gflow 0.4.16

A lightweight, single-node job scheduler written in Rust.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159

use gflow::core::info::SchedulerInfo;
use gflow::core::job::{Job, JobState};
use gflow::core::reservation::{GpuReservation, ReservationStatus};
use std::collections::BTreeMap;
use std::time::{SystemTime, UNIX_EPOCH};

use super::schemas::{GpuAvailabilityOutput, QueuePressureGroupOutput, QueuePressureOutput};

#[derive(Debug, Default)]
struct QueuePressureGroupAccumulator {
    queued: usize,
    running: usize,
    requested_gpus: u32,
}

pub(super) fn build_queue_pressure_output(
    info: SchedulerInfo,
    jobs: Vec<Job>,
    reservations: Vec<GpuReservation>,
) -> QueuePressureOutput {
    let generated_at = SystemTime::now()
        .duration_since(UNIX_EPOCH)
        .map(|duration| duration.as_secs())
        .unwrap_or(0);
    let available_gpus = info
        .gpus
        .iter()
        .filter(|gpu| gpu.available)
        .map(|gpu| gpu.index)
        .collect::<Vec<_>>();
    let unavailable_gpus = info
        .gpus
        .iter()
        .filter(|gpu| !gpu.available)
        .map(|gpu| GpuAvailabilityOutput {
            index: gpu.index,
            reason: gpu.reason.clone(),
        })
        .collect::<Vec<_>>();

    let mut running_jobs = 0usize;
    let mut queued_jobs = 0usize;
    let mut held_jobs = 0usize;
    let mut queued_requested_gpus = 0u32;
    let mut running_allocated_gpus = 0u32;
    let mut blocked_reasons = BTreeMap::new();
    let mut users = BTreeMap::<String, QueuePressureGroupAccumulator>::new();
    let mut projects = BTreeMap::<String, QueuePressureGroupAccumulator>::new();

    for job in &jobs {
        match job.state {
            JobState::Running => {
                running_jobs += 1;
                running_allocated_gpus += job
                    .gpu_ids
                    .as_ref()
                    .map(|ids| ids.len() as u32)
                    .unwrap_or(job.gpus);
                accumulate_queue_group(&mut users, job.submitted_by.as_ref(), job, false);
                if let Some(project) = &job.project {
                    accumulate_queue_group(&mut projects, project.as_ref(), job, false);
                }
            }
            JobState::Queued => {
                queued_jobs += 1;
                queued_requested_gpus = queued_requested_gpus.saturating_add(job.gpus);
                *blocked_reasons.entry(job_reason_label(job)).or_insert(0) += 1;
                accumulate_queue_group(&mut users, job.submitted_by.as_ref(), job, true);
                if let Some(project) = &job.project {
                    accumulate_queue_group(&mut projects, project.as_ref(), job, true);
                }
            }
            JobState::Hold => {
                held_jobs += 1;
                *blocked_reasons.entry(job_reason_label(job)).or_insert(0) += 1;
                accumulate_queue_group(&mut users, job.submitted_by.as_ref(), job, true);
                if let Some(project) = &job.project {
                    accumulate_queue_group(&mut projects, project.as_ref(), job, true);
                }
            }
            _ => {}
        }
    }

    let now = SystemTime::now();
    let reservations_active = reservations
        .iter()
        .filter(|reservation| {
            reservation.status == ReservationStatus::Active || reservation.is_active(now)
        })
        .count();

    QueuePressureOutput {
        generated_at,
        total_gpus: info.gpus.len(),
        available_gpus,
        unavailable_gpus,
        running_jobs,
        queued_jobs,
        held_jobs,
        queued_requested_gpus,
        running_allocated_gpus,
        blocked_reasons,
        users: queue_group_outputs(users),
        projects: queue_group_outputs(projects),
        reservations_total: reservations.len(),
        reservations_active,
    }
}

fn accumulate_queue_group(
    groups: &mut BTreeMap<String, QueuePressureGroupAccumulator>,
    name: &str,
    job: &Job,
    queued: bool,
) {
    let group = groups.entry(name.to_string()).or_default();
    if queued {
        group.queued += 1;
        group.requested_gpus = group.requested_gpus.saturating_add(job.gpus);
    } else {
        group.running += 1;
    }
}

fn queue_group_outputs(
    groups: BTreeMap<String, QueuePressureGroupAccumulator>,
) -> Vec<QueuePressureGroupOutput> {
    let mut outputs = groups
        .into_iter()
        .map(|(name, group)| QueuePressureGroupOutput {
            name,
            queued: group.queued,
            running: group.running,
            requested_gpus: group.requested_gpus,
        })
        .collect::<Vec<_>>();
    outputs.sort_by(|left, right| {
        right
            .requested_gpus
            .cmp(&left.requested_gpus)
            .then_with(|| right.queued.cmp(&left.queued))
            .then_with(|| left.name.cmp(&right.name))
    });
    outputs.truncate(10);
    outputs
}

fn job_reason_label(job: &Job) -> String {
    if let Some(reason) = job.reason.as_deref() {
        return reason.to_string();
    }

    match job.state {
        JobState::Hold => "JobHeldUser".to_string(),
        JobState::Queued => "Resources".to_string(),
        _ => "unknown".to_string(),
    }
}