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
|
# frozen_string_literal: true
require 'date'
# For each Viking partition, report the number of jobs pending for a long time
# due to QoS reasons.
class SlurmPendingOnQos
def initialize(collector, config)
@collector = collector
@partition_thresholds = {
nodes: 604_800,
week: 1_209_600,
month: 2_419_200,
himem: 604_800,
himem_week: 1_209_600,
gpu: 604_800,
interactive: 900,
test: 900,
preempt: 2_419_200
}
end
def raid
@partition_thresholds.each do |partition, threshold|
squeue_cmd = [
'/usr/bin/squeue',
'--format="%R,%V"',
'--noheader',
"--partition=#{partition}",
'--state=PENDING'
].join(' ')
data = `#{squeue_cmd}`.split("\n").grep(/QOS/).map do |row|
row.split(',')
end
count = data.count do |columns|
(Time.now.to_i - DateTime.parse(columns[1]).to_time.to_i) > threshold
end
@collector.report!(
'pending_on_qos',
count,
help: 'Number of jobs pending beyond a threshold for QoS reasons',
type: 'gauge',
labels: {
partition: partition.to_s
}
)
end
end
end
|
