#libreops — Public Fediverse posts

homelabs
are both

an exercise
in digital archeology

and

writing prophetic
self-fulfilling
science fiction

#libreOps

homelabs
are both

an exercise
in digital archeology

and

writing prophetic
self-fulfilling
science fiction

#libreOps

homelabs
are both

an exercise
in digital archeology

and

writing prophetic
self-fulfilling
science fiction

#libreOps

homelabs
are both

an exercise
in digital archeology

and

writing prophetic
self-fulfilling
science fiction

#libreOps

homelabs
are both

an exercise
in digital archeology

and

writing prophetic
self-fulfilling
science fiction

#libreOps

Accidentally double queuing and attempting to use 150% of available CPUs, crashing the build system?

It's more likely than you think!

Thankfully, with this one simple config option, you can ensure your heavy jobs wait their turn!

https://laminar.ohwg.net/docs.html#Contexts

#LibreOps

Build failed after 2 hours, 43 minutes...but manually re-running make in the same context (after installing a missing package) finished the build in about a minute.

Re-running in a fresh workspace just to validate that it works...

...but also, was able to manually determine that it's an issue with lacking a module, rather than the off-by-one guile version that I thought in the first place...

...so, useful foundations, but not...the solution for my "no code for module" exception.

#LibreOps

Context, this is the single-board computer that acts as the development build orchestrator. It's got a couple gigs of RAM and a 4-core with aarch64 with one core pretty much pegged.

Using #Laminar for these #LibreOps, it's my go-to for an open source build system.

Context, this is the single-board computer that acts as the development build orchestrator. It's got a couple gigs of RAM and a 4-core with aarch64 with one core pretty much pegged.

Using #Laminar for these #LibreOps, it's my go-to for an open source build system.

Context, this is the single-board computer that acts as the development build orchestrator. It's got a couple gigs of RAM and a 4-core with aarch64 with one core pretty much pegged.

Using #Laminar for these #LibreOps, it's my go-to for an open source build system.

Context, this is the single-board computer that acts as the development build orchestrator. It's got a couple gigs of RAM and a 4-core with aarch64 with one core pretty much pegged.

Using #Laminar for these #LibreOps, it's my go-to for an open source build system.

In other news, between yesterday and today I have:
- unloaded 1.27 metric tons of stuff in front of server rack
- pulled 4 blades & 7 other
- 200kg of UPS
- disassembled, carried up flight of stairs, and re-assembled the rack plus air removal unit
- mildly staged the essentials (switch, staging blade, deployment SBC) in the rack

Not done, but wiped out. Gonna be another day or three before I re-weave all the cabling, I doubt I'll do much more than rest my poor aching body tomorrow.

#libreOps

Yo #guix knowers, any idea why `guix describe` shows different results for the root and other accounts?

Asking because "https://git.savannah.gnu.org..." is deprecated/depricating, and after setting the new URL in the system-level channel file, it's fixed for root but not for the other accounts.

#libreOps

Yo #guix knowers, any idea why `guix describe` shows different results for the root and other accounts?

Asking because "https://git.savannah.gnu.org..." is deprecated/depricating, and after setting the new URL in the system-level channel file, it's fixed for root but not for the other accounts.

#libreOps

Yo #guix knowers, any idea why `guix describe` shows different results for the root and other accounts?

Asking because "https://git.savannah.gnu.org..." is deprecated/depricating, and after setting the new URL in the system-level channel file, it's fixed for root but not for the other accounts.

#libreOps

Content type issue has been resolved, and we have signs of life from the API...this is excellent, it means I can start releasing code.

Well, sort of. I need to finish the auth part first, but that's relatively simple compared to what we just did.

Here's to #lisp and #guix for making this possible! #libreOps

Content type issue has been resolved, and we have signs of life from the API...this is excellent, it means I can start releasing code.

Well, sort of. I need to finish the auth part first, but that's relatively simple compared to what we just did.

Here's to #lisp and #guix for making this possible! #libreOps

Content type issue has been resolved, and we have signs of life from the API...this is excellent, it means I can start releasing code.

Well, sort of. I need to finish the auth part first, but that's relatively simple compared to what we just did.

Here's to #lisp and #guix for making this possible! #libreOps

Content type issue has been resolved, and we have signs of life from the API...this is excellent, it means I can start releasing code.

Well, sort of. I need to finish the auth part first, but that's relatively simple compared to what we just did.

Here's to #lisp and #guix for making this possible! #libreOps

Content type issue has been resolved, and we have signs of life from the API...this is excellent, it means I can start releasing code.

Well, sort of. I need to finish the auth part first, but that's relatively simple compared to what we just did.

Here's to #lisp and #guix for making this possible! #libreOps

This is how Linux CIFS driver'sidsfromsid works, "discovered" from the source code: when a Samba server finds a file owned by a Unix user or group that it doesn't know about, Samba (and Windows) server uses two special SIDs to represent unmapped Unix accounts: S-1-22-1-uid (S-1-5-88-1-uid) and S-1-22-2-gid (S-1-5-88-2-gid). When Linux's cifs.ko sees them, it automatically maps them back to Unix accounts (same UID/GID as server) on the local client machine. But it only works if the fileshare is owned by someone on Unix that Samba doesn't know - if Samba already knows it, Samba maps them to a real SID in the ACL, so idsfromsid won't work. How is it supposed to be properly used remains a mystery... https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=975221eca5fbfdb4b6b1d17c9e540d4d7627ce18 #libreops

This is how Linux CIFS driver'sidsfromsid works, "discovered" from the source code: when a Samba server finds a file owned by a Unix user or group that it doesn't know about, Samba (and Windows) server uses two special SIDs to represent unmapped Unix accounts: S-1-22-1-uid (S-1-5-88-1-uid) and S-1-22-2-gid (S-1-5-88-2-gid). When Linux's cifs.ko sees them, it automatically maps them back to Unix accounts (same UID/GID as server) on the local client machine. But it only works if the fileshare is owned by someone on Unix that Samba doesn't know - if Samba already knows it, Samba maps them to a real SID in the ACL, so idsfromsid won't work. How is it supposed to be properly used remains a mystery... https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=975221eca5fbfdb4b6b1d17c9e540d4d7627ce18 #libreops

This is how Linux CIFS driver'sidsfromsid works, "discovered" from the source code: when a Samba server finds a file owned by a Unix user or group that it doesn't know about, Samba (and Windows) server uses two special SIDs to represent unmapped Unix accounts: S-1-22-1-uid (S-1-5-88-1-uid) and S-1-22-2-gid (S-1-5-88-2-gid). When Linux's cifs.ko sees them, it automatically maps them back to Unix accounts (same UID/GID as server) on the local client machine. But it only works if the fileshare is owned by someone on Unix that Samba doesn't know - if Samba already knows it, Samba maps them to a real SID in the ACL, so idsfromsid won't work. How is it supposed to be properly used remains a mystery... https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=975221eca5fbfdb4b6b1d17c9e540d4d7627ce18 #libreops

This is how Linux CIFS driver'sidsfromsid works, "discovered" from the source code: when a Samba server finds a file owned by a Unix user or group that it doesn't know about, Samba (and Windows) server uses two special SIDs to represent unmapped Unix accounts: S-1-22-1-uid (S-1-5-88-1-uid) and S-1-22-2-gid (S-1-5-88-2-gid). When Linux's cifs.ko sees them, it automatically maps them back to Unix accounts (same UID/GID as server) on the local client machine. But it only works if the fileshare is owned by someone on Unix that Samba doesn't know - if Samba already knows it, Samba maps them to a real SID in the ACL, so idsfromsid won't work. How is it supposed to be properly used remains a mystery... https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=975221eca5fbfdb4b6b1d17c9e540d4d7627ce18 #libreops

The Unix permission mapping features (cifsacl, idsfromsid, modefromsid) provided by Linux's CIFS driver for SMB v2+ are completely undocumented! In other words, nobody on the Internet (other than a few Samba and Microsoft Azure developers) knows how to mount a modern (non SMBv1) file share on Linux with proper Unix permissions. Wonderful! /s #libreops

The Unix permission mapping features (cifsacl, idsfromsid, modefromsid) provided by Linux's CIFS driver for SMB v2+ are completely undocumented! In other words, nobody on the Internet (other than a few Samba and Microsoft Azure developers) knows how to mount a modern (non SMBv1) file share on Linux with proper Unix permissions. Wonderful! /s #libreops

The Unix permission mapping features (cifsacl, idsfromsid, modefromsid) provided by Linux's CIFS driver for SMB v2+ are completely undocumented! In other words, nobody on the Internet (other than a few Samba and Microsoft Azure developers) knows how to mount a modern (non SMBv1) file share on Linux with proper Unix permissions. Wonderful! /s #libreops

The Unix permission mapping features (cifsacl, idsfromsid, modefromsid) provided by Linux's CIFS driver for SMB v2+ are completely undocumented! In other words, nobody on the Internet (other than a few Samba and Microsoft Azure developers) knows how to mount a modern (non SMBv1) file share on Linux with proper Unix permissions. Wonderful! /s #libreops

Otro conjunto de aplicaciones y servicios (DNS) con buena pinta -> https://libreops.cc/

#libreops

Otro conjunto de aplicaciones y servicios (DNS) con buena pinta -> https://libreops.cc/

#libreops

Otro conjunto de aplicaciones y servicios (DNS) con buena pinta -> https://libreops.cc/

#libreops

Όλα καλά Λιμπράκια μου? Τι πάθαμε σήμερα και ήμασταν κάτω?

#libreops

Όλα καλά Λιμπράκια μου? Τι πάθαμε σήμερα και ήμασταν κάτω?

#libreops

Όλα καλά Λιμπράκια μου? Τι πάθαμε σήμερα και ήμασταν κάτω?

#libreops

Όλα καλά Λιμπράκια μου? Τι πάθαμε σήμερα και ήμασταν κάτω?

#libreops

Are any of my mutuals managing grafana dashboards via IAC? We adopted lib-grafonnet a few months prior to the project going unmaintained, and now its deprecated and generating dashboard code that is non-complaint with current versions.

This is the type of tech-debt nobody wants to carry, so how have you solved it?

In an ideal world, grafana would publish a sanctioned API library that makes this a non-issue as they would maintain it.

What are y'all doing?

#LibreOps #Observability #BoostOK

"Who needs a space heater when you can run FP32 FMA stress test on CPU 0, CPU 1, GPU 0 and GPU 1..." Repost with obligatory wattmeter reading added - great for the winter! #libreops

"Who needs a space heater when you can run FP32 FMA stress test on CPU 0, CPU 1, GPU 0 and GPU 1..." Repost with obligatory wattmeter reading added - great for the winter! #libreops

"Who needs a space heater when you can run FP32 FMA stress test on CPU 0, CPU 1, GPU 0 and GPU 1..." Repost with obligatory wattmeter reading added - great for the winter! #libreops

"Who needs a space heater when you can run FP32 FMA stress test on CPU 0, CPU 1, GPU 0 and GPU 1..." Repost with obligatory wattmeter reading added - great for the winter! #libreops

New test bench is up and running. No need to worry about massive AWS server bills again. #libreops

New test bench is up and running. No need to worry about massive AWS server bills again. #libreops

New test bench is up and running. No need to worry about massive AWS server bills again. #libreops

New test bench is up and running. No need to worry about massive AWS server bills again. #libreops

Super janky watercooling installed. "Who needs a space heater when you can run FP32 FMA stress test on CPU 0, CPU 1, GPU 0 and GPU 1... #libreops

Super janky watercooling installed. "Who needs a space heater when you can run FP32 FMA stress test on CPU 0, CPU 1, GPU 0 and GPU 1... #libreops

Super janky watercooling installed. "Who needs a space heater when you can run FP32 FMA stress test on CPU 0, CPU 1, GPU 0 and GPU 1... #libreops

Super janky watercooling installed. "Who needs a space heater when you can run FP32 FMA stress test on CPU 0, CPU 1, GPU 0 and GPU 1... #libreops

Dual Xeon E5-2680 v4 arrived. Architecture: Broadwell-EP. Core count: 14C/28T x2. Price: $10 x 2. System cost: $200. Performance is not great from today's point of view but still a reasonable test bench to develop for multi-core scaling and NUMA... #libreops

Dual Xeon E5-2680 v4 arrived. Architecture: Broadwell-EP. Core count: 14C/28T x2. Price: $10 x 2. System cost: $200. Performance is not great from today's point of view but still a reasonable test bench to develop for multi-core scaling and NUMA... #libreops

Just ordered a dual-socket Intel Xeon E5-2680 v4 system instead of the AMD Zen 2 EPYC as originally planned. This aging platform has poor performance compared to the newer Zen 2, but it's boring in a good way. For $200, you get a machine representative of a typical NUMA server of the mid-2010s without the fear of forgetting to power your AWS instance off. It has a set of well-documented and known-good hardware DRAM and QPI performance counters ( https://github.com/RRZE-HPC/likwid/wiki/BroadwellEP ). Meanwhile very little is known about Zen 2. ( https://github.com/RRZE-HPC/likwid/wiki/Zen2 ). A case of "develop on Intel and deploy on AMD." If it works well on Broadwell-E, it probably should work well on everything else... #libreops

@libreops
Από που μπορώ να αποκτήσω / παραγγείλω αυτοκόλλητα;

#LibreOpsgr #libreops