"'Using a combination of packet sniffing, #eBPF, and async-profiler we managed to identify the root cause of slow produce requests in our Kafka cluster. We then tested a couple of solutions to the problem: data=writeback journaling mode, fast commits, and changing the file system to XFS.[…] With XFS, the number of produce requests exceeding 65ms (our SLO) was lowered by 82%.'"
"'"Upstream [#Linux#kernel] deprecated the V4 format for #XFS with commit b96cb835[1]. […] #Fedora 40 will be the last release that supports these older XFS filesystems."'"
Great news for those who care ! #Grub has fixed #XFS issue after a couple of moths of it being in limbo.. Will be bringing it back next release of #XeroLinux.. YaY ! #FOSS#OpenSource#Linux
My Arch #Linux desktop crashed during an update and now has a corrupt XFS. Mounting fails because of the corruption. xfs_repair can't fix it because of unapplied log changes.
The XFS partition is on top of a pair of LVM-mirrored drives. A different XFS partition on the same drives mounts fine.
I'd rather not destroy the log and take the data loss but that's the next step I'm looking at.
Nitrux uses OpenRC instead of systemd, unlike most mainstream distros. It also features the Liquorix kernel, described as “an enthusiast Linux kernel designed for uncompromised responsiveness in interactive systems, enabling low latency in A/V production, and reduced frame time deviations in games.”...
»I burned out years ago trying to juggle the roles senior developer, reviewer, tester, triager (crappily), release manager, and (at times) manager liaison. There's enough work here in this one subsystem for a team of 20 FT, but instead we're squeezed to half that. I thought if I could hold on just a bit longer I could help to maintain the focus on long term development to improve the experience for users. I was wrong.«
People frequently say they don’t want #ZFS on their workstation because it’s a server FS. But what’s the best workstation FS? #UFS, #Ext4, #HAMMER, #Btrfs, #XFS? What’s the going benchmark workload look like? What subjective aspects matter too? (And yes, I know about damned lies of https://fsbench.filesystems.org/)
@sassageflare
My point is that entire racks full of interconnected cages full of storage, accelerators, NICs, etc. is quite common in #HPC, and #SGI basically made HPC solutions for CAD & VFX...
#Linux#kernel 6.3.5 is out, which among others fixes[1] the #XFS metadata corruption that can occur with earlier 6.3.y releases. It has been seen in the wild by a number of users and made the news.
The fix for the #Linux#kernel#XFS metadata corruption in 6.3.y that made the news has been found and is now queued for #LinuxKernel 6.3.5 [edit: rc is now out, release likely on Tue or Wed]:
「 In 1994, the paper Scalability in the XFS File System saw publication. Computers got faster since 1984, and so did storages. Notably, we are now seeing boxes with multiple CPUs, and with storages reaching into the Terabytes. The improvements to the 4.3BSD fast filing system (or the modified version in SGI IRIX called EFS) were no longer sufficient 」
— Kristian Köhntopp
Oh, CONFIG_XFS_DEBUG=y, which means: […] We randomly chose a near block allocation strategy to use to improve code coverage, not the optimal one for IO performance. Hence the CPU usage and allocation patterns that impact IO performance are simply not predictable or reproducable from run to run. So, yeah, trying to bisect […] will not be reliable....
Systemd-Free Immutable Distro Nitrux 2.9.1 Is Out Powered by Liquorix Kernel 6.4 (9to5linux.com)
Nitrux uses OpenRC instead of systemd, unlike most mainstream distros. It also features the Liquorix kernel, described as “an enthusiast Linux kernel designed for uncompromised responsiveness in interactive systems, enabling low latency in A/V production, and reduced frame time deviations in games.”...