mz_cluster_client/
metrics.rs

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
// Copyright Materialize, Inc. and contributors. All rights reserved.
//
// Use of this software is governed by the Business Source License
// included in the LICENSE file.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0.

//! Metrics shared by both compute and storage.

use std::time::Duration;

use mz_ore::metric;
use mz_ore::metrics::{
    CounterVec, DeleteOnDropCounter, DeleteOnDropGauge, GaugeVec, IntCounterVec, MetricsRegistry,
};
use mz_ore::stats::SlidingMinMax;
use prometheus::core::{AtomicF64, AtomicU64};

/// Controller metrics.
#[derive(Debug, Clone)]
pub struct ControllerMetrics {
    dataflow_wallclock_lag_seconds: GaugeVec,
    dataflow_wallclock_lag_seconds_sum: CounterVec,
    dataflow_wallclock_lag_seconds_count: IntCounterVec,
}

impl ControllerMetrics {
    /// Create a metrics instance registered into the given registry.
    pub fn new(metrics_registry: &MetricsRegistry) -> Self {
        Self {
            // The next three metrics immitate a summary metric type. The `prometheus` crate lacks
            // support for summaries, so we roll our own. Note that we also only expose the 0- and
            // the 1-quantile, i.e., minimum and maximum lag values.
            dataflow_wallclock_lag_seconds: metrics_registry.register(metric!(
                name: "mz_dataflow_wallclock_lag_seconds",
                help: "A summary of the second-by-second lag of the dataflow frontier relative \
                       to wallclock time, aggregated over the last minute.",
                var_labels: ["instance_id", "replica_id", "collection_id", "quantile"],
            )),
            dataflow_wallclock_lag_seconds_sum: metrics_registry.register(metric!(
                name: "mz_dataflow_wallclock_lag_seconds_sum",
                help: "The total sum of dataflow wallclock lag measurements.",
                var_labels: ["instance_id", "replica_id", "collection_id"],
            )),
            dataflow_wallclock_lag_seconds_count: metrics_registry.register(metric!(
                name: "mz_dataflow_wallclock_lag_seconds_count",
                help: "The total count of dataflow wallclock lag measurements.",
                var_labels: ["instance_id", "replica_id", "collection_id"],
            )),
        }
    }

    /// Return an object that tracks wallclock lag metrics for the given collection on the given
    /// cluster and replica.
    pub fn wallclock_lag_metrics(
        &self,
        collection_id: String,
        instance_id: Option<String>,
        replica_id: Option<String>,
    ) -> WallclockLagMetrics {
        let labels = vec![
            instance_id.unwrap_or_default(),
            replica_id.unwrap_or_default(),
            collection_id,
        ];

        let labels_with_quantile = |quantile: &str| {
            labels
                .iter()
                .cloned()
                .chain([quantile.to_string()])
                .collect()
        };

        let wallclock_lag_seconds_min = self
            .dataflow_wallclock_lag_seconds
            .get_delete_on_drop_metric(labels_with_quantile("0"));
        let wallclock_lag_seconds_max = self
            .dataflow_wallclock_lag_seconds
            .get_delete_on_drop_metric(labels_with_quantile("1"));
        let wallclock_lag_seconds_sum = self
            .dataflow_wallclock_lag_seconds_sum
            .get_delete_on_drop_metric(labels.clone());
        let wallclock_lag_seconds_count = self
            .dataflow_wallclock_lag_seconds_count
            .get_delete_on_drop_metric(labels);
        let wallclock_lag_minmax = SlidingMinMax::new(60);

        WallclockLagMetrics {
            wallclock_lag_seconds_min,
            wallclock_lag_seconds_max,
            wallclock_lag_seconds_sum,
            wallclock_lag_seconds_count,
            wallclock_lag_minmax,
        }
    }
}

/// Metrics tracking frontier wallclock lag for a collection.
#[derive(Debug)]
pub struct WallclockLagMetrics {
    /// Gauge tracking minimum dataflow wallclock lag.
    wallclock_lag_seconds_min: DeleteOnDropGauge<AtomicF64, Vec<String>>,
    /// Gauge tracking maximum dataflow wallclock lag.
    wallclock_lag_seconds_max: DeleteOnDropGauge<AtomicF64, Vec<String>>,
    /// Counter tracking the total sum of dataflow wallclock lag.
    wallclock_lag_seconds_sum: DeleteOnDropCounter<AtomicF64, Vec<String>>,
    /// Counter tracking the total count of dataflow wallclock lag measurements.
    wallclock_lag_seconds_count: DeleteOnDropCounter<AtomicU64, Vec<String>>,

    /// State maintaining minimum and maximum wallclock lag.
    wallclock_lag_minmax: SlidingMinMax<f32>,
}

impl WallclockLagMetrics {
    /// Observe a new wallclock lag measurement.
    pub fn observe(&mut self, lag: Duration) {
        let lag_secs = lag.as_secs_f32();

        self.wallclock_lag_minmax.add_sample(lag_secs);

        let (&min, &max) = self
            .wallclock_lag_minmax
            .get()
            .expect("just added a sample");

        self.wallclock_lag_seconds_min.set(min.into());
        self.wallclock_lag_seconds_max.set(max.into());
        self.wallclock_lag_seconds_sum.inc_by(lag_secs.into());
        self.wallclock_lag_seconds_count.inc();
    }
}