose I retired it after fixing it.