One-stop ZFS backup & replication solution
Go to file
Christian Schwarz a8e92971d0 zfs: rewrite SendStream, fix bug in Close() on FreeBSD, add platformtests
This commit was motivated by https://github.com/zrepl/zrepl/issues/495
where, on FreeBSD with OpenZFS 2.0, a SendStream.Close() call might wait indefinitely for `zfs send` to exit.
The reason is that, due to the refactoring done for redacted send & recv
(30af21b025),
the `dump_bytes` function, which writes to the pipe, executes in a separate thread (synctask taskq) iff not `HAVE_LARGE_STACKS`.
The `zfs send` process/thread waits for that taskq thread using an uninterruptible primitive.
So when we SIGKILL `zfs send`, that signal doesn't reach the right thread to interrupt the pipe write.

Theoretically this affects both Linux and FreeBSD, but most Linux users `HAVE_LARGE_STACKS` and since https://github.com/penzfs/zfs/pull/12350/files OpenZFS on FreeBSD `HAVE_LARGE_STACKS` as well.
However, at least until FreeBSD 13.1, possibly for the entire 13 lifecycle, we're going to have to live with that oddity.

Measures taken in this commit:
- Report the behavior as an upstream bug https://github.com/openzfs/zfs/issues/12500
- Change SendStream code so that it closes zrepl's read-end of the pipe (see comment in code)
- Clean up and make explicit SendStream's state handling
- Write extensive platformtests for SendStream
    - They pass on my Linux install and on FreeBSD 12
    - FreeBSD 13 still needs testing.

fixes https://github.com/zrepl/zrepl/issues/495
2021-09-19 20:11:31 +02:00
.circleci docs / CI: stop creating churn with doc commits & commit as zreplbot@ 2021-07-08 17:07:24 +02:00
.github docs: GitHub Sponsors link 2020-06-14 15:26:05 +02:00
build Update to protobuf v1.25 and grpc 1.35; bump CI to go1.12 2021-01-25 00:39:01 +01:00
cli Reformat all files with make format. 2020-08-31 23:57:45 +02:00
client status: byteprogresshistory: disable averaging as workaround for #497 2021-09-12 20:08:44 +02:00
config send/recv: job-level bandwidth limiting 2021-09-12 20:08:43 +02:00
daemon send/recv: job-level bandwidth limiting 2021-09-12 20:08:43 +02:00
dist dist/grafana: fix endpoint abstractions cache metric panel 2020-08-04 01:36:31 +02:00
docs send/recv: job-level bandwidth limiting 2021-09-12 20:08:43 +02:00
endpoint bandwidth limiting: use correct field name on error 2021-09-19 20:03:01 +02:00
logger logger: fix go-1.15-discovered conversion from int to string 2020-08-12 21:44:02 +02:00
packaging fix make deb-docker for all platforms but amd64 2021-09-13 22:54:21 +02:00
platformtest zfs: rewrite SendStream, fix bug in Close() on FreeBSD, add platformtests 2021-09-19 20:11:31 +02:00
pruning pruning/keep_last_n: correctly handle the case where count > matching snaps 2021-03-25 22:42:01 +01:00
replication replication: simplify parallel replication variables & expose them in config 2021-03-14 17:30:10 +01:00
rpc rpc/dataconn: fix log message in closeErr case 2021-09-19 18:40:39 +02:00
tlsconf build: go1.14 + address tlsconf deprecation notice 2020-03-27 12:40:57 +01:00
transport transport/tcp: ipmap tests: remove tests that cover CIDR normalization 2021-09-15 08:43:08 +02:00
util platformtest: fix bandwidth-limiting-related panics (missing BucketCapacity in sender/receiver config) 2021-09-19 20:03:01 +02:00
version build: parametrized pipeline (no more approval for release build), embed exact go version info in artifacts 2020-09-06 15:39:34 +02:00
zfs zfs: rewrite SendStream, fix bug in Close() on FreeBSD, add platformtests 2021-09-19 20:11:31 +02:00
.gitignore Rudimentary Makefile specifying requirements for a release 2017-09-30 16:40:39 +02:00
.gitmodules docs: move hugo docs to old directory 2017-11-11 23:25:12 +01:00
.golangci.yml run golangci-lint and apply suggested fixes 2019-03-27 13:12:26 +01:00
build.Dockerfile build: parametrized pipeline (no more approval for release build), embed exact go version info in artifacts 2020-09-06 15:39:34 +02:00
build.installprotoc.bash build: Linux arm64 support 2019-06-23 15:25:26 +02:00
debian build: rpm + deb targets, build-in-docker targets, CircleCI pipeline rewrite 2020-09-02 21:34:52 +02:00
go.mod send/recv: job-level bandwidth limiting 2021-09-12 20:08:43 +02:00
go.sum send/recv: job-level bandwidth limiting 2021-09-12 20:08:43 +02:00
lazy.sh Update to protobuf v1.25 and grpc 1.35; bump CI to go1.12 2021-01-25 00:39:01 +01:00
LICENSE LICENSE + docs: adjust copyright 2018-10-13 17:34:05 +02:00
main.go implement new 'zrepl status' 2021-03-14 18:24:25 +01:00
Makefile build: bump release build Go version + quickcheck default to Go 1.16 2021-02-20 17:29:17 +01:00
README.md docs + readme: actually fix Patreon badge 2020-11-01 16:10:48 +01:00

GitHub license Language: Go User Docs Support me on Patreon Donate via GitHub Sponsors Donate via Liberapay Donate via PayPal Twitter

zrepl

zrepl is a one-stop ZFS backup & replication solution.

User Documentation

User Documentation can be found at zrepl.github.io.

Bug Reports

  1. If the issue is reproducible, enable debug logging, reproduce and capture the log.
  2. Open an issue on GitHub, with logs pasted as GitHub gists / inline.

Feature Requests

  1. Does your feature request require default values / some kind of configuration? If so, think of an expressive configuration example.
  2. Think of at least one use case that generalizes from your concrete application.
  3. Open an issue on GitHub with example conf & use case attached.
  4. Optional: Post a bounty on the issue, or contact Christian Schwarz for contract work.

The above does not apply if you already implemented everything. Check out the Coding Workflow section below for details.

Building, Releasing, Downstream-Packaging

This section provides an overview of the zrepl build & release process. Check out docs/installation/compile-from-source.rst for build-from-source instructions.

Overview

zrepl is written in Go and uses Go modules to manage dependencies. The documentation is written in ReStructured Text using the Sphinx framework.

Install build dependencies using ./lazy.sh devsetup. lazy.sh uses python3-pip to fetch the build dependencies for the docs - you might want to use a venv. If you just want to install the Go dependencies, run ./lazy.sh godep.

The test suite is split into pure Go tests (make test-go) and platform tests that interact with ZFS and thus generally require root privileges (sudo make test-platform). Platform tests run on their own pool with the name zreplplatformtest, which is created using the file vdev in /tmp.

For a full code coverage profile, run make test-go COVER=1 && sudo make test-platform && make cover-merge. An HTML report can be generated using make cover-html.

Code generation is triggered by make generate. Generated code is committed to the source tree.

Build & Release Process

The Makefile is catering to the needs of developers & CI, not distro packagers. It provides phony targets for

  • local development (building, running tests, etc)
  • building a release in Docker (used by the CI & release management)
  • building .deb and .rpm packages out of the release artifacts.

Build tooling & dependencies are documented as code in lazy.sh. Go dependencies are then fetched by the go command and pip dependencies are pinned through a requirements.txt.

We use CircleCI for continuous integration. There are two workflows:

  • ci runs for every commit / branch / tag pushed to GitHub. It is supposed to run very fast (<5min and provides quick feedback to developers). It runs formatting checks, lints and tests on the most important OSes / architectures. Artifacts are published to minio.cschwarz.com (see GitHub Commit Status).

  • release runs

    • on manual triggers through the CircleCI API (in order to produce a release)
    • periodically on master Artifacts are published to minio.cschwarz.com (see GitHub Commit Status).

Releases are issued via Git tags + GitHub Releases feature. The procedure to issue a release is as follows:

  • Issue the source release:
    • Git tag the release on the master branch.
    • Push the tag.
    • Run ./docs/publish.sh to re-build & push zrepl.github.io.
  • Issue the official binary release:
    • Run the release pipeline (triggered via CircleCI API)
    • Download the artifacts to the release manager's machine.
    • Create a GitHub release, edit the changelog, upload all the release artifacts, including .rpm and .deb files.
    • Issue the GitHub release.
    • Add the .rpm and .deb files to the official zrepl repos, publish those.

Official binary releases are not re-built when Go receives an update. If the Go update is critical to zrepl (e.g. a Go security update that affects zrepl), we'd issue a new source release. The rationale for this is that whereas distros provide a mechanism for this ($zrepl_source_release-$distro_package_revision), GitHub Releases doesn't which means we'd need to update the existing GitHub release's assets, which nobody would notice (no RSS feed updates, etc.). Downstream packagers can read the changelog to determine whether they want to push that minor release into their distro or simply skip it.

Additional Notes to Distro Package Maintainers

  • Use sudo make test-platform-bin && sudo make test-platform on a test system to validate that zrepl's abstractions on top of ZFS work with the system ZFS.
  • Ship a default config that adheres to your distro's hier and logging system.
  • Ship a service manager file and please try to upstream it to this repository.
    • dist/systemd contains a Systemd unit template.
  • Ship other material provided in ./dist, e.g. in /usr/share/zrepl/.
  • Have a look at the Makefile's ZREPL_VERSION variable and how it passed to Go's ldFlags. This is how zrepl version knows what version number to show. Your build system should set the ldFlags flags appropriately and add a prefix or suffix that indicates that the given zrepl binary is a distro build, not an official one.
  • Make sure you are informed about new zrepl versions, e.g. by subscribing to GitHub's release RSS feed.

Contributing Code

  • Open an issue when starting to hack on a new feature
  • Commits should reference the issue they are related to
  • Docs improvements not documenting new features do not require an issue.

Breaking Changes

Backward-incompatible changes must be documented in the git commit message and are listed in docs/changelog.rst.

Glossary & Naming Inconsistencies

In ZFS, dataset refers to the objects filesystem, ZVOL and snapshot.
However, we need a word for filesystem & ZVOL but not a snapshot, bookmark, etc.

Toward the user, the following terminology is used:

  • filesystem: a ZFS filesystem or a ZVOL
  • filesystem version: a ZFS snapshot or a bookmark

Sadly, the zrepl implementation is inconsistent in its use of these words: variables and types are often named dataset when they in fact refer to a filesystem.

There will not be a big refactoring (an attempt was made, but it's destroying too much history without much gain).

However, new contributions & patches should fix naming without further notice in the commit message.