Skip to content

Worker managed cpu utilization #1073

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
Dym03 wants to merge 1 commit into
base: main
Choose a base branch
from
+182 −24
Open

Worker managed cpu utilization #1073

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
98 changes: 87 additions & 11 deletions crates/hyperqueue/src/dashboard/ui/screens/cluster/worker/cpu_util_table.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,14 +1,16 @@
use crate::common::format::human_size;
use crate::dashboard::ui::screens::cluster::worker::UtilizationRenderMode;
use ratatui::layout::{Constraint, Rect};
use ratatui::style::Style;
use ratatui::widgets::{Cell, Row, Table};
use std::cmp;
use tako::hwstats::MemoryStats;
use tako::resources::ResourceIndex;

use crate::dashboard::ui::styles;
use crate::dashboard::ui::terminal::DashboardFrame;
use crate::dashboard::ui::widgets::progressbar::{
ProgressPrintStyle, get_progress_bar_color, render_progress_bar_at,
ProgressPrintStyle, get_progress_bar_color, get_progress_bar_cpu_color, render_progress_bar_at,
};
use crate::dashboard::utils::calculate_average;

Expand All @@ -19,6 +21,8 @@ const CPU_METER_WIDTH: u8 = CPU_METER_PROGRESSBAR_WIDTH + 4;
pub fn render_cpu_util_table(
cpu_util_list: &[f64],
mem_util: &MemoryStats,
used_cpus: &[ResourceIndex],
util_render_mode: &UtilizationRenderMode,
rect: Rect,
frame: &mut DashboardFrame,
table_style: Style,
Expand All @@ -31,38 +35,50 @@ pub fn render_cpu_util_table(
let width = constraints.len();
let height = (cpu_util_list.len() as f64 / width as f64).ceil() as usize;

let mut rows: Vec<Vec<(f64, usize)>> = vec![vec![]; height];
for (position, &cpu_util) in cpu_util_list.iter().enumerate() {
let row = position % height;
rows[row].push((cpu_util, position));
let mut rows: Vec<Vec<(f64, usize, bool)>> = vec![vec![]; height];
if *util_render_mode == UtilizationRenderMode::Worker {
rows = get_utilization_sorted_by_usage(cpu_util_list, used_cpus, height)
} else {
for (position, &cpu_util) in cpu_util_list.iter().enumerate() {
let row = position % height;
let used = used_cpus.contains(&ResourceIndex::new(position as u32));
rows[row].push((cpu_util, position, used));
}
}

let rows: Vec<Row> = rows
.into_iter()
.map(|targets| {
let columns: Vec<Cell> = targets
.into_iter()
.map(|(cpu_util, position)| {
.map(|(cpu_util, position, used)| {
let progress = cpu_util / 100.00;
let style = match util_render_mode {
UtilizationRenderMode::Global => get_progress_bar_color(progress),
UtilizationRenderMode::Worker => get_progress_bar_cpu_color(progress, used),
};

Cell::from(render_progress_bar_at(
Some(format!("{position:>3} ")),
progress,
CPU_METER_PROGRESSBAR_WIDTH,
ProgressPrintStyle::default(),
))
.style(get_progress_bar_color(progress))
.style(style)
})
.collect();
Row::new(columns)
})
.collect();

let avg_cpu = calculate_average(cpu_util_list);

let mem_used = mem_util.total - mem_util.free;
let (which_util, num_cpus, avg_cpu) =
create_title_info(cpu_util_list, used_cpus, util_render_mode);

let title = styles::table_title(format!(
"Worker Utilization ({} CPUs), Avg CPU = {:.0}%, Mem = {:.0}% ({}/{})",
cpu_util_list.len(),
"{} Utilization ({} CPUs), Avg CPU = {:.0}%, Mem = {:.0}% ({}/{})",
which_util,
num_cpus,
avg_cpu,
(mem_used as f64 / mem_util.total as f64) * 100.0,
human_size(mem_used),
Expand All @@ -89,3 +105,63 @@ fn get_column_constraints(rect: Rect, num_cpus: usize) -> Vec<Constraint> {
)
.collect()
}

fn get_utilization_sorted_by_usage(
cpu_util_list: &[f64],
used_cpus: &[ResourceIndex],
height: usize,
) -> Vec<Vec<(f64, usize, bool)>> {
let mut all_cpus: Vec<(f64, usize, bool)> = cpu_util_list
.iter()
.enumerate()
.map(|(position, &cpu_util)| {
let used = used_cpus.contains(&ResourceIndex::new(position as u32));
(cpu_util, position, used)
})
.collect();

all_cpus.sort_by_key(|&(_, _, used)| std::cmp::Reverse(used));

let mut rows: Vec<Vec<(f64, usize, bool)>> = vec![vec![]; height];
for (index, cpu_data) in all_cpus.into_iter().enumerate() {
let row = index % height;
rows[row].push(cpu_data);
}
rows
}

fn create_title_info(
cpu_util_list: &[f64],
used_cpus: &[ResourceIndex],
util_render_mode: &UtilizationRenderMode,
) -> (String, usize, f64) {
let which_util = match util_render_mode {
UtilizationRenderMode::Global => "Node".to_string(),
UtilizationRenderMode::Worker => "Worker".to_string(),
};

let num_cpus = match util_render_mode {
UtilizationRenderMode::Global => cpu_util_list.len(),
UtilizationRenderMode::Worker => used_cpus.len(),
};

let avg_usage = match util_render_mode {
UtilizationRenderMode::Global => calculate_average(cpu_util_list),
UtilizationRenderMode::Worker => {
let used_cpu_util_list: Vec<f64> = cpu_util_list
.iter()
.enumerate()
.filter_map(|(idx, utilization)| {
if used_cpus.contains(&ResourceIndex::new(idx as u32)) {
Some(*utilization)
} else {
None
}
})
.collect();
calculate_average(&used_cpu_util_list)
}
};

(which_util, num_cpus, avg_usage)
}
89 changes: 76 additions & 13 deletions crates/hyperqueue/src/dashboard/ui/screens/cluster/worker/mod.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,6 +12,7 @@ use crate::dashboard::ui::widgets::text::draw_text;
use crossterm::event::{KeyCode, KeyEvent};
use ratatui::layout::{Constraint, Direction, Layout, Rect};
use tako::hwstats::MemoryStats;
use tako::resources::{CPU_RESOURCE_NAME, ResourceIndex};
use tako::{JobTaskId, WorkerId};

mod cpu_util_table;
Expand All @@ -26,6 +27,7 @@ pub struct WorkerDetail {
worker_tasks_table: TasksTable,

utilization: Option<Utilization>,
utilization_render_mode: UtilizationRenderMode,
}

impl Default for WorkerDetail {
Expand All @@ -36,13 +38,37 @@ impl Default for WorkerDetail {
worker_config_table: Default::default(),
worker_tasks_table: TasksTable::non_interactive(),
utilization: None,
utilization_render_mode: UtilizationRenderMode::Worker,
}
}
}

#[derive(PartialEq)]
enum UtilizationRenderMode {
Global,
Worker,
}

impl UtilizationRenderMode {
fn next(&mut self) {
*self = match self {
UtilizationRenderMode::Global => UtilizationRenderMode::Worker,
UtilizationRenderMode::Worker => UtilizationRenderMode::Global,
}
}

fn next_text(&self) -> &str {
match self {
UtilizationRenderMode::Global => "Show worker CPU utilization",
UtilizationRenderMode::Worker => "Show global CPU utilization",
}
}
}

struct Utilization {
cpu: Vec<f64>,
memory: MemoryStats,
used_cpus: Vec<ResourceIndex>,
}

impl WorkerDetail {
Expand All @@ -66,12 +92,24 @@ impl WorkerDetail {
frame,
style_header_text(),
);
draw_text("<backspace>: Back", layout.footer, frame, style_footer());

draw_text(
format!(
"<backspace>: Back, <c>: {}",
self.utilization_render_mode.next_text()
)
.as_str(),
layout.footer,
frame,
style_footer(),
);

if let Some(util) = &self.utilization {
render_cpu_util_table(
&util.cpu,
&util.memory,
&util.used_cpus,
&self.utilization_render_mode,
layout.current_utilization,
frame,
table_style_deselected(),
Expand All @@ -95,21 +133,45 @@ impl WorkerDetail {
if let Some(worker_id) = self.worker_id {
self.utilization_history.update(data, worker_id);

if let Some((cpu_util, mem_util)) = data
if let Some(overview) = data
.workers()
.query_worker_overview_at(worker_id, data.current_time())
.and_then(|overview| overview.item.hw_state.as_ref())
.map(|hw_state| {
(
&hw_state.state.cpu_usage.cpu_per_core_percent_usage,
&hw_state.state.memory_usage,
)
})
{
self.utilization = Some(Utilization {
cpu: cpu_util.iter().map(|&v| v as f64).collect(),
memory: mem_util.clone(),
});
let worker_used_cpus: Vec<ResourceIndex> = match self.utilization_render_mode {
UtilizationRenderMode::Worker => overview
.item
.running_tasks
.iter()
.flat_map(|(_id, task_resource_alloc)| {
task_resource_alloc
.resources
.iter()
.filter_map(|resource_alloc| {
if resource_alloc.resource == CPU_RESOURCE_NAME {
Some(resource_alloc.indices.iter().map(|(index, _)| *index))
} else {
None
}
})
})
.flatten()
.collect(),
UtilizationRenderMode::Global => vec![],
};

if let Some(hw_state) = overview.item.hw_state.as_ref() {
self.utilization = Some(Utilization {
cpu: hw_state
.state
.cpu_usage
.cpu_per_core_percent_usage
.iter()
.map(|&v| v as f64)
.collect(),
memory: hw_state.state.memory_usage.clone(),
used_cpus: worker_used_cpus,
})
}
}

let tasks_info: Vec<(JobTaskId, &TaskInfo)> =
Expand All @@ -127,6 +189,7 @@ impl WorkerDetail {
pub fn handle_key(&mut self, key: KeyEvent) {
match key.code {
KeyCode::Backspace => self.worker_tasks_table.clear_selection(),
KeyCode::Char('c') => self.utilization_render_mode.next(),
_ => self.worker_tasks_table.handle_key(key),
}
}
Expand Down
19 changes: 19 additions & 0 deletions crates/hyperqueue/src/dashboard/ui/widgets/progressbar.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,6 +30,25 @@ pub fn get_progress_bar_color(progress: f64) -> Style {
}
}

pub fn get_progress_bar_cpu_color(progress: f64, used: bool) -> Style {
let color = if !used {
Color::Gray
} else if progress <= GREEN_THRESHOLD {
Color::Green
} else if progress <= YELLOW_THRESHOLD {
Color::Yellow
} else {
Color::Red
};

Style {
fg: Some(color),
bg: None,
add_modifier: Modifier::empty(),
sub_modifier: Modifier::empty(),
}
}

/**
* Creates a string progress bar for 0 < progress < 1
*/
Expand Down
Loading