# Per-user disk quotas for the CALC@HEI compute nodes (chacha, disco). # # Columns: # root_hard limit on / (home directories) # data_hard limit on /data (chacha) or /data/disk01 (disco) # shared_hard limit on /data/space (disco only, NFS-exported) # # Soft limits are derived at 90% of the hard limit by sync_quotas.sh. # Units: K, M, G, T. Use 0G for "no limit" only with a comment saying why. # # Apply with: nodes/scripts/sync_quotas.sh (dry run) # nodes/scripts/sync_quotas.sh --apply # # This file replaces the hand-edited /root/users_quota.list, which had not been # updated since 2025-02-28 and left 14 accounts on chacha and 3 on disco with no # quota at all. One of them reached 266G and filled / to 99%, which took # slurmctld's StateSaveLocation down with it (calc/ops/journals.md, 2026-08-20). # # The shared_hard column is NOT a flat default: /data/space carries per-user # limits that were tuned by hand over time (620G, 300G, 250G, 1000G, …). Those # are reproduced here as they were found live on 2026-08-20. Flattening them to # a single value would narrow most of them and, for the five users who had no # limit at all, land far below what they already hold. # --- defaults: 20G home, 100G local datasets; shared per-user, see above ----- abdullah.binjos 20G 100G 100G alexandre.veuthey 30G 400G 100G # interim home; /data raised, see below aydin.anbarani 20G 100G 100G cedric.travelletti 20G 100G 200G # had no shared limit, holds 117G fabrizio.sossan 20G 100G 100G nour-eddine.idomar 20G 100G 100G clement.colace 10G 20G 50G darko.petrovic 20G 100G 300G dion.osmani 20G 200G 200G fereshteh.jafari 20G 100G 620G florian.desmons 20G 100G 100G francesco.carrino 20G 100G 100G francisco.ribeiro 10G 150G 150G francois.butikofer 10G 20G 50G glorymary.givi 20G 100G 200G # TrunX simulations; shared quota raised from 100G (2026-09-16) gregoire.castelain 20G 20G 250G gregory.mermoud 20G 100G 100G hiten.goel 20G 100G 100G jaime.barrancohernandez 20G 100G 100G laura.belli 50G 100G 100G # 42G live on disco, see note below laura.finarelli 20G 100G 100G loic.azzalini 30G 1200G 2900G # /data 1 TB usable, agreed 2026-08-20 loic.christen 20G 300G 100G # /data raised, see note below louis.heredero 20G 100G 100G louis.lettry 20G 100G 100G lucien.troillet 20G 100G 100G marc.gillioz 30G 100G 100G # 25G live on disco, see note below marta.rende 20G 100G 100G martin.tavernier 20G 100G 100G mattechlab 20G 100G 100G mauro.leidi 20G 100G 100G nachiket.chondhikar 20G 100G 100G pmudry 20G 100G 100G remi 20G 100G 100G roland.willa 20G 20G 120G sebastien.borloz 10G 20G 20G simon.duerr 20G 100G 1000G # 51G on disco disk01, see note below steve.devenes 20G 100G 100G theophile.gentilhomme 20G 100G 100G yannic.delisle 20G 100G 100G yannick.bovier 20G 100G 100G yiwei.jia 20G 300G 100G # /data raised, see note below # Raised above the 20G default because the account is already over it. Drop to # the default once the listed reclaimable data is gone: # alexandre.veuthey 27G on chacha, of which 15G is .cache/uv (prunable) # laura.belli 42G on disco # marc.gillioz 25G on disco # # loic.azzalini's /data is 1200G, not 1000G, on purpose: the soft limit is # derived at 90% of the hard one, so a 1T hard limit would put the soft limit # at 922G — under the 986G he already holds — and block his writes when the # grace expired. 1200G leaves the agreed 1 TB usable below the warning. # # His 2900G on /data/space is a different allocation, not part of that # agreement: he holds 2277G there and had no limit at all. The number records # what is already on disk (usage +25%, rounded up) so the backfill does not # break him — it is not an approved allowance. /data/space is 7T with 2.3T # free and he is 53% of what is used. Needs its own conversation, see # calc/ops/todo.md#disk-quotas-2026-08-20. # # alexandre.veuthey, loic.christen and yiwei.jia were over the 100G /data # default and under grace from 2026-08-20. Raised to their usage +25% rounded # up (280G->400G, 237G->300G, 201G->300G) rather than asking them to delete # 418G between them: chacha /data dropped to 77% when martin.barry's data went, # so the 100G figure -- a flat default carried over from the 2025 list, set # when /data was at 97% -- no longer had scarcity behind it. Not a policy # decision on what a fair /data allowance is; that is still open in # calc/ops/todo.md#disk-quotas-2026-08-20. # # data_hard is one column for two different filesystems — /data on chacha and # /data/disk01 on disco — whose per-user limits drifted apart while they were # set by hand. simon.duerr is the case that showed it: 20G on chacha, 1000G on # disco. His 100G here clears the 51G he holds on disco without loosening # chacha, where he stores nothing, but the conflation stands and the next # divergence will need per-host columns. See # calc/ops/todo.md#disk-quotas-2026-08-20. # # Not listed, deliberately: # guillaume.bessard live UID unconfirmed (calc/ops/todo.md), keeps his # hand-set 10G/20G until it is # pamela.delgado in the 2025 list, no longer has an account # martin.barry departed; access revoked, roster and SLURM entries # already gone, ~907G of data deleted 2026-08-20 # (calc/ops/journals.md) # kevin.voisin, marco.caporizzi, jeremy.duc, olivier.amacker, # alexandre.venturi decommissioned 2026-08-11, homes still present on # chacha at ~0 bytes (calc/ops/todo.md)