When pushing a large chain of CLs, builds can fail with OOM issues as
many Nix evaluations of the depot are happening simultaneously.
To work around this, we limit the concurrency of simultaneous Nix
evaluations (i.e. the `:llama` step). This can slow down the start of
builds in a large chain of small changes, but that is a better
tradeoff than failing the builds entirely and making people click
buttons.
Change-Id: If351aaad22d52e2bcf871377f22ab1df594c518d
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7501
Reviewed-by: sterni <sternenseemann@systemli.org>
Autosubmit: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
This is already allowed de facto, since there seems to be a special
exception for reading from derivation outputs. What is forbidden, is
access to files imported to the store (even via builtins.toFile) and
derivation files. The latter is required for doing dependency analysis
on arbitrary derivations, unfortunately.
Access to the store allows kind of evil things, but it should
be (hopefully) hard to do this by accident, and accessing derivation
files is not impure, though it relies on store implementation internals
so to speak.
Change-Id: I33a7de83ef0ee20a7076690329d62f6caffffe5f
Reviewed-on: https://cl.tvl.fyi/c/depot/+/6835
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
Reviewed-by: grfn <grfn@gws.fyi>
Upstream nixpkgs removed a lot of aliases this time, so we needed to do
the following transformations. It's a real shame that aliases only
really become discoverable easily when they are removed.
* runCommandNoCC -> runCommand
* gmailieer -> lieer
We also need to work around the fact that home-manager hasn't catched
on to this rename.
* mysql -> mariadb
* pkgconfig -> pkg-config
This also affects our Nix fork which needs to be bumped.
* prometheus_client -> prometheus-client
* rxvt_unicode -> rxvt-unicode-unwrapped
* nix-review -> nixpkgs-review
* oauth2_proxy -> oauth2-proxy
Additionally, some Go-related builders decided to drop support for
passing the sha256 hash in directly, so we need to use the generic hash
arguments.
Change-Id: I84aaa225ef18962937f8616a9ff064822f0d5dc3
Reviewed-on: https://cl.tvl.fyi/c/depot/+/6792
Autosubmit: sterni <sternenseemann@systemli.org>
Tested-by: BuildkiteCI
Reviewed-by: grfn <grfn@gws.fyi>
Reviewed-by: flokli <flokli@flokli.de>
Reviewed-by: tazjin <tazjin@tvl.su>
Reviewed-by: wpcarro <wpcarro@gmail.com>
This is in preparation for a subsequent CL that will do much more
significant changes in //nix/buildkite.
Change-Id: I80a8d67d3a7d593854c8d711572483c2581e7881
Reviewed-on: https://cl.tvl.fyi/c/depot/+/5824
Reviewed-by: ezemtsov <eugene.zemtsov@gmail.com>
Tested-by: BuildkiteCI
For external users of the pipeline construction, the token might be in
a different path than `/run/agenix/buildkite-graphql-token`.
It is made configurable through the BUILDKITE_TOKEN_PATH environment
variable. This should be configured on the pipeline level to apply to
all steps.
Change-Id: I23c52e2d705e4134b8b013f8603f92e5533a6e44
Reviewed-on: https://cl.tvl.fyi/c/depot/+/5424
Autosubmit: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
Reviewed-by: asmundo <asmundo@gmail.com>
There is no need for this step to be part of the static pipeline (it
should not run if the build fails anyways).
Change-Id: I71400a452d6f8f4708d146b346eaffda5da2f766
Reviewed-on: https://cl.tvl.fyi/c/depot/+/5049
Tested-by: BuildkiteCI
Autosubmit: tazjin <tazjin@tvl.su>
Reviewed-by: ezemtsov <eugene.zemtsov@gmail.com>
This will create `build-chunk-$n.json` files for steps that should run
_before_ duck, and `post-chunk-$n.json` files for steps that should
run after duck.
The post steps are not yet uploaded to Buildkite, but we also don't
have any right now.
Change-Id: I7e1b59cf55a8bf1d97266f6e988aa496959077bf
Reviewed-on: https://cl.tvl.fyi/c/depot/+/5047
Tested-by: BuildkiteCI
Reviewed-by: ezemtsov <eugene.zemtsov@gmail.com>
Autosubmit: tazjin <tazjin@tvl.su>
Using this instead of a conditional leads to nicer output in the UI,
but has no semantic difference.
Change-Id: I5b368d663f417d256e4792d2d46b84fc50d42d0e
Reviewed-on: https://cl.tvl.fyi/c/depot/+/5045
Reviewed-by: ezemtsov <eugene.zemtsov@gmail.com>
Tested-by: BuildkiteCI
Autosubmit: tazjin <tazjin@tvl.su>
This step is independent of the build result and can be scheduled at
the beginning while pipeline eval is still in progress.
Change-Id: I2ee268e4c333efa654dcb12c0b1562b43231d241
Reviewed-on: https://cl.tvl.fyi/c/depot/+/5044
Tested-by: BuildkiteCI
Autosubmit: tazjin <tazjin@tvl.su>
Reviewed-by: ezemtsov <eugene.zemtsov@gmail.com>
Previously we only stored the drvmap, but we will also need the build
chunks to refactor the generation of dynamic post-steps.
Change-Id: I256fffe13af8f8c4521835257f5d87dda323b248
Reviewed-on: https://cl.tvl.fyi/c/depot/+/5043
Tested-by: BuildkiteCI
Autosubmit: tazjin <tazjin@tvl.su>
Reviewed-by: ezemtsov <eugene.zemtsov@gmail.com>
This CI pipeline in Buildkite verifies the external (josh-provided)
view of the depot at //views/kit.
See https://buildkite.com/tvl/tvl-kit
Note that this always triggers a build of HEAD. This is because we
don't know the transformed commit ID, and we currently have no way to
pass a ref through. The pipeline is configured to skip intermediate
builds.
I asked Buildkite for some ideas on how to improve this, lets see.
Change-Id: I6c60fb1ea7606c1c90219ef04fd7bada64661529
Reviewed-on: https://cl.tvl.fyi/c/depot/+/5010
Tested-by: BuildkiteCI
Autosubmit: tazjin <tazjin@tvl.su>
Reviewed-by: sterni <sternenseemann@systemli.org>
This is no longer TVL-specific and should live here with the other
generalised stuff.
Change-Id: I95a1b4c0321f34812162d6fd40568269abf639dd
Reviewed-on: https://cl.tvl.fyi/c/depot/+/5006
Tested-by: BuildkiteCI
Autosubmit: tazjin <tazjin@tvl.su>
Reviewed-by: ezemtsov <eugene.zemtsov@gmail.com>
Removes all TVL-specific values in favour of environment variables
supplied by Buildkite.
This makes it possible to reuse this script outside of TVL.
Change-Id: Ic543bc41e4c81e65ee349ad241c515231e97ab30
Reviewed-on: https://cl.tvl.fyi/c/depot/+/5005
Tested-by: BuildkiteCI
Autosubmit: tazjin <tazjin@tvl.su>
Reviewed-by: ezemtsov <eugene.zemtsov@gmail.com>
This changes the logic for build pipeline generation to inspect
an (optional) parentTargetMap attribute which contains the derivation
map of a target commit.
Targets that existed in a parent commit with the same drv hash will be
skipped, as they are not considered to have changed.
This does not yet wire up any logic for retrieving the target map from
storage, meaning that at this commit all targets are always built.
The intention is that we will have logic to fetch the target
map (initially from Buildkite artefact storage), which we then pass to
the depot via externalArgs when actually generating the pipeline.
Change-Id: I3373c60aaf4b56b94c6ab64e2e5eef68dea9287c
Reviewed-on: https://cl.tvl.fyi/c/depot/+/4946
Tested-by: BuildkiteCI
Reviewed-by: sterni <sternenseemann@systemli.org>
Always create a structure that maps all targets to derivations, and
persist it as a JSON file.
This relates to some of the ideas expressed in:
https://docs.google.com/document/d/16A0a5oUxH1VoiSM8hyFyLW0WiUYpNo2e2D6FTW4BlH8/edit
The file is always uploaded to Buildkite as an artifact. This allows
for retrieving it based on the commit ID in a Buildkite GraphQL query.
By default, Buildkite stores artefacts for 6 months. Storage location
can be overridden (with custom retention) through some environment
variables, but for now at TVL the Buildkite-managed storage is fine.
See also: https://buildkite.com/docs/pipelines/artifacts
In the subsequent filtering implementation, when diffing commits
across a time-range that exceeds artefact retention time, we should
simply default to building everything.
Change-Id: I6d808461cd1c1fdd6983ba8c8ef075736d42caa7
Reviewed-on: https://cl.tvl.fyi/c/depot/+/3662
Tested-by: BuildkiteCI
Reviewed-by: sterni <sternenseemann@systemli.org>
This reverts commit 5e036ed9fc.
Reason for revert: This introduced a logic error since the remaining
step runs at the wrong point in the pipeline. Temporarily reverting to
having duplicated waits in order to clean up later.
Change-Id: Ifa6ece50dd22924f02efd7b790a5863ca1189af7
Reviewed-on: https://cl.tvl.fyi/c/depot/+/4841
Tested-by: BuildkiteCI
Reviewed-by: tazjin <tazjin@tvl.su>
Autosubmit: tazjin <tazjin@tvl.su>
Turns the anchor derivation into something that can actually be
built (a call creating a propagated build inputs file), and builds it.
This should fix the anchoring logic we have on canon.
Change-Id: If6a7662b82e2e396388980f65e332cf67a45b46e
Reviewed-on: https://cl.tvl.fyi/c/depot/+/4763
Tested-by: BuildkiteCI
Autosubmit: tazjin <mail@tazj.in>
Reviewed-by: sterni <sternenseemann@systemli.org>
This now happens in //nix/buildkite instead
Change-Id: Ie9e239ee4f28ac34aa4d3279dac55d70a2cb9d86
Reviewed-on: https://cl.tvl.fyi/c/depot/+/4764
Tested-by: BuildkiteCI
Reviewed-by: sterni <sternenseemann@systemli.org>
Produces more useful output and also makes for a good target for the
upcoming extraSteps logic.
Change-Id: Ifd389d433d9e27f97940a48999f4fba35646e37a
Reviewed-on: https://cl.tvl.fyi/c/depot/+/4727
Tested-by: BuildkiteCI
Autosubmit: tazjin <mail@tazj.in>
Reviewed-by: sterni <sternenseemann@systemli.org>
Extracts the logic for generating our Buildkite pipeline (which has
been copy&pasted and slightly modified in some places outside of
depot) into a generic //nix/buildkite library.
This should cause no change in functionality.
Change-Id: Iad3201713945de41279b39e4f1b847f697c179f7
Reviewed-on: https://cl.tvl.fyi/c/depot/+/4726
Autosubmit: tazjin <mail@tazj.in>
Tested-by: BuildkiteCI
Reviewed-by: sterni <sternenseemann@systemli.org>
If available, provide a link back to Gerrit on the overview page of a
build.
Uses the default style (i.e. style unset), which makes it
non-intrusive visually.
Change-Id: I4271d589d548015b75762fd0584f3958bfcc53e5
Reviewed-on: https://cl.tvl.fyi/c/depot/+/4442
Tested-by: BuildkiteCI
Reviewed-by: grfn <grfn@gws.fyi>
This fix was recommended by Buildkite and is explained in the comment.
Change-Id: I3f1c1c07cba0b417857d69c021c8af4750d645c4
Reviewed-on: https://cl.tvl.fyi/c/depot/+/4334
Tested-by: BuildkiteCI
Reviewed-by: sterni <sternenseemann@systemli.org>
The number of jobs in the depot pipeline is reaching the limits of the
Buildkite backend's ability for a single pipeline upload. Based on a
conversation with their support my understanding is that this has to
do with internal locking mechanisms at Buildkite.
To work around this, we can instead chunk the pipeline into several
smaller chunks that are uploaded serially.
This commit introduces logic to chunk the pipeline accordingly. The
chunk size chosen is 256 for now (a multiple of our number of agents,
which is useful if we can get builds from the first chunk to start
before the next ones are uploaded).
Note that this chunk size is significantly below even the current
number of targets (~460 as of this commit), but choosing a lower chunk
size might alleviate problems we've been seeing with timeouts during
pipeline uploads.
Change-Id: I77030aaf8b874c330218b78c77d15216e13b9af7
Reviewed-on: https://cl.tvl.fyi/c/depot/+/4332
Tested-by: BuildkiteCI
Reviewed-by: wpcarro <wpcarro@gmail.com>
Autosubmit: tazjin <mail@tazj.in>
We can gcroot the derivation files and drop this step, but have
elected not to do so for the moment, see cl/3436.
Change-Id: I993a1f3921e9f21e18fa260e76d3dd15ffa556bd
Reviewed-on: https://cl.tvl.fyi/c/depot/+/4327
Tested-by: BuildkiteCI
Autosubmit: sterni <sternenseemann@systemli.org>
Reviewed-by: tazjin <mail@tazj.in>
This step would get inserted at the wrong point in the build pipeline
otherwise, causing a dependency cycle and causing the pipeline to fail.
Change-Id: I534568eec77f74ae6c47276820f8a9e99493a3ea
This simplifies the fallback logic used in case of Nix evaluation
failure and makes it so that the evaluation step itself is the one
that is marked as failed in Buildkite.
This is possible because the pipeline upload command will insert new
steps at the point where it runs in the pipeline, and not later.
Change-Id: I870534c004ebc457a1602623c4e5f9c0c68e28fc
Instead of manually tracking the build status through Buildkite
metadata, use the Buildkite GraphQL API in the `🦆` build
step (i.e. the one that determines the status of the entire pipeline
to be reported back to Gerrit) to fetch the number of failed jobs.
This way we have less manual state accounting in the pipeline.
The downside is that the GraphQL query embedded here is a little hard
to read.
Notes:
* This needs an access token for Buildkite. We already have one for
besadii which is also run by the agents, so I've given it GraphQL
permissions and reused it.
* I almost introduced a very rare bug here: My initial intuition was
to simply `exit $FAILED_JOBS` - in the extremely rare case where
`$FAILED_JOBS % 256 = 0` this would mean we would ... fail to fail
the build :)
Change-Id: I61976b11b591d722494d3010a362b544efe2cb25
It seems that shell variables don't work as expected inside the
Buildkite pipeline, so usage of variables has been removed.
We also don't echo the revision anymore because of that, but it does
still appear in the log of `git push`.
Change-Id: I124e3b09af896da898f2a78715ed371651a1c5f8
Reviewed-on: https://cl.tvl.fyi/c/depot/+/3780
Tested-by: BuildkiteCI
Reviewed-by: grfn <grfn@gws.fyi>
This makes the revision number available much earlier (before the rest
of the pipeline runs, while Nix eval is happening) which should only
be a few seconds after a commit to canon.
It is also more readable in this shape.
Change-Id: Iccbb17dfef6afe68f54fda41e8d10c4dc52b08c2
Reviewed-on: https://cl.tvl.fyi/c/depot/+/3775
Tested-by: BuildkiteCI
Reviewed-by: grfn <grfn@gws.fyi>
This automatically pushes a new ref at refs/r/$revision to Gerrit
whenever a CI run completes on canon.
Revision numbers can be fetched from Gerrit with this command:
git fetch gerrit "refs/r/*:refs/r/*"
Note that this build step requires credentials to be provisioned on
the CI runner machine.
Change-Id: I37bb14346832f891240aa47bb55affaace3d5f21
We changed the configured pipeline in Buildkite to upload
`static-pipeline.yaml` instead of containing the steps of that
pipeline itself.
This makes it easier to test changes to builds and such, but adds
another build step with scheduling overhead etc.
However - we can work around this by killing one of the existing build
steps. There's no reason the failure status zeroing (required for
status reporting) shouldn't be part of the pipeline setup, so I've
moved it there instead and nuked that step.
This should mean that the pipeline is configurable from within the
repo, but without slowing anything down.
Change-Id: I206ecc02647de42a461e33c02879ab84daf5ed2b
Reviewed-on: https://cl.tvl.fyi/c/depot/+/3461
Tested-by: BuildkiteCI
Reviewed-by: sterni <sternenseemann@systemli.org>
Skip build steps if they have already been built, reducing pipelines
to the things that actually changed between builds. On canon all
targets are always built (we require this for anchoring).
Note that this is not perfect, garbage collection and competing
pipelines may affect each other.
Also note that we have some impure targets that change on every
commit.
Change-Id: Ic6bae3b6c8e1e7fd2116ec252f5089f471854ab6
Reviewed-on: https://cl.tvl.fyi/c/depot/+/3427
Tested-by: BuildkiteCI
Reviewed-by: sterni <sternenseemann@systemli.org>
Reviewed-by: grfn <grfn@gws.fyi>
We currently evaluate every target twice -- once when the depot pipeline
is built and once when actually running the build step in question. Nix
evaluation is quite slow especially given heavy use of import from
derivation in depot, so avoiding the second evaluation is desireable.
Evaluating a derivation yields a `drv` file in the nix store which can
be passed to `nix-store --realise` in order to build it eliminating the
need to wait for evaluation. We can obtain the path to the `drv` file
while building the pipeline via `target.drvPath` and remember it for the
build later.
However we need to work around a flaw (or oversight) in Nix's dependency
tracking via string context: This is based on derivations, not output
path (because this is what evaluation deals with, likely). This is no
problem per se, but an issue is that Nix can't express a dependency on
a `drv` file without any of its output paths. This means for us that we
either have to build all output paths at evaluation time (which we don't
want, obviously) or to deal with the fact that the `drv` file we need
may be garbage collected at any moment after discarding the string
context -- then nix is unable to track the reference from the pipeline
to the `drv` file in the store.
So to prevent a race condition between the pipeline and the garbage
collector we fall back to the normal nix-build invocation as we did
before.
Change-Id: I9ef8bd233085dc6e30eba54f403ea03ac2d35748
Reviewed-on: https://cl.tvl.fyi/c/depot/+/3426
Tested-by: BuildkiteCI
Reviewed-by: tazjin <mail@tazj.in>
--show-trace should make it easier to debug tricky evaluation errors
without running nix-build -A ops.pipelines.depot locally again.
Change-Id: Ice540562c3b389fc2a49ec1fc0adacb17db2a528
Reviewed-on: https://cl.tvl.fyi/c/depot/+/2947
Tested-by: BuildkiteCI
Reviewed-by: tazjin <mail@tazj.in>
This change is required to run the ⚓ step on canon builds.
Change-Id: Ib3cebac67c9f5337b27a948f120b0a9ba834ef2a
Reviewed-on: https://cl.tvl.fyi/c/depot/+/2932
Tested-by: BuildkiteCI
Reviewed-by: sterni <sternenseemann@systemli.org>
Reviewed-by: glittershark <grfn@gws.fyi>
Adds a conditional build step that only runs on the canon branch, and
only if 🦆 (the status reporting step) succeeds, which creates a
new Nix GC root for all depot targets named `depot-canon`.
In practice this might be a bit racey, as canon builds are not
guaranteed to succeed in order (though it is likely). This shouldn't
matter much in practice: We only want to prevent rebuilds of the whole
world.
This fixes b/102
Change-Id: Id3d0bf4158bffcb1ed6929888a29d31609b6ece1
Reviewed-on: https://cl.tvl.fyi/c/depot/+/2904
Tested-by: BuildkiteCI
Reviewed-by: glittershark <grfn@gws.fyi>
This file represents the static pipeline which is configured in the
Buildkite web UI. Updates to this file should be applied in the admin
interface.
These steps are responsible for launching the dynamic pipeline
evaluation, or falling back to the fallback pipeline if evaluation fails.
Change-Id: I6d7dd623cde65e8c69faea729f737c9bba00c2fb
Reviewed-on: https://cl.tvl.fyi/c/depot/+/2103
Tested-by: BuildkiteCI
Reviewed-by: glittershark <grfn@gws.fyi>
This adds a simple fallback Buildkite pipeline configuration which
always fails the pipeline, but correctly reports back the failure
status.
Note that this also requires changes in the Buildkite configuration
that is not in version-control.
Relates to b/66.
Change-Id: I6802a6f76448c3893798a06d514e6ccba0f50dd2
Reviewed-on: https://cl.tvl.fyi/c/depot/+/2102
Tested-by: BuildkiteCI
Reviewed-by: glittershark <grfn@gws.fyi>
We have naturally evolved a distinction between logical and physical
targets.
Physical targets are those which correspond directly to a tree
location on disk and can be built with `-A path.to.files`, while
logical targets are those that are exported from within an expression
but do not have a corresponding file on disk.
This change adds support for exporting logical targets from any tree
location by adding a `meta.targets` attribute containing keys into
itself, which will be consumed by the CI target gathering logic and
included in the generated pipeline.
Note that the labels for subtargets are syntactically different to
emphasise that they do not correspond to a file location. For example,
this change enables 'ops.nixos.whitbySystem' as a subtarget, which is
labeled in CI as `ops/nixos:whitbySystem`.
Change-Id: Ied09647a62c2ba98e3914548e3742ad422c63ecf
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1893
Tested-by: BuildkiteCI
Reviewed-by: glittershark <grfn@gws.fyi>
Create the pipeline by outputting a file that contains nix-build
invocations for each target's *derivation path*.
Each invocation has a generated Nix expression passed to it with `-E`
which fetches the correct target from the tree while correctly
handling targets with strange characters (such as in Go-packages).
This makes it possible to run target-level granular pipelines. We're
getting somewhere!
Change-Id: Ia6946e389dafd1d4926130bb8891446d6e17133b
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1855
Tested-by: BuildkiteCI
Reviewed-by: glittershark <grfn@gws.fyi>
Reviewed-by: lukegb <lukegb@tvl.fyi>
Automatically walk the entire depot tree and pick out things that are
"buildable", then include them in the attribute `ci.targets` (which is
now also the target for CI builds).
A long time ago, in a land far away, we (well, I, at the time) had a
prototype of this which ran into constant issues with infinite
recursions while trying to walk the tree. In fact, this is why
readTree originally gained the `__readTree`-attribute which marks
things that were imported automatically.
Based on some code edef whipped up earlier (with the breakthrough
being that we also add the attribute to top-level folders, which
suddenly resolves a whole bunch of problems), I've now implemented
this actually working version.
At the moment all builds still happen as one big bag of builds, but at
some point we will granularise this.
Change-Id: I86f12ce7f63dae98e7e5c6646a4e9d220de783f2
Reviewed-on: https://cl.tvl.fyi/c/depot/+/1854
Tested-by: BuildkiteCI
Reviewed-by: kanepyork <rikingcoding@gmail.com>
Reviewed-by: glittershark <grfn@gws.fyi>
step up in the pipeline
