Parca Agent is an always-on sampling profiler that uses eBPF to capture raw profiling data with very low overhead. It observes user-space and kernel-space stacktraces 100 times per second and builds pprof formatted profiles from the extracted data. Read more details in the design documentation.
The collected data can be viewed locally via HTTP endpoints and then be configured to be sent to a Parca server to be queried and analyzed over time.
It discovers targets through:
- Kubernetes: Discovering all the containers on the node the Parca agent is running on. (On by default, but can be disabled using
- systemd: A list of systemd units to be profiled on a node can be configured for the Parca agent to pick up. (Use the
--systemd-unitsflag to list the units to profile, eg.
--systemd-units=docker.serviceto profile the docker daemon)
See the Kubernetes Getting Started.
Profiles available for compiled languages (eg. C, C++, Go, Rust):
- Soon: Network usage, Allocations
The following types of profiles require explicit instrumentation:
- Runtime specific information such as Goroutines
The HTTP endpoints can be used to inspect the active profilers, by visiting port
7071 of the process (the host-port that the agent binds to can be configured using the
On a minikube cluster that might look like the following:
And by clicking "Show Profile" in one of the rows, the currently collected profile will be rendered once the collection finishes (this can take up to 10 seconds).
A raw profile can also be downloaded here by clicking "Download Pprof". Note that in the case of native stack traces such as produced from compiled language like C, C++, Go, Rust, etc. are not symbolized and if this pprof profile is analyzed using the standard pprof tooling the symbols will need to be available to the tooling.
To debug potential errors, enable debug logging using
Usage: parca-agent --node=STRING Flags: -h, --help Show context-sensitive help. --log-level="info" Log level. --http-address=":7071" Address to bind HTTP server to. --node=STRING Name node the process is running on. If on Kubernetes, this must match the Kubernetes node name. --external-label=KEY=VALUE;... Label(s) to attach to all profiles. --store-address=STRING gRPC address to send profiles and symbols to. --bearer-token=STRING Bearer token to authenticate with store. --bearer-token-file=STRING File to read bearer token from to authenticate with store. --insecure Send gRPC requests via plaintext instead of TLS. --insecure-skip-verify Skip TLS certificate verification. --sampling-ratio=1.0 Sampling ratio to control how many of the discovered targets to profile. Defaults to 1.0, which is all. --kubernetes Discover containers running on this node to profile automatically. --pod-label-selector=STRING Label selector to control which Kubernetes Pods to select. --systemd-units=SYSTEMD-UNITS,... systemd units to profile on this node. --temp-dir="/tmp" Temporary directory path to use for object files. --socket-path=STRING The filesystem path to the container runtimes socket. Leave this empty to use the defaults. --profiling-duration=10s The agent profiling duration to use. Leave this empty to use the defaults. --systemd-cgroup-path="/sys/fs/cgroup/systemd/system.slice" The cgroupfs path to a systemd slice.
To discover systemd units, the names must be passed to the agent. For example, to profile the docker daemon pass
To sample all targets, either to save resources on storage or reduce overhead, use the
--sampling-ratio flag. For example, to profile only 50% of the discovered targets use
Kubernetes label selector
To further sample targets on Kubernetes use the
--pod-label-selector= flag. For example to only profile Pods with the
app.kubernetes.io/name=my-web-app label, use
- Additional language support for just-in-time (JIT) compilers, and dynamic languages (non-exhaustive list):
- Additional types of profiles:
- Memory allocations
- Network usage
Parca Agent requires to be run as
root user (or
CAP_SYS_ADMIN). Various security precautions have been taken to protect users running Parca Agent. See details in Security Considerations.
To report a security vulnerability see this guide.
Check out our Contributing Guide to get started!
- Aqua Security for creating libbpfgo (cgo bindings for libbpf), while we contributed several features to it, they have made it spectacularly easy for us to contribute and it has been a great collaboration. Their use of libbpf in tracee has also been a helpful resource.
- Kinvolk for creating Inspektor Gadget some parts of this project were inspired by parts of it.