Tag: linux

runc, crun, and youki all run your containers, but they differ wildly in startup speed, language, and cgroups v2 support. Here's which to pick.

iptables is in soft retirement. Here's why the nft shim isn't a long-term plan and what direct nftables syntax actually buys you.

Use age + age-plugin-yubikey to encrypt files with a key that literally cannot leave your hardware token. No GPG drama required.

systemd-resolved does way more than mangle /etc/resolv.conf. Here's how its stub resolver, split DNS, DoT, DNSSEC, and resolvectl actually work.

Atuin replaces your broken $HISTFILE with a SQLite DB, fzf-style search, and optional E2E-encrypted sync across every machine you own.

io_uring slashes syscall overhead vs epoll and AIO. Covers the submission/completion queue model, liburing examples, security caveats, and when to skip it.

The kernel docs still warn about it. The Btrfs wiki still warns about it. People still build it and lose data. Here's the write hole, why RAID 1/10 are fine, and what to use instead.

SnapRAID calculates parity on a schedule instead of in real time, so it mixes drive sizes and pairs with MergerFS. Best fit for media that rarely changes.

Creating a RAID array is the easy part. The drama starts six months later when one drive starts dying. Here's the survival guide for everything that happens after mdadm --create.

Hardware RAID controllers used to be necessary. Now they're often a liability. Here's when to flash that LSI card to IT mode, when to keep it, and why ZFS specifically refuses to share.

RAID-Z isn't just RAID 5 with a new name. Variable stripe widths, no write hole, and dRAID's distributed spares change the rebuild math entirely.

RAID 5 and 6 don't scale gracefully past 8 drives. RAID 50 and RAID 60 stripe across multiple parity sub-arrays so rebuilds only stress one group. Here's when nesting actually pays off.