Currently all blobs are uploaded serially when ingesting NARs. If a NAR
contains many, small blobs, ingestion may become slow if there is a lot
of round-trip latency to the blob service.
This makes the NAR ingester use the ConcurrentBlobUploader which allows
for buffering small blobs in memory so they can be uploaded concurrently
to the blob service without blocking further deserialization.
Change-Id: I093a73770232df12d9a11e5d901b99c08505c3cb
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11694
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Since cl/…, a PathInfoService doesn't need to implement `calculate_nar`
anymore, so most of them don't actually have a handle to a
{Blob,Directory}Service anymore.
This means, we can simplify the construction of them for test cases
a lot.
Change-Id: I100e9e1c9b00a049b4d6136c57aad4cdb04461c6
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11691
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Tested-by: BuildkiteCI
Drop the (unused) feature flags here, and get rid of some crate
dependencies.
Change-Id: Id64852b498725467d56abb12b548301cfba6a760
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11678
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
This gives multi-listener support, as well as listening on named socket
FDs.
Change-Id: I92b441f854e2faf80074463d0ca6bdc23cbd890a
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11663
Tested-by: BuildkiteCI
Reviewed-by: yuka <yuka@yuka.dev>
Autosubmit: flokli <flokli@flokli.de>
This aligns the tonic version we use directly in Tvix to the one pulled
in by the OTLP stack.
Change-Id: I658528c8dabb7cd6948f1207ddcdef1984e82037
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11666
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: yuka <yuka@yuka.dev>
Tested-by: BuildkiteCI
Make some space for the rust implementation.
Change-Id: I924dc1657be10abe5a11951c3b9de50bae06db19
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11662
Tested-by: BuildkiteCI
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: yuka <yuka@yuka.dev>
This allows querying two PathInfoService implementations sequentially,
and inserts into the "near" one if it's not there yet.
There is no negative cache, and put / listing is not implemented (for
now).
Change-Id: I24c3d0e0c3c2f0524a6cc7b2f3cbc33eb20cf92b
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11636
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: Connor Brewster <cbrewster@hey.com>
… in Cargo.toml.
This gets an imperative `cargo clippy` run to pick up that config,
so `-A clippy::blocks_in_conditions` doesn't need to be explicitly
specified anymore.
Change-Id: I32b6cc50c77c22cba0d816d0db508c2f94b2c383
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11659
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: edef <edef@edef.eu>
Tested-by: BuildkiteCI
Add instrumentation to the get() and put() implementations of all
PathInfoService.
Use the nixbase32 representation of the digest, not the base64 one.
Change-Id: Iea79bbd363bf20f23985e877c6fc1793bbee6a7e
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11630
Reviewed-by: picnoir picnoir <picnoir@alternativebit.fr>
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
This one doesn't require us to deal with poisoning, is upgradeable and
the right thing to use when locking access to data, not IO resources.
Change-Id: I78634953a73404500d28f51f1d93a87e215c8149
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11612
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: Connor Brewster <cbrewster@hey.com>
We don't want to use the std::sync::RwLock here, as it blocks.
This also means we don't need to deal with the error cases anymore.
The list() implementation is updated to use try_stream, which means we
can now avoid collecting everything into a Vec before returning from it.
Change-Id: I9057dcc410dc553e6b1be3f20d5ee830569e8218
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11611
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
This provides an implementation of PathInfoService storing PathInfo in
memory up to a certain capacity, then evicting these that have been used
the least recently.
Change-Id: I9d738687caf4f181a957f72245f26b92832313cd
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11622
Tested-by: BuildkiteCI
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Autosubmit: flokli <flokli@flokli.de>
These are not used anymore.
Change-Id: I9c348391c9600e9319f171faf3eda7175ebf7076
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11621
Tested-by: BuildkiteCI
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: Connor Brewster <cbrewster@hey.com>
These are not used anymore.
Change-Id: I6c16b4d80ddaabcb75fec3ea3e32b923b7719485
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11620
Tested-by: BuildkiteCI
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Autosubmit: flokli <flokli@flokli.de>
This shouldn't be part of the PathInfoService trait.
Pretty much none of the PathInfoServices do implement it, and requiring
them to implement it means they also cannot make use of this calculation
already being done by other PathInfoServices.
Move it out into its own NarCalculationService trait, defined somewhere
at tvix_store::nar, and have everyone who wants to trigger nar
calculation use nar_calculation_service directly, which now is an
additional field in TvixStoreIO for example.
It being moved outside the PathInfoService trait doesn't prohibit
specific implementations to implement it (like the GRPC client for the
`PathInfoService` does.
This is currently wired together in a bit of a hacky fashion - as of
now, everything uses the naive implementation that traverses blob and
directoryservice, rather than composing it properly. I want to leave
that up to a later CL, dealing with other parts of store composition
too.
Change-Id: I18d07ea4301d4a07651b8218bc5fe95e4e307208
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11619
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
This does IO, which might take a longer amount of time than what we want
to be blocking the normal executor.
Use spawn_blocking instead. I didn't add it for the constructors, as we
only call these once.
Change-Id: I9a1063099bac9582ca9681043c58c1edc780c5ff
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11618
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Tested-by: BuildkiteCI
We already have the same code in make_grpc_path_info_service_client.
Change-Id: Ibcd60831af8a061a2c3bb2f960f52a43d06cf6fa
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11615
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
We reuse the prev_name allocation for Entry, instead of allocating and
returning a separate Vec.
We encode the `prev_name: None` case as an empty vector, since we don't
allow empty names anyway, and the sorting is equivalent.
Change-Id: I975b37ff873805f5ff099bc82128706891052247
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11607
Reviewed-by: Brian Olsen <me@griff.name>
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
If our underlying reader supports AsyncBufRead, then we can too.
Change-Id: If4b948c983400ca591c1c475bbcf7dc00d562040
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11545
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Rust 1.77 supports async recursion as long as there is some form of
indirection (ie. `Box::pin`). This removes the need to use the
async-recursion crate.
Change-Id: Ic9613ab7f32016f0103032a861edff92e2fb8b41
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11596
Reviewed-by: flokli <flokli@flokli.de>
Autosubmit: Connor Brewster <cbrewster@hey.com>
Tested-by: BuildkiteCI
Rename read_nar to ingest_nar, and have it use the async nar reader
version, and the ingest_entries machinery.
This means we can now drop all code dealing with manually assembling
castore nodes.
Update our consumer, NixHTTPPathInfoService to use the new API.
As we now accept an AsyncRead, we don't need to do any blocking here
anymore, and can use the same async-compression crate as in the fetching
logic (and support some more compression formats out of the box).
Change-Id: I8646d20bd8603f8da47b5c84bc9e4ac236eb7f1a
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11580
Tested-by: BuildkiteCI
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: Connor Brewster <cbrewster@hey.com>
We implement DirectoryService for Arc<DirectoryService> and
Box<DirectoryService>, this is sufficient.
Change-Id: I0a5a81cbc4782764406b5bca57f908ace6090737
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11586
Tested-by: BuildkiteCI
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Write this a bit more compact, by using map_err(|e| …) and ?.
Ideally we'd get rid of the error mapping entirely, by using proper
error types, but that's left for a followup.
Change-Id: I68dc72b162ac89c5ff82d8c2bc26e1c808a0affd
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11584
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: Connor Brewster <cbrewster@hey.com>
The emulator and bigtable client are quite big. Remove them from the
default //tvix:shell.
Put the tests behind a `integration` feature flag, and add a variant
with that enabled to CI, and drop the bigtable tools from //tvix:shell.
Change-Id: Ie042097a0d6fc26542faa96c139b77298ccb160a
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11582
Reviewed-by: edef <edef@edef.eu>
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
The sled backend doesn't perform very well with blobs in there,
especially as it's not doing any chunking.
Switch to the `objectstore+file://` instead, which does do CDC.
Change-Id: Ic0d8836c6fc811b80c7202e3ee7f44a05a4f8dfa
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11554
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Tested-by: BuildkiteCI
Have ingest_entries return an Error type with only three kinds:
- Error while uploading a specific Directory
- Error while finalizing the directory upload
- Error from the producer
Move all ingestion method-specific errors to the individual
implementations.
Change-Id: I2a015cb7ebc96d084cbe2b809f40d1b53a15daf3
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11557
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: Connor Brewster <cbrewster@hey.com>
This wasn't really used - to ingest logs in meachine-readable form, the
OTLP infrastructure is more suitable to provide structured logs than
parsing JSON from std{err}, as it also captures span information.
Also, the non-JSON output is a bit too spammy, as remarked in cl/11483 -
change it to `compact`.
Change-Id: I48007b84ba076ab566abbb6131a02868fe0eb397
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11526
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: picnoir picnoir <picnoir@alternativebit.fr>
We can speedup uploads further by not uploading all store paths
sequentially, but in parallel.
We still don't respect the reference graph, so nothing changed here.
Change-Id: I40edb5725fe7a15170f5b9f8f4c8bec2130ecca3
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11492
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Tested-by: BuildkiteCI
Move error types and filesystem-specific functions to a separate file,
and keep the fs:: namespace in public exports.
Change-Id: I5e9e83ad78d9aea38553fafc293d3e4f8c31a8c1
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11486
Tested-by: BuildkiteCI
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Autosubmit: flokli <flokli@flokli.de>
Previously the store ingestion code was coupled to `walkdir::DirEntry`s
produced by the `walkdir` crate which made it impossible to reuse
ingesting from other sources like tarballs or NARs.
This introduces a `IngestionEntry` which carries enough information for
store ingestion and a future for computing the Blake3 digest of files.
This allows the producer to perform file uploads in a way that makes
sense for the source, ie. the filesystem upload could concurrently
upload multiple files at the same time, while the NAR ingestor will need
to ingest the entire blob before yielding the next blob in the stream.
In the future we can buffer small blobs and upload them concurrently,
but the full blob still needs to be read from the NAR before advancing.
Change-Id: I6d144063e2ba5b05e765bac1f27d41b3c8e7b283
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11462
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Also, rename the DUMMY_NAME constant in the fixtures to DUMMY_PATH,
which aligns more with the ToString representation and from_bytes
conversions we have on StorePath[Ref].
Change-Id: I39763c9dfa84c5d86f2fd0171b3a4d36fd72f267
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11464
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: Connor Brewster <cbrewster@hey.com>
This exposes `user.tvix.castore.{blob,directory}.digest` xattr keys for
files and directories:
```
❯ getfattr -d /tmp/tvix/06jrrv6wwp0nc1m7fr5bgdw012rfzfx2-nano-7.2-info
getfattr: Removing leading '/' from absolute path names
user.tvix.castore.directory.digest="b3:SuYDcUM9RpWcnA40tYB1BtYpR0xw72v3ymhKDQbBfe4="
❯ getfattr -d /tmp/tvix/156a89x10c3kaby9rgf3fi4k0p6r9wl1-etc-shells
getfattr: Removing leading '/' from absolute path names
user.tvix.castore.blob.digest="b3:pZkwZoHN+/VQ8wkaX0wYVXZ0tV/HhtKlSqiaWDK7uRs="
```
It's currently mostly used for debugging, though it might be useful for
tvix-castore-aware syncing programs using the filesystem too.
Change-Id: I26ac3cb9fe51ffbf7f880519f26741549cb5ab6a
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11422
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: raitobezarius <tvl@lahfa.xyz>
Tested-by: BuildkiteCI
Reviewed-by: Brian Olsen <me@griff.name>
This allows copying individual store paths from the local /nix/store
into tvix-store.
As we don't support getting this information from Nix yet, we currently
expect metadata to be provided externally:
Nix' `exportReferencesGraph` feature contains pretty much all data we
need for this. Expect a list of this information at a well-known key
(`closure`, similar to nixpkgs' `pkgs/build-support/binary-cache/
default.nix`).
We currently simply upload all store paths sequentially, without any
parallelism or awareness in how the reference graph looks like.
As long as the connected stores don't enforce this, this is fine, at
least for now.
Change-Id: Ib83c998465adddfdb110db994843c44e26b3d3d8
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11397
Reviewed-by: raitobezarius <tvl@lahfa.xyz>
Tested-by: BuildkiteCI
Autosubmit: flokli <flokli@flokli.de>
There's little reason to keep the nar writer using Async{Read,Write}
traits from futures, while everything else async in tvix (and
nix-compat) uses tokio.
Change-Id: I8cd1efcd0dd5bb76471de997603c7b701a5095de
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11391
Tested-by: BuildkiteCI
Reviewed-by: raitobezarius <tvl@lahfa.xyz>
Reviewed-by: Brian Olsen <me@griff.name>
This actually does a bit of allocation. There's two Vecs, one for
references and one for signatures.
We can get rid of the Vec at nar_hash.
Change-Id: Ie025309b6678f83f5b961d49ff75dcfc7da145a1
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11395
Reviewed-by: raitobezarius <tvl@lahfa.xyz>
Tested-by: BuildkiteCI
Put behind the "cloud" backend, like in the `tvix-castore` crate.
Change-Id: Ib38d198baf11ab2a4b6dc405121676147c424611
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11362
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Tested-by: BuildkiteCI
Have this return a StorePathRef<'_>, rather than a StorePath, and leave
it up to the caller to possibly convert it to a owned StorePath.
This avoids some allocations, if we only want to validate.
Change-Id: I5cf8e246fe02bd4e631f46a5cb86d3f77a728a0d
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11361
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Now, it supports almost everything except `recursive = false;`, i.e. `flat`-ingestion
because we have no knob exposed in the tvix store import side to do it.
This has been tested to work.
Change-Id: I2e9da10ceccdfbf45b43c532077ed45d6306aa98
Reviewed-on: https://cl.tvl.fyi/c/depot/+/10597
Tested-by: BuildkiteCI
Autosubmit: raitobezarius <tvl@lahfa.xyz>
Reviewed-by: flokli <flokli@flokli.de>
Instead of enforcing NAR SHA256 all the time, we generalize the
`PathInfo` constructor to take a `CAHash` argument which can drive
whether we are having a flat, NAR or text scheme.
With this, it is now possible to implement flat schemes in our
evaluation builtins, e.g. `builtins.path`.
Change-Id: I15bfee0ef4f0f428bfbd2f30c57c012cdcf6a976
Signed-off-by: Ryan Lahfa <tvl@lahfa.xyz>
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11286
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
As for grpc_pathinfoservice, we test this in the rstest-based tests,
with more implementations.
tests/utils.rs is unused now.
Change-Id: I62b192170d181e4fb432bf7dfff2057cb048b52c
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11279
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Tested-by: BuildkiteCI
Remove usage of the gen_{blob,directory}_service() helper functions
from utils. We populate Memory{Blob,Directory}Services here directly, as
test_case and rstest doesn't compose well.
Change-Id: I0fb48aadb8c818f508b18ceb83c85eb91359442a
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11278
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Tested-by: BuildkiteCI
