At the current unoptimized memory utilization of the various data structures,
100k IPv6 routes consumes in the ballpark of 3-4GiB, which risks OOMing our
386 test machine.
Until we have the optimizations to (drastically) reduce that consumption,
skip the test that bloats too much for 32-bit machines.
Signed-off-by: David Anderson <danderson@tailscale.com>
Previously, the build ended up embedding an empty string, which made
the shell wrapper rebuild gocross on every invocation. This is still
reasonably fast, but fixing the bypass shaves 80% off gocross's overhead
when no rebuild is needed.
Signed-off-by: David Anderson <danderson@tailscale.com>
A bunch of us invoke tool/go from outside the repo that hosts gocross,
as a way of accessing our version-controlled toolchain. This removes
assumptions from gocross that it's being invoked within the repository
that contains its source code and toolchain configuration.
Fixestailscale/corp#9627
Signed-off-by: David Anderson <danderson@tailscale.com>
This used to make sense, but after a refactor somewhere along the line
this results in trying to download from a malformed URL and generally
confusing failures.
Signed-off-by: David Anderson <danderson@tailscale.com>
Xcode changed how/what data it exports to build steps at some point
recently, so our old way of figuring out the minimum support version
for clang stopped working.
Updates tailscale/corp#4095
Signed-off-by: David Anderson <danderson@tailscale.com>
Sometimes, our cached toolchain ends up being an older version of
Go, older than our go.mod allows. In that scenario, gocross-wrapper.sh
would find a usable toolchain, but then fail to compile gocross.
This change makes the wrapper script check that the cached toolchain's
minor version is good enough to build tailscale.com, and re-bootstraps
in shell if not.
Signed-off-by: David Anderson <danderson@tailscale.com>
I thought our versioning scheme would make go.mod include a commit hash
even on stable builds. I was wrong. Fortunately, the rest of this code
wants anything that 'git rev-parse' understands (to convert it into a full
git hash), and tags qualify.
Signed-off-by: David Anderson <danderson@tailscale.com>
By default, cmd/dist only prints the output of failed commands.
With this, you can turn all the noisy output back on.
Updates tailscale/corp#9045
Signed-off-by: David Anderson <danderson@tailscale.com>
The helper suppresses output if the command runs successfully. If the
command fails, it dumps the buffered output to stdout before returning
the error. This means the happy path isn't swamped by debug noise or
xcode being intensely verbose about what kind of day it's having,
but you still get debug output when something goes wrong.
Updates tailscale/corp#9045
Signed-off-by: David Anderson <danderson@tailscale.com>
Github requires explicitly listing every single job within a workflow
that is required for status checks, instead of letting you list entire
workflows. This is ludicrous, and apparently this nonsense is the
workaround.
Signed-off-by: David Anderson <danderson@tailscale.com>
This makes gocross and its bootstrap script understand an absolute
path in go.toolchain.rev to mean "use the given toolchain directly".
Signed-off-by: David Anderson <danderson@tailscale.com>
We need to build gocross from multiple repos, but Go's innate
git hash embedding only works when you build gocross from this repo,
not when you build it from elsewhere via 'go build
tailscale.com/tool/gocross'. Instead, explicitly embed the version
found with 'git rev-parse HEAD', which will work from any git repo.
Signed-off-by: David Anderson <danderson@tailscale.com>
This avoids accidentally overwriting variables from the input
environment, which might non-deterministically change the behavior
of gocross.
Signed-off-by: David Anderson <danderson@tailscale.com>
Turns out directing the printed script into the bootstrap location leads
to irritating "text file busy" problems and then having to muck about with
tempfiles and chmod and all that. Instead, have gocross write everything
with the right values.
Signed-off-by: David Anderson <danderson@tailscale.com>
So that when importing and using gocross from other repos, there's
an easy way to get at the right wrapper script that's in sync with
the gocross binary.
Signed-off-by: David Anderson <danderson@tailscale.com>
trimmed builds don't have absolute path information in executable
metadata, which leads the runtime.Caller approach failing
mysteriously in yarn with complaints about relative package paths.
So, instead of using embedded package metadata to find paths,
expect that we're being invoked within the tailscale repo, and
locate the tsconnect directory that way.
Signed-off-by: David Anderson <danderson@tailscale.com>
CI status doesn't collapse into "everything OK" if a job gets
skipped. Instead, always run the job, but skip its only step in PRs.
Signed-off-by: David Anderson <danderson@tailscale.com>
Replaces the former shell goop, which was a shell reimplementation
of a subset of version/mkversion.
Signed-off-by: David Anderson <danderson@tailscale.com>
With this change, you can collect version info from either a git
checkout of the tailscale.com Go module (this repo), or a git
checkout of a repo that imports the tailscale.com Go module.
Signed-off-by: David Anderson <danderson@tailscale.com>
So that I just get a quick PR to approve and merge instead of
periodically discovering that the SRI hash has bitrotted.
Signed-off-by: David Anderson <danderson@tailscale.com>
OSS-Fuzz doesn't update their version of Go as quickly as we do, so
we sometimes end up with OSS-Fuzz being unable to build our code for
a few weeks. We don't want CI to be red for that entire time, but
we also don't want to forget to reenable fuzzing when OSS-Fuzz does
start working again.
This change makes two configurations worthy of a CI pass:
- Fuzzing works, and we expected it to work. This is a normal
happy state.
- Fuzzing didn't compile, and we expected it to not compile. This
is the "OSS-Fuzz temporarily broken" state.
If fuzzing is unexpectedly broken, or unexpectedly not broken, that's
a CI failure because we need to either address a fuzz finding, or
update TS_FUZZ_CURRENTLY_BROKEN to reflect the state of OSS-Fuzz.
Signed-off-by: David Anderson <danderson@tailscale.com>
Github's matrix runner formats the race variant as '(amd64, true)' if we
use race=true. So, change the way the variable is defined so that it says
'(amd64, race)' even if that makes the if statements a bit more complex.
Signed-off-by: David Anderson <danderson@tailscale.com>
Instead of having a dozen files that contribute CI steps with
inconsistent configs, this one file lists out everything that,
for us, constitutes "a CI run". It also enables the slack
notification webhook to notify us exactly once on a mass breakage,
rather than once for every sub-job that fails.
Signed-off-by: David Anderson <danderson@tailscale.com>
The use of GOOS to mean "the compiler's host architecture" ends up
overriding whatever GOOS the user passed in, resulting in befuddling
errors like "unsupported GOOS/GOARCH pair linux/wasm" when the caller
requests js/wasm.
Signed-off-by: David Anderson <danderson@tailscale.com>
We've never used the "[ci skip]" magic commit header in our history,
across all our repos. This seems to be boilerplate we imported years
ago and have since been copying around our CI configs.
Signed-off-by: David Anderson <danderson@tailscale.com>
In the switch to static toolchains, we removed a legacy oddity from the
toolchain URL structure, but forgot to update printdep.
Signed-off-by: David Anderson <danderson@tailscale.com>
Before (note attempted use of absent date and commit hash):
"short": "1.37.0-dev",
"long": "1.37.0-dev-t",
After:
"short": "1.37.0-ERR-BuildInfo",
"long": "1.37.0-ERR-BuildInfo",
Signed-off-by: David Anderson <danderson@tailscale.com>
Commit 59c254579e moved a lot of work
from functions that could be eliminated at compile time (because
tests against runtime.GOOS are compile-time constant), into code
that must always run before main().
So, revert that, and instead optimize the package only by moving the
remaining string processing code behind sync.Onces.
Signed-off-by: David Anderson <danderson@tailscale.com>
We no longer carry an SRI hash for the toolchain, so flake
updating is no longer needed for toolchain changes.
Signed-off-by: David Anderson <danderson@tailscale.com>
Also removes the toolchain builds from flake.nix. For now the flake
build uses upstream Go 1.20, a followup change will switch it back to
our custom toolchain.
Updates tailscale/corp#9005
Signed-off-by: David Anderson <danderson@tailscale.com>
There is no stable release yet, and for alpha we want people on the
unstable build while we iterate.
Updates #502
Signed-off-by: David Anderson <danderson@tailscale.com>
We still accept the previous TS_AUTH_KEY for backwards compatibility, but the documented option name is the spelling we use everywhere else.
Updates #6321
Signed-off-by: David Anderson <danderson@tailscale.com>
This makes `tailscale debug watch-ipn` safe to use for troubleshooting
user issues, in addition to local debugging during development.
Signed-off-by: David Anderson <danderson@tailscale.com>
There is no unstability inherent in this package, it's just
unstable if you choose to import the flake at the main branch.
Signed-off-by: David Anderson <danderson@tailscale.com>
In some configurations, user explicitly do not want to store
tailscale state in k8s secrets, because doing that leads to
some annoying permission issues with sidecar containers.
With this change, TS_KUBE_SECRET="" and TS_STATE_DIR=/foo
will force storage to file when running in kubernetes.
Fixes#6704.
Signed-off-by: David Anderson <danderson@tailscale.com>
The operator creates a fair bit of internal cluster state to manage proxying,
dumping it all in the default namespace is handy for development but rude
for production.
Updates #502
Signed-off-by: David Anderson <danderson@tailscale.com>
We used to need to do timed requeues in a few places in the reconcile logic,
and the easiest way to do that was to plumb reconcile.Result return values
around. But now we're purely event-driven, so the only thing we care about
is whether or not an error occurred.
Incidentally also fix a very minor bug where headless services would get
completely ignored, rather than reconciled into the correct state. This
shouldn't matter in practice because you can't transition from a headful
to a headless service without a deletion, but for consistency let's avoid
having a path that takes no definite action if a service of interest does
exist.
Updates #502.
Signed-off-by: David Anderson <danderson@tailscale.com>
Previously, we had to do blind timed requeues while waiting for
the tailscale hostname, because we looked up the hostname through
the API. But now the proxy container image writes back its hostname
to the k8s secret, so we get an event-triggered reconcile automatically
when the time is right.
Updates #502
Signed-off-by: David Anderson <danderson@tailscale.com>
As is convention in the k8s world, use zap for structured logging. For
development, OPERATOR_LOGGING=dev switches to a more human-readable output
than JSON.
Updates #502
Signed-off-by: David Anderson <danderson@tailscale.com>
Our reconcile loop gets triggered again when the StatefulSet object
finally disappears (in addition to when its deletion starts, as indicated
by DeletionTimestamp != 0). So, we don't need to queue additional
reconciliations to proceed with the remainder of the cleanup, that
happens organically.
Signed-off-by: David Anderson <danderson@tailscale.com>
Tests cover configuring a proxy through an annotation rather than a
LoadBalancerClass, and converting between those two modes at runtime.
Updates #502.
Signed-off-by: David Anderson <danderson@tailscale.com>
For other test cases, the operator is going to produce similar generated
objects in several codepaths, and those objects are large. Move them out
to helpers so that the main test code stays a bit more intelligible.
The top-level Service that we start and end with remains in the main test
body, because its shape at the start and end is one of the main things that
varies a lot between test cases.
Updates #502.
Signed-off-by: David Anderson <danderson@tailscale.com>
The test verifies one of the successful reconcile paths, where
a client requests an exposed service via a LoadBalancer class.
Updates #502.
Signed-off-by: David Anderson <danderson@tailscale.com>
Also introduces an intermediary interface for the tailscale client, in
preparation for operator tests that fake out the Tailscale API interaction.
Updates #502.
Signed-off-by: David Anderson <danderson@tailscale.com>
This was initially developed in a separate repo, but for build/release
reasons and because go module management limits the damage of importing
k8s things now, moving it into this repo.
At time of commit, the operator enables exposing services over tailscale,
with the 'tailscale' loadBalancerClass. It also currently requires an
unreleased feature to access the Tailscale API, so is not usable yet.
Updates #502.
Signed-off-by: David Anderson <danderson@tailscale.com>
We still have to shell out to `tailscale up` because the container image's
API includes "arbitrary flags to tailscale up", unfortunately. But this
should still speed up startup a little, and also enables k8s-bound containers
to update their device information as new netmap updates come in.
Fixes#6657
Signed-off-by: David Anderson <danderson@tailscale.com>
