ChipTuner's avatar
ChipTuner
ChipTuner@gitcitadel.com
npub1qdjn...fqm7
Building software they don't like. Free, as in freedom. Low-level and server engineer: libnoscrypt, NVault, vnlib. Staff @GitCitadel https://geyser.fund/project/gitcitadel
ChipTuner's avatar
ChipTuner 1 month ago
Nostr: where devs get to tell everyone how they reinvented the wheel completely unchallenged.
ChipTuner's avatar
ChipTuner 1 month ago
Doing a little more maintenance than usual today for @GitCitadel and our .eu services. There might be some brief interruptions with #alexandria and our #lightning wallet. All other services such as relays will be untouched :)
ChipTuner's avatar
ChipTuner 1 month ago
If you're like me and whined for the past 15 years that file backup and archival systems suck. You definitely should consider restic (backrest is the web gui). Major wins for me are robustness. Sometimes you have to migrate storage, you lose a drive, move servers, or switch cloud storage. Restic has handled all of that well for me and has built-in tools to ensure said robustness. Compression and de-duplication are insane, right now I have multiple 50gb+ repos that take up less than 2gb on disk. Need to move a repo? Just mv or cp or rsync, no biggie. It also has over 10 storage backends, big ones for me are local disk, smb, nfs, S3 and sftp/ssh. client-side encryption is default btw, pbkh + aes256cbc I believe is the only support alg, which is fine. @Ingwie Phoenix (aka. birb) has the credit for show me backrest which has worked amazing for over a year now for me across multiple repositories, backups and servers.
ChipTuner's avatar
ChipTuner 1 month ago
Docker hub behind cloudflare is blocking my cloud IPs (personal and production server ips). Going to find out if logging into the registry will fix it. I can't be the only one having these issues right? I already mirror github projects because github chooses to tell me to GFY. I guess I'm gonna need to build EVERYTHING from source now?
ChipTuner's avatar
ChipTuner 1 month ago
I miss this setup sometimes. Also I still have that 3u supermicro chassis in storage if anyone wants it :) image
ChipTuner's avatar
ChipTuner 1 month ago
For those who preach "you can just do X". How sustainable did that turn out to be? One, two, three weeks later? A year later? Just curious. I think few want sustainable growth and progress, they want to make a rapid change and tell everyone about it. If you really want to impress yourself, you have to "just keep doing X"
ChipTuner's avatar
ChipTuner 1 month ago
I know no one listens to me, but yeah... Unless you commit to your default branch, you know the one that plebs pull from when they don't read your docs, your GitHub followers think your dead. GitHub is just social media and you should treat it as such. So when you make rapid commits just to look good, know that lazy users will probably be pull broken code. And once again the incentives are ass fucking backwards. View quoted note →
ChipTuner's avatar
ChipTuner 1 month ago
Tip for those of you running proxmox (or other corosync based systems) in a cluster (more than 2 nodes). I would suggest in adding a physical witness node if you can afford to. A witness node generally is an extra server that just participates votes in the cluster to help maintain quorum. So you don't run any critical services on it, it just serves to keep remaining nodes aware of their environment. A friend to talk to really. The reason for this is, lets say you have 1/3 nodes down (2/3 running). For maintenance or maybe a repair that will take a day or more. If you "lose" one of the two remaining nodes for literally any reason, including network hiccup or power etc. Both nodes will kill themselves and take your whole cluster down. This is called STONITH. Now as long as one of the two nodes were able to communicate with the witness, the remaining compute node can maintain quorum and will attempt to stay alive and keep it's services running, while the other node kills itself and reboots. Services will respawn on the remaining compute node then attempt to drain back to the dead compute node if it comes back online. Reducing your downtime. The idea is to have one more than you need, that's not running compute services on it. This allows it to be inexpensive and not as critical to the rest of the network. The other part is kind of psyhcological, a cheap witness will stop you from scheduling services on it, to make you think your cluster is bigger than it is. It allows you to have a balance of compute and maintenance with practical budget constraints.
ChipTuner's avatar
ChipTuner 1 month ago
On my first few days at college as 18 y/o, I got in a shouting match with the financial office desk lady/gatekeeper. The room was full of people. I gave her my info then she told me to wait so I stepped back. She then shouted across the room "SIR YOU DECLINED YOUR LOAN OFFERS FOR $12,000/year!" "I KNOW, LOANS DON"T MAKE TUITION CHEAPER" "SIR WE CAN"T HELP YOU UNTIL YOU ACCEPT YOUR LOANS" "MA'AM, I CAME TO NEGOTIATE WHAT I CAN AFFORD. I CANT AFFORD 50K IN ADDITIONAL DEBT" "WE'LL THATS GONNA BE A HARD ONE... CUZ... mmm. SIR WE CAN'T HELP YOU THEN" "MA'AM CAN YOU JUST SIT DOWN AND GET ME SOMEONE ELSE" Office manager poked head out after all the yelling, and room full of shocked and laughing people, and took me back to the office. He told me to pound sand, so I had to cut my courses and thankfully got a full time income the following semester. It was a simpler time.