oar_resources_add(1)

💖 oar_resources_add — like, the OAR command that hands you resources 💖

okay bestie, welcome to the oar_resources_add(1) glow-up 💅 this one lives in the OAR commands family and it's literally your little helper script bestie that defines OAR resources for you. no cap.

---

✨ NAME ✨

oar_resource_add — yet another helper script to define OAR resources 🫶

---

🌸 SYNOPSIS 🌸

oar_resource_add -H <# of hosts> [options]

---

🦋 DESCRIPTION 🦋

so this tool generates the oarproperty and oarnodesetting commands for you, and those commands create OAR resources following the host / cpu / core (/ thread) hierarchy, possibly with GPU alongside. it's giving ✨structured✨

REMINDER: every single physical element (each cpu, each core, each thread, each gpu) has to have a unique identifier in the OAR resources database. no duplicates, periodt.

if some resources already exist in the database (like, say, from a previously installed cluster 👀), you can either hand over offsets on the command line, or let the auto-offset option guess them for you — that way the identifiers for your newly created resources stay unique. slay.

this tool is also a genuinely good example of how you can create OAR resources using script loops and the oarnodesetting command. if it doesn't exactly fit your needs, feel free to read the script code and adapt it 💜 no gatekeeping here.

---

🎀 OPTIONS 🎀

• -T, --use-threads — use (create) the thread property. threads are having a moment 💅
• -H, --hosts <#> — how many hosts, bestie. the count.
• -C, --cpus <#> — how many cpu per host.
• -c, --cores <#> — how many core per cpu.
• -t, --threads <#> — how many threads per core.
• -g, --gpus <#> — how many gpus per host.

WARNING, not all hardware configurations are supported 😭 for multi-GPU configurations, the # of cores per cpu must be a multiple of the # of gpus per cpu. do the math, bestie.

if you don't set this, no gpu resources is handled.

• --host-prefix <str> — hostname prefix. default is "node-" 💖
• --host-suffix <str> — hostname suffix (like ".domain" for example).
• --host0 <#> — first host id to use.
• --cpu0 <#> — first cpu id to use.
• --core0 <#> — first core id to use.
• --thread0 <#> — first thread id to use.
• --gpu0 <#> — first gpu id to use.
• --cpusets <#> — # of cpusets on host. mainly useful for development purposes when one wants to emulate more cores than systems really have 👀 (yes, pretending. we love a simulation moment).
• --cputopo <list|RR> — provide the topology for the CPUs/cores/threads affinity (see hwloc), given a ordered list of the Linux CPU ids (as used in cpuset.cpus).

or, given the RR value, try to match a round-robin distribution of the core ids, with optionally HyperThreads as well. 🔄

• --gpudevices <#> — # of gpudevices on host. mainly useful for development purposes when one wants to emulate more GPUs than systems really have 🎮✨
• --gputopo <list> — provide the topology for the CPUs/GPUs affinity (see hwloc), given a ordered list of the Linux GPU device ids.
• -A, --append <str> — append a text string (extra properties). little bonus content, we love.
• -o, --write-to <file> — write commands to file.
• -p, --no-create-properties — do not generate oarproperty commands.
• -a, --auto-offset — guess the next host/cpu/core/thread/gpu ids to use from the OAR database (unless host0/cpu0/core0/thread0/gpu0 is set).

WARNING: properties must already exist in your database or you will get errors 💀 like, actually. it will not be cute.

• -Y, --yaml — generate YAML output. for the yaml girlies 🧚
• -h, --help — display usage.

---

⚠️ WARNING ⚠️

the oar_resource_add tool does not look at the actual hardware topology of the target machines. so core and GPU device affinity to CPU may not be correct 🙈

see the hwloc commands for instance to find out the correct topology and affinity, and use the --cputopo and --gputopo options accordingly. trust the process, bestie.

---

🌈 SEE ALSO 🌈

oarnodesetting(1), oarproperty(1), oar_resources_init(1), bash(1), hwloc(7), lstopo(1)

---

👑 COPYRIGHTS 👑

Copyright 2003-2018 Laboratoire d'Informatique de Grenoble (http://www.liglab.fr). This software is licensed under the GNU General Public License Version 2 or above. There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

---

okay and that's the whole page 💖 huge shoutout to the oar_resources_add page from the OAR commands manual, dated 2023-07-10 — mwah, iconic, no notes. bye bestie, go create some resources 🫶✨🎀