ProcDump is a Linux and macOS reimagining of the classic ProcDump tool from the Sysinternals suite for Windows. It creates process dumps in response to performance, runtime, signal, and resource-tracking triggers.
The supported implementation is the Cargo workspace in this repository. Linux
and macOS process access sit behind shared Rust interfaces. Linux native dumps
use the Rust corex ELF writer by default; managed dumps use .NET diagnostics
IPC; macOS and the explicit Linux -usegcore fallback use gcore.
The Linux eBPF kernel program and injected CLR profiler remain native
components under crates/procdump/native/, built and embedded by Cargo. Their userspace
loading, monitoring, EventPipe handling, orchestration, reporting, and dump
writing are Rust.
- Rust stable with
rustfmtandclippy - Linux: Clang,
pkg-config, libelf/zlib development packages,gdb, andgcore - macOS: Xcode command-line tools,
gdb, andgcore - .NET SDK/runtime when running managed integration scenarios
Please see installation instructions here.
Please see build instructions here.
cargo build --workspace
cargo test --workspace
cargo clippy --workspace --all-targets -- -D warningscrates/procdump: safe Rust API and shared dump/monitoring implementationcrates/procdump-cli: command-line applicationcrates/procdump-capi: static C ABI and public headercrates/procdump/native/ebpf: optional Linux eBPF kernel programcrates/procdump/native/profiler: optional injected CLR profilertests/integration: unchanged compatibility scenarios and native test fixturesxtask: Cargo staging and integration-test runner
The procdump crate's default feature set supports immediate dump generation;
monitoring, .NET triggers, and restrack are additive features, while the CLI
enables full.
BREAKING CHANGE With the release of ProcDump 1.3 the switches are now aligned with the Windows ProcDump version. Please note that the Mac version currently has a limited set of triggers.
Capture Usage:
procdump [-n Count]
[-s Seconds]
[-c|-cl CPU_Usage]
[-m|-ml Commit_Usage1[,Commit_Usage2...]]
[-gcm [<GCGeneration>: | LOH: | POH:]Memory_Usage1[,Memory_Usage2...]]
[-gcgen Generation]
[-restrack [nodump]]
[-sr Sample_Rate]
[-tc Thread_Threshold]
[-fc FileDescriptor_Threshold]
[-sig Signal_Number1[,Signal_Number2...]]
[-pc|-pcl Provider:Counter[pN] Threshold]
[-e]
[-f Include_Filter,...]
[-fx Exclude_Filter]
[-mc Custom_Dump_Mask]
[-pf Polling_Frequency]
[-o]
[-log syslog|stdout]
{
{{[-w] Process_Name | [-pgid] PID} [Dump_File | Dump_Folder]}
}
Options:
-n Number of dumps to write before exiting.
-s Consecutive seconds before dump is written (default is 10).
-c CPU threshold above which to create a dump of the process.
-cl CPU threshold below which to create a dump of the process.
-m Memory commit threshold(s) (MB) above which to create dumps.
-ml Memory commit threshold(s) (MB) below which to create dumps.
-gcm [.NET] GC memory threshold(s) (MB) above which to create dumps for the specified generation or heap (default is total .NET memory usage).
-gcgen [.NET] Create dump when the garbage collection of the specified generation starts and finishes.
-restrack Enable memory leak tracking (malloc family of APIs). If used without other triggers, use 't' to manually capture a restrack report. When used with other triggers, the 'nodump' option can be used to prevent dump generation and only produce restrack report(s).
-sr Sample rate when using -restrack.
-tc Thread count threshold above which to create a dump of the process.
-fc File descriptor count threshold above which to create a dump of the process.
-sig Comma separated list of signal number(s) during which any signal results in a dump of the process.
-pc [.NET] Trigger when performance counter is at or exceeds the threshold. Format: provider_name:counter_name[pN] threshold. Supports both EventCounters and System.Diagnostics.Metrics. For histogram instruments, append [pN] to select a percentile (e.g., [p50], [p95], [p99]). Default is p50 if omitted.
-pcl [.NET] Trigger when performance counter falls below the threshold. Format: provider_name:counter_name[pN] threshold.
-e [.NET] Create dump when the process encounters an exception.
-f Filter (include) on the content of .NET exceptions (comma separated). Wildcards (*) are supported.
-fx Filter (exclude) on the content of -restrack call stacks. Wildcards (*) are supported.
-mc Custom core dump mask (in hex) indicating what memory should be included in the core dump. Please see 'man core' (/proc/[pid]/coredump_filter) for available options.
-pf Polling frequency.
-o Overwrite existing dump file.
-log Writes extended ProcDump tracing to the specified output stream (syslog or stdout).
-w Wait for the specified process to launch if it's not running.
-pgid Process ID specified refers to a process group ID.
The -restrack switch activates resource tracking, allowing for the monitoring and reporting of any resource allocations that have not been freed at the time of generating the core dump. The results are saved to a file with a '.restrack' extension. Currently, the following resource allocation/deallocation functions are tracked:
Allocation:
- malloc
- calloc
- realloc
- reallocarray
- mmap
Deallocation:
- free
- munmap
The Mac version does not currently implement resource tracking.
The following examples all target a process with pid == 1234
The following will create a core dump immediately.
sudo procdump 1234
The following will create 3 core dumps 10 seconds apart.
sudo procdump -n 3 1234
The following will create 3 core dumps 5 seconds apart.
sudo procdump -n 3 -s 5 1234
The following will create a core dump each time the process has CPU usage >= 65%, up to 3 times, with at least 10 seconds between each dump.
sudo procdump -c 65 -n 3 1234
The following will create a core dump each time the process has CPU usage >= 65%, up to 3 times, with at least 5 seconds between each dump.
sudo procdump -c 65 -n 3 -s 5 1234
The following will create a core dump when CPU usage is outside the range [10,65].
sudo procdump -cl 10 -c 65 1234
The following will create a core dump when CPU usage is >= 65% or memory usage is >= 100 MB.
sudo procdump -c 65 -m 100 1234
The following will create a core dump when memory usage is >= 100 MB followed by another dump when memory usage is >= 200MB.
sudo procdump -m 100,200 1234
The following will create a memory leak report (no dumps) every time the user presses 't':
sudo procdump -restrack 1234
The following will create 3 memory leak reports (no dumps) 5 seconds apart:
sudo procdump -n 3 -s 5 -restrack nodump 1234
The following will create a core dump and a memory leak report when memory usage is >= 100 MB
sudo procdump -m 100 -restrack 1234
The following will create a memory leak report (no dumps) when memory usage is >= 100 MB
sudo procdump -m 100 -restrack nodump 1234
The following will create a core dump and a memory leak report when memory usage is >= 100 MB by sampling every 10th memory allocation.
sudo procdump -m 100 -restrack -sr 10 1234
The following will create a core dump and a memory leak report when memory usage is >= 100 MB and exclude any call stacks that contain frames with the string "cache" in them
sudo procdump -m 100 -restrack -fx *cache* 1234
The following will create a core dump when the total .NET memory usage is >= 100 MB followed by another dump when memory usage is >= 200MB.
sudo procdump -gcm 100,200 1234
The following will create a core dump when .NET memory usage for generation 1 is >= 1 MB followed by another dump when memory usage is >= 2MB.
sudo procdump -gcm 1:1,2 1234
The following will create a core dump when .NET Large Object Heap memory usage is >= 100 MB followed by another dump when memory usage is >= 200MB.
sudo procdump -gcm LOH:100,200 1234
The following will create a core dump at the start and end of a .NET generation 1 garbage collection.
sudo procdump -gcgen 1
The following will create a core dump in the /tmp directory immediately.
sudo procdump 1234 /tmp
The following will create a core dump in the current directory with the name dump_0.1234. If -n is used, the files will be named dump_0.1234, dump_1.1234 and so on.
sudo procdump 1234 dump
The following will create a core dump when a SIGSEGV occurs.
sudo procdump -sig 11 1234
The following will create a core dump when a SIGSEGV occures where the core dump contains only anonymous private mappings.
sudo procdump -mc 1 -sig 11 1234
The following will create a core dump when the target .NET application throws a System.InvalidOperationException
sudo procdump -e -f System.InvalidOperationException 1234
The include filter supports partial and wildcard matching, so the following will create a core dump too for a System.InvalidOperationException
sudo procdump -e -f InvalidOperation 1234
or
sudo procdump -e -f "*Invali*Operation*" 1234
All options can also be used with
-w, to wait for any process with the given name.
The following waits for a process named my_application and creates a core dump immediately when it is found.
sudo procdump -w my_application
- Currently will only run on Linux Kernels version 3.5+ or macOS Sierra+.
- Does not have full feature parity with the Windows version of ProcDump, specifically stay-alive functionality.
- Ask a question on StackOverflow (tag with ProcDumpForLinux)
- Request a new feature on GitHub
- Vote for popular feature requests
- File a bug in GitHub Issues
If you are interested in fixing issues and contributing directly to the code base, please see the document How to Contribute, which covers the following:
- How to build and run from source
- The development workflow, including debugging and running tests
- Coding Guidelines
- Submitting pull requests
Please see also our Code of Conduct.
Copyright (c) Microsoft Corporation. All rights reserved.
Licensed under the MIT License.
ProcDump for Linux:
- Clones, compiles and statically links against libbpf (https://github.com/libbpf/libbpf)
- Copies symbol resolution source code from BCC (https://github.com/iovisor/bcc).
- Uses eBPF