Resource Tracker for R#
This R package provides raw access to the resource-tracker Python module via
the reticulate R package along with a convenience R6 wrapper for the
ResourceTracker class with the most common methods, such as starting and
stopping the tracker, collecting samples, and generating reports.
Installation#
This package requires Python to be installed along with the resource-tracker
Python module, which can be installed from R via:
Note that on MacOS and Windows, you also need to install psutil that's used
under the hood to collect metrics. On Linux, it's optional, and can rely on the
/proc filesystem when using a modern kernel. For more details, see the
installation instructions for the Python package.
Once the Python dependencies are resolved, you can either install the R package from CRAN:
Or the most recent (development version) from GitHub:
Getting started#
Load the package and start the resource tracker:
This now runs in the background, not blocking your R session.
Let's simulate some work for some seconds with the below, inefficient rolling mean implementation:
numbers <- 1:1e6
window <- 3
rollavg <- sapply(
seq_len(length(numbers) - window + 1),
function(i) mean(numbers[i:(i + window - 1)]))
It's time to check on the aggregated statistics of the resources used:
tracker$stats()
# List of 9
# $ process_cpu_usage :List of 2
# ..$ mean: num 1.12
# ..$ max : num 1.14
# $ process_memory :List of 2
# ..$ mean: num 728783
# ..$ max : num 798130
# $ process_gpu_usage :List of 2
# ..$ mean: num 0
# ..$ max : num 0
# $ process_gpu_vram :List of 2
# ..$ mean: num 0
# ..$ max : num 0
# $ process_gpu_utilized :List of 2
# ..$ mean: num 0
# ..$ max : num 0
# $ system_disk_space_used_gb:List of 1
# ..$ max: num 2560
# $ system_net_recv_bytes :List of 1
# ..$ sum: num 1840217
# $ system_net_sent_bytes :List of 1
# ..$ sum: num 1843725
# $ timestamp :List of 1
# ..$ duration: num 7
Or get recommendations on hardware requirements for the next run:
tracker$recommend_resources()
# List of 4
# $ cpu : int 1
# $ memory: int 1024
# $ gpu : int 0
# $ vram : int 0
tracker$recommend_server()
# List of 50
# $ vendor_id : chr "upcloud"
# $ server_id : chr "DEV-1xCPU-1GB-10GB"
# $ description : chr "Developer 1 vCPUs, 1 GB RAM"
# $ family : chr "Developer"
# $ vcpus : int 1
# $ hypervisor : chr "KVM"
# $ cpu_allocation : chr "Shared"
# $ cpu_cores : int 1
# $ cpu_architecture : chr "x86_64"
# $ cpu_manufacturer : chr "AMD"
# $ cpu_family : chr "EPYC"
# $ cpu_model : chr "7542"
# $ cpu_l1_cache : int 131072
# $ cpu_l2_cache : int 524288
# $ cpu_l3_cache : int 16777216
# $ cpu_flags : chr [1:88] "fpu" "vme" "de" "pse" ...
# $ memory_amount : int 1024
# $ storage_size : int 10
# $ inbound_traffic : num 0
# $ outbound_traffic : num 1024
# $ ipv4 : int 1
# $ price : num 0.0052
# ...
Or look at the collected samples:
tracker$system_metrics
# 'data.frame': 7 obs. of 21 variables:
# $ timestamp : POSIXct, format: "2025-08-08 00:02:28" "2025-08-08 00:02:29" ...
# $ processes : num 697 696 694 693 694 694 693
# $ utime : num 3.01 2.25 2.3 2.15 2.9 1.44 1.36
# $ stime : num 0.51 0.57 0.55 0.24 0.89 0.83 0.36
# $ cpu_usage : num 3.52 2.82 2.85 2.39 3.79 ...
# $ memory_free : num 8279604 8254872 8298804 8247688 8200296 ...
# $ memory_used : num 23920048 23946512 23901264 23951016 23993224 ...
# $ memory_buffers : num 2992 2992 2992 2992 2992 ...
# $ memory_cached : num 33344236 33342504 33343820 33345184 33350368 ...
# $ memory_active : num 47841676 47860856 47855900 47877072 47989240 ...
# $ memory_inactive : num 106028 106028 106028 106028 106028 ...
# $ disk_read_bytes : num 1138688 819200 24576 126976 36352000 ...
# $ disk_write_bytes : num 942080 0 8585216 61440 20480 ...
# $ disk_space_total_gb: num 3700 3700 3700 3700 3700 ...
# $ disk_space_used_gb : num 2560 2560 2560 2560 2560 ...
# $ disk_space_free_gb : num 1140 1140 1140 1140 1140 ...
# $ net_recv_bytes : num 710628 1204 7685 4553 1094516 ...
# $ net_sent_bytes : num 710504 1463 7042 8261 1085653 ...
# $ gpu_usage : num 0.06 0.13 0.11 0.05 0.03 0.02 0.02
# $ gpu_vram : num 571 562 558 558 558 558 558
# $ gpu_utilized : num 1 1 1 1 1 1 1
tracker$process_metrics
# 'data.frame': 7 obs. of 12 variables:
# $ timestamp : POSIXct, format: "2025-08-08 00:02:28" "2025-08-08 00:02:29" ...
# $ pid : num 941247 941247 941247 941247 941247 ...
# $ children : num 4 4 4 4 4 4 4
# $ utime : num 1.01 1.02 1.01 1 0.99 1.04 1.01
# $ stime : num 0.1 0.12 0.07 0.13 0.14 0.09 0.09
# $ cpu_usage : num 1.11 1.14 1.08 1.13 1.13 1.13 1.1
# $ memory : num 646973 668091 680918 724281 798130 ...
# $ read_bytes : num 0 0 0 0 0 0 0
# $ write_bytes : num 0 0 0 0 0 0 0
# $ gpu_usage : num 0 0 0 0 0 0 0
# $ gpu_vram : num 0 0 0 0 0 0 0
# $ gpu_utilized: num 0 0 0 0 0 0 0
tracker$get_combined_metrics()
# 'data.frame': 7 obs. of 32 variables:
# $ timestamp : POSIXct, format: "2025-08-08 00:02:28" "2025-08-08 00:02:29" ...
# $ system_processes : num 697 696 694 693 694 694 693
# $ system_utime : num 3.01 2.25 2.3 2.15 2.9 1.44 1.36
# $ system_stime : num 0.51 0.57 0.55 0.24 0.89 0.83 0.36
# $ system_cpu_usage : num 3.52 2.82 2.85 2.39 3.79 ...
# $ system_memory_free : num 8279604 8254872 8298804 8247688 8200296 ...
# $ system_memory_used : num 23920048 23946512 23901264 23951016 23993224 ...
# $ system_memory_buffers : num 2992 2992 2992 2992 2992 ...
# $ system_memory_cached : num 33344236 33342504 33343820 33345184 33350368 ...
# $ system_memory_active : num 47841676 47860856 47855900 47877072 47989240 ...
# $ system_memory_inactive : num 106028 106028 106028 106028 106028 ...
# $ system_disk_read_bytes : num 1138688 819200 24576 126976 36352000 ...
# $ system_disk_write_bytes : num 942080 0 8585216 61440 20480 ...
# $ system_disk_space_total_gb: num 3700 3700 3700 3700 3700 ...
# $ system_disk_space_used_gb : num 2560 2560 2560 2560 2560 ...
# $ system_disk_space_free_gb : num 1140 1140 1140 1140 1140 ...
# $ system_net_recv_bytes : num 710628 1204 7685 4553 1094516 ...
# $ system_net_sent_bytes : num 710504 1463 7042 8261 1085653 ...
# $ system_gpu_usage : num 0.06 0.13 0.11 0.05 0.03 0.02 0.02
# $ system_gpu_vram : num 571 562 558 558 558 558 558
# $ system_gpu_utilized : num 1 1 1 1 1 1 1
# $ process_pid : num 941247 941247 941247 941247 941247 ...
# $ process_children : num 4 4 4 4 4 4 4
# $ process_utime : num 1.01 1.02 1.01 1 0.99 1.04 1.01
# $ process_stime : num 0.1 0.12 0.07 0.13 0.14 0.09 0.09
# $ process_cpu_usage : num 1.11 1.14 1.08 1.13 1.13 1.13 1.1
# $ process_memory : num 646973 668091 680918 724281 798130 ...
# $ process_read_bytes : num 0 0 0 0 0 0 0
# $ process_write_bytes : num 0 0 0 0 0 0 0
# $ process_gpu_usage : num 0 0 0 0 0 0 0
# $ process_gpu_vram : num 0 0 0 0 0 0 0
# $ process_gpu_utilized : num 0 0 0 0 0 0 0
Note that the system_metrics and process_metrics are properties of the
ResourceTracker object, while get_combined_metrics is a method that takes
optional parameters (e.g. to render more human-friendly names or return all
values in bytes) and returns a data.frame.
And finally, let's generate a report and open it in your browser:
For an example HTML report, see this Metaflow card.
Advanced usage#
Not all features are exposed via the ResourceTracker class. For example, you
could snapshot a ResourceTracker object to a string or compressed file and
later restore it. To access these features, open a GitHub issue or you could use
the Python API directly via the resource.tracker::resource_tracker Python
module reference.