qgpureset(1B)

📦 version 6.1.3, in epel 9, nixos 26.05, nixpkgs unstable

💅 qgpureset — the PBS reset-GPU-error-counts girlie ✨

okay so real talk, this is qgpureset(1B) from the PBS manual, and she's here to clear out your Nvidia GPU's emotional baggage (the ECC error counts, bestie). let's get into it 🫶

✨ NAME ✨

qgpureset — reset GPU error counts. that's it, that's the whole vibe. 💅

✨ SYNOPSIS ✨

qgpureset -H host -g gpuid -p -v

look at her, all flags, no notes. 🔥

✨ DESCRIPTION ✨

so here's the tea: the qgpureset command will request a MOM to reset the ECC counts on one of its Nvidia GPUs. yep, the MOM — pbs_mom, mother superior of your compute node. she's the one doing the actual resetting. 🧚

the GPU's error count gets reset by sending a GPU Control batch request to the batch server. very demure, very batch-request of her. ✨

💀 important warning, no cap: changing the GPU mode requires PBS Operator or Manager privilege. and it also requires that Torque be configured with --enable-nvidia-gpu. if that config flag isn't there, this whole thing is literally not happening. periodt.

✨ OPTIONS ✨

  • -H host — specifies the host within the cluster on which the GPU is located. the argument is the name of a host that is a member of the cluster of hosts managed by the server. like, an actual member, not a random name you made up. 🌸
  • -g gpuid — specifies the ID of the GPU. which one, bestie? give her the ID. 🎀
  • -p — specifies to reset the GPU's permanent ECC error count. she's holding onto that one forever, so this is the reset for the long-term memory. 👑
  • -v — specifies to reset the GPU's volatile ECC error count. the short-term, in-the-moment one. gone, slayed, cleared. 🦋

✨ OPERANDS ✨

None. zero. nada. she doesn't take operands, only flags. 🙈

✨ STANDARD ERROR ✨

the qgpureset command will write diagnostic messages to standard error for each error occurrence. so like, if something goes wrong, you'll hear about it. loud and clear. 😭

✨ EXIT STATUS ✨

  • upon successful processing of all the operands presented to the qgpureset command, the exit status will be a value of zero. 🟢 iconic, flawless, zero.
  • if the qgpureset command fails to process any operand, the command exits with a value greater than zero. 🔴 not the vibe.

✨ SEE ALSO ✨

pbs_mom(8B) and pbs_server(8B) — go read those too, bestie, they're the supporting cast. 💜

sign-off time 🎀 so this one's from the Local source/version, and it lives on the qgpureset(1B) page. that's the whole manual, you're now officially that girl who knows how to reset GPU error counts. slay, go clear those ECC counts, and remember: permanent and volatile are two different resets, iykyk. mwah 💋🌺

📦 where this page comes from