Rongsong Shen
a8fb768526
night_mem_pct (resname: mem, percent: true) had value: 50, but xce_limit_val_t.val is a 0-1 fraction convention when percent is set, not a 0-100 number (see bug-1831 for the full pipeline trace). As configured, its actual capacity computed as 50x (5000%) the resource's current value -- a cap that could never realistically be exceeded by any actual demand, making this limit functionally dead in any real deployment. night_mem_pct was authored directly in YAML for a docs/example config with no C-literal ancestry to cross-check against, unlike limits2 (migrated verbatim from a hand-written C literal .val = 0.2 for a documented "20% mem cap") -- it used the human-intuitive but incorrect 0-100 scale instead. Changed value: 50 to 0.5. This is a real behavior change, not just a display fix: night_mem_pct will now actually constrain the night queue's mem usage at 50% for the first time in any deployment using this config. Discovered during live manual verification of scheduler-sim-diagnostics Requirement 6 against a real deployed cluster. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01RLSNJhNoULvVcUhXEFafQC
95 lines
2.4 KiB
YAML
95 lines
2.4 KiB
YAML
---
|
|
limits:
|
|
# cap total slots for ug1 across the whole cluster
|
|
- name: ug1_slots
|
|
users:
|
|
members:
|
|
- ug1
|
|
resname: slots
|
|
value: 10
|
|
|
|
# per-user slot cap inside ug2 (each member gets up to 4, not a shared pool)
|
|
- name: ug2_slots_per_user
|
|
users:
|
|
members:
|
|
- ug2
|
|
exclude_members:
|
|
- user3
|
|
resname: slots
|
|
value: 4
|
|
per_user: true
|
|
|
|
# queue-scoped memory cap, percentage of currently-available mem resource
|
|
# (bug-1831: percent-based value is a 0-1 fraction, not a 0-100 number --
|
|
# 0.5 = 50%, matching the convention limit2/limits2's original hardcoded
|
|
# C literal established; the previous "value: 50" here silently computed
|
|
# 5000% of the resource, a cap that could never actually bind)
|
|
- name: night_mem_pct
|
|
queues:
|
|
members:
|
|
- night
|
|
resname: mem
|
|
value: 0.5
|
|
percent: true
|
|
|
|
# host-scoped cap, applied per host in hg1
|
|
- name: hg1_slots_per_host
|
|
hosts:
|
|
members:
|
|
- hg1
|
|
resname: slots
|
|
value: 20
|
|
per_host: true
|
|
|
|
# combined scope: ug1 members on the normal queue
|
|
- name: ug1_normal_slots
|
|
users:
|
|
members:
|
|
- ug1
|
|
queues:
|
|
members:
|
|
- normal
|
|
resname: slots
|
|
value: 8
|
|
|
|
# migrated from src/controller/sched.c xcesched_push_configs()'s
|
|
# hardcoded limit1_config literal (all users except rshen, shared
|
|
# 3-slot cap)
|
|
- name: limit1
|
|
users:
|
|
members:
|
|
- all
|
|
exclude_members:
|
|
- rshen
|
|
resname: slots
|
|
value: 3.0
|
|
|
|
# global cap on ug1's total mem requested across the whole cluster
|
|
# (bug-1790 regression coverage: a GLOBAL limit on a per-host resource
|
|
# OTHER than "slots" -- exercises calc_wanted_host_resval_job_totals()'s
|
|
# generalization of bug-1789's fix to any per-host resource, proven live
|
|
# rather than left as untested config)
|
|
- name: ug1_global_mem
|
|
users:
|
|
members:
|
|
- ug1
|
|
resname: mem
|
|
value: 5000000000
|
|
|
|
# migrated from src/controller/sched.c xcesched_push_configs()'s
|
|
# hardcoded limit2_config literal (rshen on the xce queue, per-user/
|
|
# per-queue/per-project/per-host 20% mem cap)
|
|
- name: limits2
|
|
per_user: true
|
|
per_queue: true
|
|
per_project: true
|
|
per_host: true
|
|
users:
|
|
members:
|
|
- rshen
|
|
queues:
|
|
members:
|
|
- xce
|
|
resname: mem
|
|
value: 0.2
|
|
percent: true
|