2db8a882e8
* Centralize cache path handling. * Add better debugging support for static build infrastructure. * Centralize handling of git repository fetching for static builds. * Centralize build directory handling in the static build. And quit using a sub-directory of packaging/makeself from the source tree for it. * Remove version numbers from static build jobs. * Better organize static build jobs. 0x numbers for non-build prep-work 1x numbers for libraries that potentially impact multiple other things we vendor. 2x numbers for libraries that are direct dependencies of Netdata 3x numbers for combined libraries and tooling used by Netdata 4x numbers for general tooling used by Netdata 5x numbers for tooling used by single specific components 6x numbers for non-build prep-work for the Netdata build 7x numbers for the actual build of Netdata 8x numbers for post-build checks 9x numbers for the actual packaging * Clean up variable handling in Netdata build job. * Split post-build handling steps to their own jobs. This will make it easier to see what is actually going on in the build process. * Clean up CI messages. * Split archive creation sub-steps into indivudal jobs. * Disable shell tracing for archive creation job. It’s not needed in 99.9% of cases, and should only be enabled locally if it is needed. * Assorted fixes for code restructuring. * Tidy up paths for runtime check. * Fix CI handling of artifacts. |
||
|---|---|---|
| .. | ||
| jobs | ||
| build-static.sh | ||
| build-x86_64-static.sh | ||
| build.sh | ||
| bundled-packages.version | ||
| functions.sh | ||
| install-alpine-packages.sh | ||
| install-or-update.sh | ||
| makeself-header.sh | ||
| makeself-help-header.txt | ||
| makeself-license.txt | ||
| makeself.lsm | ||
| makeself.sh | ||
| post-installer.sh | ||
| README.md | ||
| run-all-jobs.sh | ||
| uname2platform.sh | ||
Netdata static binary build
We publish pre-built static builds of Netdata for Linux systems. Currently, these are published for 64-bit x86, ARMv7,
AArch64, and POWER8+ hardware. These static builds are able to operate in a mostly self-contained manner and only
require a POSIX compliant shell and a supported init system. These static builds install under /opt/netdata. If
you are on a platform which we provide static builds for but do not provide native packages for, a static build
will be used by default for installation.
If you want to enforce the usage of a static build and have the installer return a failure if one is not available,
you can do so by adding --static-only to the options you pass to the installer.
Requirements
- Container runtime tool (Docker or Podman)
Building a static binary package
Before you begin, make sure that your repo and the repo's submodules are clean from any previous builds and up to date. Otherwise, perform a cleanup
To build the static binary 64-bit distribution package, into the root folder on the netdata repo, run:
./packaging/makeself/build-static.sh x86_64
The program will:
- setup a new docker container with Alpine Linux
- install the required alpine packages (the build environment, needed libraries, etc)
- download and compile third party apps that are packaged with Netdata (
bash,curl, etc) - compile Netdata
Once finished, a file named netdata-vX.X.X-gGITHASH-x86_64-DATE-TIME.run will be created in the current directory. This is the Netdata binary package that can be run to install Netdata on any other computer.
You can build static binaries for other architectures such as armv7l, aarch64, and ppc64le.
Building binaries with debug info
To build Netdata binaries with debugging / tracing information in them, use:
cd /path/to/netdata.git
./packaging/makeself/build-static.sh x86_64 debug
These binaries are not optimized (they are a bit slower), they have certain features disables (like log flood protection), other features enables (like debug flags) and are not stripped (the binary files are bigger, since they now include source code tracing information).
Debugging Netdata binaries
Once you have installed a binary package with debugging info, you will need to install valgrind and run this command to start Netdata:
PATH="/opt/netdata/bin:${PATH}" valgrind --undef-value-errors=no /opt/netdata/bin/srv/netdata -D
The above command, will run Netdata under valgrind. While Netdata runs under valgrind it will be 10x slower and use a lot more memory.
If Netdata crashes, valgrind will print a stack trace of the issue. Open a github issue to let us know.
To stop Netdata while it runs under valgrind, press Control-C on the console.
If you omit the parameter
--undef-value-errors=noto valgrind, you will get hundreds of errors about conditional jumps that depend on uninitialized values. This is normal. Valgrind has heuristics to prevent it from printing such errors for system libraries, but for the static Netdata binary, all the required libraries are built into Netdata. So, valgrind cannot apply its heuristics and prints them.