chore(tvix/[ca]store): Drop sled support completely in favor of redb

Over the past couple of months we've been using redb instead of sled as
the default filesystem-based database in PS and DS. I am confident that
we can get rid of sled completely now, and just keep redb.

Change-Id: I11fa1e4453e280253855f8eade990b37eb6965ae
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12567
Reviewed-by: yuka <yuka@yuka.dev>
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Autosubmit: Ilan Joselevich <personal@ilanjoselevich.com>
This commit is contained in:
Ilan Joselevich 2024-10-03 19:50:30 +02:00
parent 56b306f884
commit ab6e8d28aa
14 changed files with 18 additions and 812 deletions

87
tvix/Cargo.lock generated
View file

@ -524,12 +524,6 @@ version = "3.14.0"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7f30e7476521f6f8af1a1c4c0b8cc94f0bee37d91763d0ca2665f299b6cd8aec" checksum = "7f30e7476521f6f8af1a1c4c0b8cc94f0bee37d91763d0ca2665f299b6cd8aec"
[[package]]
name = "byteorder"
version = "1.5.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "1fd0f2584146f6f2ef48085050886acf353beff7305ebd1ae69500e27c67f64b"
[[package]] [[package]]
name = "bytes" name = "bytes"
version = "1.7.1" version = "1.7.1"
@ -1323,16 +1317,6 @@ dependencies = [
"percent-encoding", "percent-encoding",
] ]
[[package]]
name = "fs2"
version = "0.4.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "9564fc758e15025b46aa6643b1b77d047d1a56a1aea6e01002ac0c7026876213"
dependencies = [
"libc",
"winapi",
]
[[package]] [[package]]
name = "fuse-backend-rs" name = "fuse-backend-rs"
version = "0.12.0" version = "0.12.0"
@ -1463,15 +1447,6 @@ dependencies = [
"slab", "slab",
] ]
[[package]]
name = "fxhash"
version = "0.2.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c31b6d751ae2c7f11320402d34e41349dd1016f8d5d45e48c4312bc8625af50c"
dependencies = [
"byteorder",
]
[[package]] [[package]]
name = "gcp_auth" name = "gcp_auth"
version = "0.12.2" version = "0.12.2"
@ -2284,7 +2259,7 @@ dependencies = [
"lru", "lru",
"mimalloc", "mimalloc",
"nix-compat", "nix-compat",
"parking_lot 0.12.3", "parking_lot",
"prost", "prost",
"prost-build", "prost-build",
"rstest", "rstest",
@ -2514,7 +2489,7 @@ dependencies = [
"hyper", "hyper",
"itertools 0.13.0", "itertools 0.13.0",
"md-5", "md-5",
"parking_lot 0.12.3", "parking_lot",
"percent-encoding", "percent-encoding",
"quick-xml", "quick-xml",
"rand", "rand",
@ -2703,17 +2678,6 @@ version = "2.2.0"
source = "registry+https://github.com/rust-lang/crates.io-index" source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "bb813b8af86854136c6922af0598d719255ecb2179515e6e7730d468f05c9cae" checksum = "bb813b8af86854136c6922af0598d719255ecb2179515e6e7730d468f05c9cae"
[[package]]
name = "parking_lot"
version = "0.11.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7d17b78036a60663b797adeaee46f5c9dfebb86948d1255007a1d6be0271ff99"
dependencies = [
"instant",
"lock_api",
"parking_lot_core 0.8.6",
]
[[package]] [[package]]
name = "parking_lot" name = "parking_lot"
version = "0.12.3" version = "0.12.3"
@ -2721,21 +2685,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f1bf18183cf54e8d6059647fc3063646a1801cf30896933ec2311622cc4b9a27" checksum = "f1bf18183cf54e8d6059647fc3063646a1801cf30896933ec2311622cc4b9a27"
dependencies = [ dependencies = [
"lock_api", "lock_api",
"parking_lot_core 0.9.9", "parking_lot_core",
]
[[package]]
name = "parking_lot_core"
version = "0.8.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "60a2cfe6f0ad2bfc16aefa463b497d5c7a5ecd44a23efa72aa342d90177356dc"
dependencies = [
"cfg-if",
"instant",
"libc",
"redox_syscall 0.2.16",
"smallvec",
"winapi",
] ]
[[package]] [[package]]
@ -3233,15 +3183,6 @@ dependencies = [
"libc", "libc",
] ]
[[package]]
name = "redox_syscall"
version = "0.2.16"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "fb5a58c1855b4b6819d59012155603f0b22ad30cad752600aadfcb695265519a"
dependencies = [
"bitflags 1.3.2",
]
[[package]] [[package]]
name = "redox_syscall" name = "redox_syscall"
version = "0.3.5" version = "0.3.5"
@ -3842,22 +3783,6 @@ dependencies = [
"autocfg", "autocfg",
] ]
[[package]]
name = "sled"
version = "0.34.7"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7f96b4737c2ce5987354855aed3797279def4ebf734436c6aa4552cf8e169935"
dependencies = [
"crc32fast",
"crossbeam-epoch",
"crossbeam-utils",
"fs2",
"fxhash",
"libc",
"log",
"parking_lot 0.11.2",
]
[[package]] [[package]]
name = "smallvec" name = "smallvec"
version = "1.13.1" version = "1.13.1"
@ -4698,7 +4623,7 @@ dependencies = [
"lazy_static", "lazy_static",
"libc", "libc",
"object_store", "object_store",
"parking_lot 0.12.3", "parking_lot",
"petgraph", "petgraph",
"pin-project", "pin-project",
"pin-project-lite", "pin-project-lite",
@ -4712,7 +4637,6 @@ dependencies = [
"serde_qs", "serde_qs",
"serde_tagged", "serde_tagged",
"serde_with", "serde_with",
"sled",
"tempfile", "tempfile",
"thiserror", "thiserror",
"threadpool", "threadpool",
@ -4895,7 +4819,7 @@ dependencies = [
"lru", "lru",
"mimalloc", "mimalloc",
"nix-compat", "nix-compat",
"parking_lot 0.12.3", "parking_lot",
"pin-project-lite", "pin-project-lite",
"prost", "prost",
"prost-build", "prost-build",
@ -4909,7 +4833,6 @@ dependencies = [
"serde_qs", "serde_qs",
"serde_with", "serde_with",
"sha2", "sha2",
"sled",
"tempfile", "tempfile",
"thiserror", "thiserror",
"tokio", "tokio",

View file

@ -1911,19 +1911,6 @@ rec {
}; };
resolvedDefaultFeatures = [ "default" ]; resolvedDefaultFeatures = [ "default" ];
}; };
"byteorder" = rec {
crateName = "byteorder";
version = "1.5.0";
edition = "2021";
sha256 = "0jzncxyf404mwqdbspihyzpkndfgda450l0893pz5xj685cg5l0z";
authors = [
"Andrew Gallant <jamslam@gmail.com>"
];
features = {
"default" = [ "std" ];
};
resolvedDefaultFeatures = [ "default" "std" ];
};
"bytes" = rec { "bytes" = rec {
crateName = "bytes"; crateName = "bytes";
version = "1.7.1"; version = "1.7.1";
@ -2820,7 +2807,7 @@ rec {
"nightly" = [ "crossbeam-utils/nightly" ]; "nightly" = [ "crossbeam-utils/nightly" ];
"std" = [ "alloc" "crossbeam-utils/std" ]; "std" = [ "alloc" "crossbeam-utils/std" ];
}; };
resolvedDefaultFeatures = [ "alloc" "default" "std" ]; resolvedDefaultFeatures = [ "alloc" "std" ];
}; };
"crossbeam-utils" = rec { "crossbeam-utils" = rec {
crateName = "crossbeam-utils"; crateName = "crossbeam-utils";
@ -4149,29 +4136,6 @@ rec {
}; };
resolvedDefaultFeatures = [ "alloc" "default" "std" ]; resolvedDefaultFeatures = [ "alloc" "default" "std" ];
}; };
"fs2" = rec {
crateName = "fs2";
version = "0.4.3";
edition = "2015";
sha256 = "04v2hwk7035c088f19mfl5b1lz84gnvv2hv6m935n0hmirszqr4m";
authors = [
"Dan Burkert <dan@danburkert.com>"
];
dependencies = [
{
name = "libc";
packageId = "libc";
target = { target, features }: (target."unix" or false);
}
{
name = "winapi";
packageId = "winapi";
target = { target, features }: (target."windows" or false);
features = [ "handleapi" "processthreadsapi" "winerror" "fileapi" "winbase" "std" ];
}
];
};
"fuse-backend-rs" = rec { "fuse-backend-rs" = rec {
crateName = "fuse-backend-rs"; crateName = "fuse-backend-rs";
version = "0.12.0"; version = "0.12.0";
@ -4625,23 +4589,6 @@ rec {
}; };
resolvedDefaultFeatures = [ "alloc" "async-await" "async-await-macro" "channel" "default" "futures-channel" "futures-io" "futures-macro" "futures-sink" "io" "memchr" "sink" "slab" "std" ]; resolvedDefaultFeatures = [ "alloc" "async-await" "async-await-macro" "channel" "default" "futures-channel" "futures-io" "futures-macro" "futures-sink" "io" "memchr" "sink" "slab" "std" ];
}; };
"fxhash" = rec {
crateName = "fxhash";
version = "0.2.1";
edition = "2015";
sha256 = "037mb9ichariqi45xm6mz0b11pa92gj38ba0409z3iz239sns6y3";
libPath = "lib.rs";
authors = [
"cbreeden <github@u.breeden.cc>"
];
dependencies = [
{
name = "byteorder";
packageId = "byteorder";
}
];
};
"gcp_auth" = rec { "gcp_auth" = rec {
crateName = "gcp_auth"; crateName = "gcp_auth";
version = "0.12.2"; version = "0.12.2";
@ -7087,7 +7034,7 @@ rec {
} }
{ {
name = "parking_lot"; name = "parking_lot";
packageId = "parking_lot 0.12.3"; packageId = "parking_lot";
} }
{ {
name = "prost"; name = "prost";
@ -7877,7 +7824,7 @@ rec {
} }
{ {
name = "parking_lot"; name = "parking_lot";
packageId = "parking_lot 0.12.3"; packageId = "parking_lot";
} }
{ {
name = "percent-encoding"; name = "percent-encoding";
@ -8625,40 +8572,7 @@ rec {
"loom" = [ "dep:loom" ]; "loom" = [ "dep:loom" ];
}; };
}; };
"parking_lot 0.11.2" = rec { "parking_lot" = rec {
crateName = "parking_lot";
version = "0.11.2";
edition = "2018";
sha256 = "16gzf41bxmm10x82bla8d6wfppy9ym3fxsmdjyvn61m66s0bf5vx";
authors = [
"Amanieu d'Antras <amanieu@gmail.com>"
];
dependencies = [
{
name = "instant";
packageId = "instant";
}
{
name = "lock_api";
packageId = "lock_api";
}
{
name = "parking_lot_core";
packageId = "parking_lot_core 0.8.6";
}
];
features = {
"arc_lock" = [ "lock_api/arc_lock" ];
"deadlock_detection" = [ "parking_lot_core/deadlock_detection" ];
"nightly" = [ "parking_lot_core/nightly" "lock_api/nightly" ];
"owning_ref" = [ "lock_api/owning_ref" ];
"serde" = [ "lock_api/serde" ];
"stdweb" = [ "instant/stdweb" ];
"wasm-bindgen" = [ "instant/wasm-bindgen" ];
};
resolvedDefaultFeatures = [ "default" ];
};
"parking_lot 0.12.3" = rec {
crateName = "parking_lot"; crateName = "parking_lot";
version = "0.12.3"; version = "0.12.3";
edition = "2021"; edition = "2021";
@ -8673,7 +8587,7 @@ rec {
} }
{ {
name = "parking_lot_core"; name = "parking_lot_core";
packageId = "parking_lot_core 0.9.9"; packageId = "parking_lot_core";
} }
]; ];
features = { features = {
@ -8685,52 +8599,7 @@ rec {
}; };
resolvedDefaultFeatures = [ "default" ]; resolvedDefaultFeatures = [ "default" ];
}; };
"parking_lot_core 0.8.6" = rec { "parking_lot_core" = rec {
crateName = "parking_lot_core";
version = "0.8.6";
edition = "2018";
sha256 = "1p2nfcbr0b9lm9rglgm28k6mwyjwgm4knipsmqbgqaxdy3kcz8k0";
authors = [
"Amanieu d'Antras <amanieu@gmail.com>"
];
dependencies = [
{
name = "cfg-if";
packageId = "cfg-if";
}
{
name = "instant";
packageId = "instant";
}
{
name = "libc";
packageId = "libc";
target = { target, features }: (target."unix" or false);
}
{
name = "redox_syscall";
packageId = "redox_syscall 0.2.16";
target = { target, features }: ("redox" == target."os" or null);
}
{
name = "smallvec";
packageId = "smallvec";
}
{
name = "winapi";
packageId = "winapi";
target = { target, features }: (target."windows" or false);
features = [ "winnt" "ntstatus" "minwindef" "winerror" "winbase" "errhandlingapi" "handleapi" ];
}
];
features = {
"backtrace" = [ "dep:backtrace" ];
"deadlock_detection" = [ "petgraph" "thread-id" "backtrace" ];
"petgraph" = [ "dep:petgraph" ];
"thread-id" = [ "dep:thread-id" ];
};
};
"parking_lot_core 0.9.9" = rec {
crateName = "parking_lot_core"; crateName = "parking_lot_core";
version = "0.9.9"; version = "0.9.9";
edition = "2018"; edition = "2018";
@ -10207,23 +10076,6 @@ rec {
"python" = [ "dep:pyo3" "dep:pyo3-build-config" ]; "python" = [ "dep:pyo3" "dep:pyo3-build-config" ];
}; };
}; };
"redox_syscall 0.2.16" = rec {
crateName = "redox_syscall";
version = "0.2.16";
edition = "2018";
sha256 = "16jicm96kjyzm802cxdd1k9jmcph0db1a4lhslcnhjsvhp0mhnpv";
libName = "syscall";
authors = [
"Jeremy Soller <jackpot51@gmail.com>"
];
dependencies = [
{
name = "bitflags";
packageId = "bitflags 1.3.2";
}
];
};
"redox_syscall 0.3.5" = rec { "redox_syscall 0.3.5" = rec {
crateName = "redox_syscall"; crateName = "redox_syscall";
version = "0.3.5"; version = "0.3.5";
@ -12340,69 +12192,6 @@ rec {
}; };
resolvedDefaultFeatures = [ "default" "std" ]; resolvedDefaultFeatures = [ "default" "std" ];
}; };
"sled" = rec {
crateName = "sled";
version = "0.34.7";
edition = "2018";
sha256 = "0dcr2s7cylj5mb33ci3kpx7fz797jwvysnl5airrir9cgirv95kz";
authors = [
"Tyler Neely <t@jujit.su>"
];
dependencies = [
{
name = "crc32fast";
packageId = "crc32fast";
}
{
name = "crossbeam-epoch";
packageId = "crossbeam-epoch";
}
{
name = "crossbeam-utils";
packageId = "crossbeam-utils";
}
{
name = "fs2";
packageId = "fs2";
target = { target, features }: (("linux" == target."os" or null) || ("macos" == target."os" or null) || ("windows" == target."os" or null));
}
{
name = "fxhash";
packageId = "fxhash";
}
{
name = "libc";
packageId = "libc";
}
{
name = "log";
packageId = "log";
}
{
name = "parking_lot";
packageId = "parking_lot 0.11.2";
}
];
devDependencies = [
{
name = "log";
packageId = "log";
}
];
features = {
"backtrace" = [ "dep:backtrace" ];
"color-backtrace" = [ "dep:color-backtrace" ];
"compression" = [ "zstd" ];
"default" = [ "no_metrics" ];
"io_uring" = [ "rio" ];
"no_logs" = [ "log/max_level_off" ];
"pretty_backtrace" = [ "color-backtrace" ];
"rio" = [ "dep:rio" ];
"testing" = [ "event_log" "lock_free_delays" "compression" "failpoints" "backtrace" ];
"zstd" = [ "dep:zstd" ];
};
resolvedDefaultFeatures = [ "default" "no_metrics" ];
};
"smallvec" = rec { "smallvec" = rec {
crateName = "smallvec"; crateName = "smallvec";
version = "1.13.1"; version = "1.13.1";
@ -15409,7 +15198,7 @@ rec {
} }
{ {
name = "parking_lot"; name = "parking_lot";
packageId = "parking_lot 0.12.3"; packageId = "parking_lot";
} }
{ {
name = "petgraph"; name = "petgraph";
@ -15448,10 +15237,6 @@ rec {
name = "serde_with"; name = "serde_with";
packageId = "serde_with"; packageId = "serde_with";
} }
{
name = "sled";
packageId = "sled";
}
{ {
name = "thiserror"; name = "thiserror";
packageId = "thiserror"; packageId = "thiserror";
@ -16212,7 +15997,7 @@ rec {
} }
{ {
name = "parking_lot"; name = "parking_lot";
packageId = "parking_lot 0.12.3"; packageId = "parking_lot";
} }
{ {
name = "pin-project-lite"; name = "pin-project-lite";
@ -16257,10 +16042,6 @@ rec {
name = "sha2"; name = "sha2";
packageId = "sha2"; packageId = "sha2";
} }
{
name = "sled";
packageId = "sled";
}
{ {
name = "thiserror"; name = "thiserror";
packageId = "thiserror"; packageId = "thiserror";
@ -17930,7 +17711,7 @@ rec {
features = { features = {
"debug" = [ "impl-debug" ]; "debug" = [ "impl-debug" ];
}; };
resolvedDefaultFeatures = [ "basetsd" "consoleapi" "errhandlingapi" "fileapi" "handleapi" "knownfolders" "minwindef" "ntstatus" "objbase" "processenv" "processthreadsapi" "shellapi" "shlobj" "std" "stringapiset" "synchapi" "sysinfoapi" "winbase" "wincon" "winerror" "winnt" "winuser" ]; resolvedDefaultFeatures = [ "basetsd" "consoleapi" "errhandlingapi" "fileapi" "handleapi" "knownfolders" "minwindef" "objbase" "processenv" "shellapi" "shlobj" "std" "stringapiset" "synchapi" "sysinfoapi" "winbase" "wincon" "winerror" "winnt" "winuser" ];
}; };
"winapi-i686-pc-windows-gnu" = rec { "winapi-i686-pc-windows-gnu" = rec {
crateName = "winapi-i686-pc-windows-gnu"; crateName = "winapi-i686-pc-windows-gnu";

View file

@ -123,7 +123,6 @@ serde_tagged = "0.3.0"
serde_with = "3.9.0" serde_with = "3.9.0"
sha1 = "0.10.6" sha1 = "0.10.6"
sha2 = "0.10.8" sha2 = "0.10.8"
sled = "0.34.7"
smol_str = "0.2.2" smol_str = "0.2.2"
tabwriter = "1.4" tabwriter = "1.4"
tempfile = "3.12.0" tempfile = "3.12.0"

View file

@ -36,8 +36,8 @@ Now, spin up tvix-daemon, connecting to some (local) backends:
``` ```
tvix-store --otlp=false daemon \ tvix-store --otlp=false daemon \
--blob-service-addr=objectstore+file://$PWD/blobs \ --blob-service-addr=objectstore+file://$PWD/blobs \
--directory-service-addr=sled://$PWD/directories.sled \ --directory-service-addr=redb://$PWD/directories.redb \
--path-info-service-addr=sled://$PWD/pathinfo.sled & --path-info-service-addr=redb://$PWD/pathinfo.redb &
``` ```
Copy some data into tvix-store (we use `nar-bridge` for this for now): Copy some data into tvix-store (we use `nar-bridge` for this for now):
@ -58,8 +58,8 @@ In case you want to have `tvix-store virtiofs` open the stores directly, kill
``` ```
pkill tvix-store pkill tvix-store
export BLOB_SERVICE_ADDR=objectstore+file://$PWD/blobs export BLOB_SERVICE_ADDR=objectstore+file://$PWD/blobs
export DIRECTORY_SERVICE_ADDR=sled://$PWD/directories.sled export DIRECTORY_SERVICE_ADDR=redb://$PWD/directories.redb
export PATH_INFO_SERVICE_ADDR=sled://$PWD/pathinfo.sled export PATH_INFO_SERVICE_ADDR=redb://$PWD/pathinfo.redb
``` ```
#### Interactive shell #### Interactive shell

View file

@ -19,7 +19,6 @@ object_store = { workspace = true, features = ["http"] }
parking_lot = { workspace = true } parking_lot = { workspace = true }
pin-project-lite = { workspace = true } pin-project-lite = { workspace = true }
prost = { workspace = true } prost = { workspace = true }
sled = { workspace = true }
thiserror = { workspace = true } thiserror = { workspace = true }
tokio-stream = { workspace = true, features = ["fs", "net"] } tokio-stream = { workspace = true, features = ["fs", "net"] }
tokio-util = { workspace = true, features = ["io", "io-util", "codec"] } tokio-util = { workspace = true, features = ["io", "io-util", "codec"] }

View file

@ -13,10 +13,6 @@ use super::DirectoryService;
/// The following URIs are supported: /// The following URIs are supported:
/// - `memory:` /// - `memory:`
/// Uses a in-memory implementation. /// Uses a in-memory implementation.
/// - `sled:`
/// Uses a in-memory sled implementation.
/// - `sled:///absolute/path/to/somewhere`
/// Uses sled, using a path on the disk for persistency. Can be only opened
/// from one process at the same time. /// from one process at the same time.
/// - `redb:` /// - `redb:`
/// Uses a in-memory redb implementation. /// Uses a in-memory redb implementation.
@ -55,8 +51,6 @@ mod tests {
use tempfile::TempDir; use tempfile::TempDir;
lazy_static! { lazy_static! {
static ref TMPDIR_SLED_1: TempDir = TempDir::new().unwrap();
static ref TMPDIR_SLED_2: TempDir = TempDir::new().unwrap();
static ref TMPDIR_REDB_1: TempDir = TempDir::new().unwrap(); static ref TMPDIR_REDB_1: TempDir = TempDir::new().unwrap();
static ref TMPDIR_REDB_2: TempDir = TempDir::new().unwrap(); static ref TMPDIR_REDB_2: TempDir = TempDir::new().unwrap();
} }
@ -64,16 +58,6 @@ mod tests {
#[rstest] #[rstest]
/// This uses an unsupported scheme. /// This uses an unsupported scheme.
#[case::unsupported_scheme("http://foo.example/test", false)] #[case::unsupported_scheme("http://foo.example/test", false)]
/// This configures sled in temporary mode.
#[case::sled_valid_temporary("sled://", true)]
/// This configures sled with /, which should fail.
#[case::sled_invalid_root("sled:///", false)]
/// This configures sled with a host, not path, which should fail.
#[case::sled_invalid_host("sled://foo.example", false)]
/// This configures sled with a valid path path, which should succeed.
#[case::sled_valid_path(&format!("sled://{}", &TMPDIR_SLED_1.path().to_str().unwrap()), true)]
/// This configures sled with a host, and a valid path path, which should fail.
#[case::sled_invalid_host_with_valid_path(&format!("sled://foo.example{}", &TMPDIR_SLED_2.path().to_str().unwrap()), false)]
/// This correctly sets the scheme, and doesn't set a path. /// This correctly sets the scheme, and doesn't set a path.
#[case::memory_valid("memory://", true)] #[case::memory_valid("memory://", true)]
/// This sets a memory url host to `foo` /// This sets a memory url host to `foo`

View file

@ -12,7 +12,6 @@ mod object_store;
mod order_validator; mod order_validator;
mod redb; mod redb;
mod simple_putter; mod simple_putter;
mod sled;
#[cfg(test)] #[cfg(test)]
pub mod tests; pub mod tests;
mod traverse; mod traverse;
@ -27,7 +26,6 @@ pub use self::object_store::{ObjectStoreDirectoryService, ObjectStoreDirectorySe
pub use self::order_validator::{LeavesToRootValidator, OrderValidator, RootToLeavesValidator}; pub use self::order_validator::{LeavesToRootValidator, OrderValidator, RootToLeavesValidator};
pub use self::redb::{RedbDirectoryService, RedbDirectoryServiceConfig}; pub use self::redb::{RedbDirectoryService, RedbDirectoryServiceConfig};
pub use self::simple_putter::SimplePutter; pub use self::simple_putter::SimplePutter;
pub use self::sled::{SledDirectoryService, SledDirectoryServiceConfig};
pub use self::traverse::descend_to; pub use self::traverse::descend_to;
pub use self::utils::traverse_directory; pub use self::utils::traverse_directory;
@ -138,7 +136,6 @@ pub(crate) fn register_directory_services(reg: &mut Registry) {
reg.register::<Box<dyn ServiceBuilder<Output = dyn DirectoryService>>, super::directoryservice::MemoryDirectoryServiceConfig>("memory"); reg.register::<Box<dyn ServiceBuilder<Output = dyn DirectoryService>>, super::directoryservice::MemoryDirectoryServiceConfig>("memory");
reg.register::<Box<dyn ServiceBuilder<Output = dyn DirectoryService>>, super::directoryservice::CacheConfig>("cache"); reg.register::<Box<dyn ServiceBuilder<Output = dyn DirectoryService>>, super::directoryservice::CacheConfig>("cache");
reg.register::<Box<dyn ServiceBuilder<Output = dyn DirectoryService>>, super::directoryservice::GRPCDirectoryServiceConfig>("grpc"); reg.register::<Box<dyn ServiceBuilder<Output = dyn DirectoryService>>, super::directoryservice::GRPCDirectoryServiceConfig>("grpc");
reg.register::<Box<dyn ServiceBuilder<Output = dyn DirectoryService>>, super::directoryservice::SledDirectoryServiceConfig>("sled");
reg.register::<Box<dyn ServiceBuilder<Output = dyn DirectoryService>>, super::directoryservice::RedbDirectoryServiceConfig>("redb"); reg.register::<Box<dyn ServiceBuilder<Output = dyn DirectoryService>>, super::directoryservice::RedbDirectoryServiceConfig>("redb");
#[cfg(feature = "cloud")] #[cfg(feature = "cloud")]
{ {

View file

@ -1,263 +0,0 @@
use futures::stream::BoxStream;
use prost::Message;
use std::ops::Deref;
use std::path::Path;
use std::sync::Arc;
use tonic::async_trait;
use tracing::{instrument, warn};
use super::utils::traverse_directory;
use super::{Directory, DirectoryGraph, DirectoryPutter, DirectoryService, LeavesToRootValidator};
use crate::composition::{CompositionContext, ServiceBuilder};
use crate::{proto, B3Digest, Error};
#[derive(Clone)]
pub struct SledDirectoryService {
db: sled::Db,
}
impl SledDirectoryService {
pub fn new<P: AsRef<Path>>(p: P) -> Result<Self, sled::Error> {
if p.as_ref() == Path::new("/") {
return Err(sled::Error::Unsupported(
"cowardly refusing to open / with sled".to_string(),
));
}
let config = sled::Config::default()
.use_compression(false) // is a required parameter
.path(p);
let db = config.open()?;
Ok(Self { db })
}
pub fn new_temporary() -> Result<Self, sled::Error> {
let config = sled::Config::default().temporary(true);
let db = config.open()?;
Ok(Self { db })
}
}
#[async_trait]
impl DirectoryService for SledDirectoryService {
#[instrument(skip(self, digest), fields(directory.digest = %digest))]
async fn get(&self, digest: &B3Digest) -> Result<Option<Directory>, Error> {
let resp = tokio::task::spawn_blocking({
let db = self.db.clone();
let digest = digest.clone();
move || db.get(digest.as_slice())
})
.await?
.map_err(|e| {
warn!("failed to retrieve directory: {}", e);
Error::StorageError(format!("failed to retrieve directory: {}", e))
})?;
match resp {
// The directory was not found, return
None => Ok(None),
// The directory was found, try to parse the data as Directory message
Some(data) => match proto::Directory::decode(&*data) {
Ok(directory) => {
// Validate the retrieved Directory indeed has the
// digest we expect it to have, to detect corruptions.
let actual_digest = directory.digest();
if actual_digest != *digest {
return Err(Error::StorageError(format!(
"requested directory with digest {}, but got {}",
digest, actual_digest
)));
}
let directory = directory.try_into().map_err(|e| {
warn!("failed to retrieve directory: {}", e);
Error::StorageError(format!("failed to retrieve directory: {}", e))
})?;
Ok(Some(directory))
}
Err(e) => {
warn!("unable to parse directory {}: {}", digest, e);
Err(Error::StorageError(e.to_string()))
}
},
}
}
#[instrument(skip(self, directory), fields(directory.digest = %directory.digest()))]
async fn put(&self, directory: Directory) -> Result<B3Digest, Error> {
tokio::task::spawn_blocking({
let db = self.db.clone();
move || {
let digest = directory.digest();
// store it
db.insert(
digest.as_slice(),
proto::Directory::from(directory).encode_to_vec(),
)
.map_err(|e| Error::StorageError(e.to_string()))?;
Ok(digest)
}
})
.await?
}
#[instrument(skip_all, fields(directory.digest = %root_directory_digest))]
fn get_recursive(
&self,
root_directory_digest: &B3Digest,
) -> BoxStream<'static, Result<Directory, Error>> {
traverse_directory(self.clone(), root_directory_digest)
}
#[instrument(skip_all)]
fn put_multiple_start(&self) -> Box<(dyn DirectoryPutter + 'static)>
where
Self: Clone,
{
Box::new(SledDirectoryPutter {
tree: self.db.deref().clone(),
directory_validator: Some(Default::default()),
})
}
}
#[derive(serde::Deserialize)]
#[serde(deny_unknown_fields)]
pub struct SledDirectoryServiceConfig {
is_temporary: bool,
#[serde(default)]
/// required when is_temporary = false
path: Option<String>,
}
impl TryFrom<url::Url> for SledDirectoryServiceConfig {
type Error = Box<dyn std::error::Error + Send + Sync>;
fn try_from(url: url::Url) -> Result<Self, Self::Error> {
// sled doesn't support host, and a path can be provided (otherwise
// it'll live in memory only).
if url.has_host() {
return Err(Error::StorageError("no host allowed".to_string()).into());
}
// TODO: expose compression and other parameters as URL parameters?
Ok(if url.path().is_empty() {
SledDirectoryServiceConfig {
is_temporary: true,
path: None,
}
} else {
SledDirectoryServiceConfig {
is_temporary: false,
path: Some(url.path().to_string()),
}
})
}
}
#[async_trait]
impl ServiceBuilder for SledDirectoryServiceConfig {
type Output = dyn DirectoryService;
async fn build<'a>(
&'a self,
_instance_name: &str,
_context: &CompositionContext,
) -> Result<Arc<dyn DirectoryService>, Box<dyn std::error::Error + Send + Sync + 'static>> {
match self {
SledDirectoryServiceConfig {
is_temporary: true,
path: None,
} => Ok(Arc::new(SledDirectoryService::new_temporary()?)),
SledDirectoryServiceConfig {
is_temporary: true,
path: Some(_),
} => Err(Error::StorageError(
"Temporary SledDirectoryService can not have path".into(),
)
.into()),
SledDirectoryServiceConfig {
is_temporary: false,
path: None,
} => Err(Error::StorageError("SledDirectoryService is missing path".into()).into()),
SledDirectoryServiceConfig {
is_temporary: false,
path: Some(path),
} => Ok(Arc::new(SledDirectoryService::new(path)?)),
}
}
}
/// Buffers Directory messages to be uploaded and inserts them in a batch
/// transaction on close.
pub struct SledDirectoryPutter {
tree: sled::Tree,
/// The directories (inside the directory validator) that we insert later,
/// or None, if they were already inserted.
directory_validator: Option<DirectoryGraph<LeavesToRootValidator>>,
}
#[async_trait]
impl DirectoryPutter for SledDirectoryPutter {
#[instrument(level = "trace", skip_all, fields(directory.digest=%directory.digest()), err)]
async fn put(&mut self, directory: Directory) -> Result<(), Error> {
match self.directory_validator {
None => return Err(Error::StorageError("already closed".to_string())),
Some(ref mut validator) => {
validator
.add(directory)
.map_err(|e| Error::StorageError(e.to_string()))?;
}
}
Ok(())
}
#[instrument(level = "trace", skip_all, ret, err)]
async fn close(&mut self) -> Result<B3Digest, Error> {
match self.directory_validator.take() {
None => Err(Error::InvalidRequest("already closed".to_string())),
Some(validator) => {
// Insert all directories as a batch.
tokio::task::spawn_blocking({
let tree = self.tree.clone();
move || {
// retrieve the validated directories.
let directories = validator
.validate()
.map_err(|e| Error::StorageError(e.to_string()))?
.drain_leaves_to_root()
.collect::<Vec<_>>();
// Get the root digest, which is at the end (cf. insertion order)
let root_digest = directories
.last()
.ok_or_else(|| Error::InvalidRequest("got no directories".to_string()))?
.digest();
let mut batch = sled::Batch::default();
for directory in directories {
batch.insert(
directory.digest().as_slice(),
proto::Directory::from(directory).encode_to_vec(),
);
}
tree.apply_batch(batch).map_err(|e| {
Error::StorageError(format!("unable to apply batch: {}", e))
})?;
Ok(root_digest)
}
})
.await?
}
}
}
}

View file

@ -23,7 +23,6 @@ use self::utils::make_grpc_directory_service_client;
#[rstest] #[rstest]
#[case::grpc(make_grpc_directory_service_client().await)] #[case::grpc(make_grpc_directory_service_client().await)]
#[case::memory(directoryservice::from_addr("memory://").await.unwrap())] #[case::memory(directoryservice::from_addr("memory://").await.unwrap())]
#[case::sled(directoryservice::from_addr("sled://").await.unwrap())]
#[case::redb(directoryservice::from_addr("redb://").await.unwrap())] #[case::redb(directoryservice::from_addr("redb://").await.unwrap())]
#[case::objectstore(directoryservice::from_addr("objectstore+memory://").await.unwrap())] #[case::objectstore(directoryservice::from_addr("objectstore+memory://").await.unwrap())]
#[cfg_attr(all(feature = "cloud", feature = "integration"), case::bigtable(directoryservice::from_addr("bigtable://instance-1?project_id=project-1&table_name=table-1&family_name=cf1").await.unwrap()))] #[cfg_attr(all(feature = "cloud", feature = "integration"), case::bigtable(directoryservice::from_addr("bigtable://instance-1?project_id=project-1&table_name=table-1&family_name=cf1").await.unwrap()))]

View file

@ -25,7 +25,6 @@ serde_json = { workspace = true }
serde_with = { workspace = true } serde_with = { workspace = true }
serde_qs = { workspace = true } serde_qs = { workspace = true }
sha2 = { workspace = true } sha2 = { workspace = true }
sled = { workspace = true }
thiserror = { workspace = true } thiserror = { workspace = true }
tokio = { workspace = true, features = ["fs", "macros", "net", "rt", "rt-multi-thread", "signal"] } tokio = { workspace = true, features = ["fs", "macros", "net", "rt", "rt-multi-thread", "signal"] }
tokio-listener = { workspace = true, features = ["clap", "multi-listener", "sd_listen", "tonic012"] } tokio-listener = { workspace = true, features = ["clap", "multi-listener", "sd_listen", "tonic012"] }

View file

@ -12,11 +12,6 @@ use url::Url;
/// The following URIs are supported: /// The following URIs are supported:
/// - `memory:` /// - `memory:`
/// Uses a in-memory implementation. /// Uses a in-memory implementation.
/// - `sled:`
/// Uses a in-memory sled implementation.
/// - `sled:///absolute/path/to/somewhere`
/// Uses sled, using a path on the disk for persistency. Can be only opened
/// from one process at the same time.
/// - `redb:` /// - `redb:`
/// Uses a in-memory redb implementation. /// Uses a in-memory redb implementation.
/// - `redb:///absolute/path/to/somewhere` /// - `redb:///absolute/path/to/somewhere`
@ -66,8 +61,6 @@ mod tests {
use tvix_castore::directoryservice::{DirectoryService, MemoryDirectoryServiceConfig}; use tvix_castore::directoryservice::{DirectoryService, MemoryDirectoryServiceConfig};
lazy_static! { lazy_static! {
static ref TMPDIR_SLED_1: TempDir = TempDir::new().unwrap();
static ref TMPDIR_SLED_2: TempDir = TempDir::new().unwrap();
static ref TMPDIR_REDB_1: TempDir = TempDir::new().unwrap(); static ref TMPDIR_REDB_1: TempDir = TempDir::new().unwrap();
static ref TMPDIR_REDB_2: TempDir = TempDir::new().unwrap(); static ref TMPDIR_REDB_2: TempDir = TempDir::new().unwrap();
} }
@ -77,16 +70,6 @@ mod tests {
#[rstest] #[rstest]
/// This uses a unsupported scheme. /// This uses a unsupported scheme.
#[case::unsupported_scheme("http://foo.example/test", false)] #[case::unsupported_scheme("http://foo.example/test", false)]
/// This configures sled in temporary mode.
#[case::sled_temporary("sled://", true)]
/// This configures sled with /, which should fail.
#[case::sled_invalid_root("sled:///", false)]
/// This configures sled with a host, not path, which should fail.
#[case::sled_invalid_host("sled://foo.example", false)]
/// This configures sled with a valid path path, which should succeed.
#[case::sled_valid_path(&format!("sled://{}", &TMPDIR_SLED_1.path().to_str().unwrap()), true)]
/// This configures sled with a host, and a valid path path, which should fail.
#[case::sled_invalid_host_with_valid_path(&format!("sled://foo.example{}", &TMPDIR_SLED_2.path().to_str().unwrap()), false)]
/// This correctly sets the scheme, and doesn't set a path. /// This correctly sets the scheme, and doesn't set a path.
#[case::memory_valid("memory://", true)] #[case::memory_valid("memory://", true)]
/// This sets a memory url host to `foo` /// This sets a memory url host to `foo`

View file

@ -6,7 +6,6 @@ mod memory;
mod nix_http; mod nix_http;
mod redb; mod redb;
mod signing_wrapper; mod signing_wrapper;
mod sled;
#[cfg(any(feature = "fuse", feature = "virtiofs"))] #[cfg(any(feature = "fuse", feature = "virtiofs"))]
mod fs; mod fs;
@ -32,7 +31,6 @@ pub use self::memory::{MemoryPathInfoService, MemoryPathInfoServiceConfig};
pub use self::nix_http::{NixHTTPPathInfoService, NixHTTPPathInfoServiceConfig}; pub use self::nix_http::{NixHTTPPathInfoService, NixHTTPPathInfoServiceConfig};
pub use self::redb::{RedbPathInfoService, RedbPathInfoServiceConfig}; pub use self::redb::{RedbPathInfoService, RedbPathInfoServiceConfig};
pub use self::signing_wrapper::{KeyFileSigningPathInfoServiceConfig, SigningPathInfoService}; pub use self::signing_wrapper::{KeyFileSigningPathInfoServiceConfig, SigningPathInfoService};
pub use self::sled::{SledPathInfoService, SledPathInfoServiceConfig};
#[cfg(test)] #[cfg(test)]
pub(crate) use self::signing_wrapper::test_signing_service; pub(crate) use self::signing_wrapper::test_signing_service;
@ -94,7 +92,6 @@ pub(crate) fn register_pathinfo_services(reg: &mut Registry) {
reg.register::<Box<dyn ServiceBuilder<Output = dyn PathInfoService>>, LruPathInfoServiceConfig>("lru"); reg.register::<Box<dyn ServiceBuilder<Output = dyn PathInfoService>>, LruPathInfoServiceConfig>("lru");
reg.register::<Box<dyn ServiceBuilder<Output = dyn PathInfoService>>, MemoryPathInfoServiceConfig>("memory"); reg.register::<Box<dyn ServiceBuilder<Output = dyn PathInfoService>>, MemoryPathInfoServiceConfig>("memory");
reg.register::<Box<dyn ServiceBuilder<Output = dyn PathInfoService>>, NixHTTPPathInfoServiceConfig>("nix"); reg.register::<Box<dyn ServiceBuilder<Output = dyn PathInfoService>>, NixHTTPPathInfoServiceConfig>("nix");
reg.register::<Box<dyn ServiceBuilder<Output = dyn PathInfoService>>, SledPathInfoServiceConfig>("sled");
reg.register::<Box<dyn ServiceBuilder<Output = dyn PathInfoService>>, RedbPathInfoServiceConfig>("redb"); reg.register::<Box<dyn ServiceBuilder<Output = dyn PathInfoService>>, RedbPathInfoServiceConfig>("redb");
reg.register::<Box<dyn ServiceBuilder<Output = dyn PathInfoService>>, KeyFileSigningPathInfoServiceConfig>("keyfile-signing"); reg.register::<Box<dyn ServiceBuilder<Output = dyn PathInfoService>>, KeyFileSigningPathInfoServiceConfig>("keyfile-signing");
#[cfg(feature = "cloud")] #[cfg(feature = "cloud")]

View file

@ -1,190 +0,0 @@
use super::PathInfoService;
use crate::proto::PathInfo;
use async_stream::try_stream;
use futures::stream::BoxStream;
use nix_compat::nixbase32;
use prost::Message;
use std::path::Path;
use std::sync::Arc;
use tonic::async_trait;
use tracing::{instrument, warn};
use tvix_castore::composition::{CompositionContext, ServiceBuilder};
use tvix_castore::Error;
/// SledPathInfoService stores PathInfo in a [sled](https://github.com/spacejam/sled).
///
/// The PathInfo messages are stored as encoded protos, and keyed by their output hash,
/// as that's currently the only request type available.
pub struct SledPathInfoService {
db: sled::Db,
}
impl SledPathInfoService {
pub fn new<P: AsRef<Path>>(p: P) -> Result<Self, sled::Error> {
if p.as_ref() == Path::new("/") {
return Err(sled::Error::Unsupported(
"cowardly refusing to open / with sled".to_string(),
));
}
let config = sled::Config::default()
.use_compression(false) // is a required parameter
.path(p);
let db = config.open()?;
Ok(Self { db })
}
pub fn new_temporary() -> Result<Self, sled::Error> {
let config = sled::Config::default().temporary(true);
let db = config.open()?;
Ok(Self { db })
}
}
#[async_trait]
impl PathInfoService for SledPathInfoService {
#[instrument(level = "trace", skip_all, fields(path_info.digest = nixbase32::encode(&digest)))]
async fn get(&self, digest: [u8; 20]) -> Result<Option<PathInfo>, Error> {
let resp = tokio::task::spawn_blocking({
let db = self.db.clone();
move || db.get(digest.as_slice())
})
.await?
.map_err(|e| {
warn!("failed to retrieve PathInfo: {}", e);
Error::StorageError(format!("failed to retrieve PathInfo: {}", e))
})?;
match resp {
None => Ok(None),
Some(data) => {
let path_info = PathInfo::decode(&*data).map_err(|e| {
warn!("failed to decode stored PathInfo: {}", e);
Error::StorageError(format!("failed to decode stored PathInfo: {}", e))
})?;
Ok(Some(path_info))
}
}
}
#[instrument(level = "trace", skip_all, fields(path_info.root_node = ?path_info.node))]
async fn put(&self, path_info: PathInfo) -> Result<PathInfo, Error> {
// Call validate on the received PathInfo message.
let store_path = path_info
.validate()
.map_err(|e| Error::InvalidRequest(format!("failed to validate PathInfo: {}", e)))?;
// In case the PathInfo is valid, we were able to parse a StorePath.
// Store it in the database, keyed by its digest.
// This overwrites existing PathInfo objects.
tokio::task::spawn_blocking({
let db = self.db.clone();
let k = *store_path.digest();
let data = path_info.encode_to_vec();
move || db.insert(k, data)
})
.await?
.map_err(|e| {
warn!("failed to insert PathInfo: {}", e);
Error::StorageError(format! {
"failed to insert PathInfo: {}", e
})
})?;
Ok(path_info)
}
fn list(&self) -> BoxStream<'static, Result<PathInfo, Error>> {
let db = self.db.clone();
let mut it = db.iter().values();
Box::pin(try_stream! {
// Don't block the executor while waiting for .next(), so wrap that
// in a spawn_blocking call.
// We need to pass around it to be able to reuse it.
while let (Some(elem), new_it) = tokio::task::spawn_blocking(move || {
(it.next(), it)
}).await? {
it = new_it;
let data = elem.map_err(|e| {
warn!("failed to retrieve PathInfo: {}", e);
Error::StorageError(format!("failed to retrieve PathInfo: {}", e))
})?;
let path_info = PathInfo::decode(&*data).map_err(|e| {
warn!("failed to decode stored PathInfo: {}", e);
Error::StorageError(format!("failed to decode stored PathInfo: {}", e))
})?;
yield path_info
}
})
}
}
#[derive(serde::Deserialize)]
#[serde(deny_unknown_fields)]
pub struct SledPathInfoServiceConfig {
is_temporary: bool,
#[serde(default)]
/// required when is_temporary = false
path: Option<String>,
}
impl TryFrom<url::Url> for SledPathInfoServiceConfig {
type Error = Box<dyn std::error::Error + Send + Sync>;
fn try_from(url: url::Url) -> Result<Self, Self::Error> {
// sled doesn't support host, and a path can be provided (otherwise
// it'll live in memory only).
if url.has_host() {
return Err(Error::StorageError("no host allowed".to_string()).into());
}
// TODO: expose compression and other parameters as URL parameters?
Ok(if url.path().is_empty() {
SledPathInfoServiceConfig {
is_temporary: true,
path: None,
}
} else {
SledPathInfoServiceConfig {
is_temporary: false,
path: Some(url.path().to_string()),
}
})
}
}
#[async_trait]
impl ServiceBuilder for SledPathInfoServiceConfig {
type Output = dyn PathInfoService;
async fn build<'a>(
&'a self,
_instance_name: &str,
_context: &CompositionContext,
) -> Result<Arc<dyn PathInfoService>, Box<dyn std::error::Error + Send + Sync + 'static>> {
match self {
SledPathInfoServiceConfig {
is_temporary: true,
path: None,
} => Ok(Arc::new(SledPathInfoService::new_temporary()?)),
SledPathInfoServiceConfig {
is_temporary: true,
path: Some(_),
} => Err(
Error::StorageError("Temporary SledPathInfoService can not have path".into())
.into(),
),
SledPathInfoServiceConfig {
is_temporary: false,
path: None,
} => Err(Error::StorageError("SledPathInfoService is missing path".into()).into()),
SledPathInfoServiceConfig {
is_temporary: false,
path: Some(path),
} => Ok(Arc::new(SledPathInfoService::new(path)?)),
}
}
}

View file

@ -9,7 +9,6 @@ use rstest_reuse::{self, *};
use super::PathInfoService; use super::PathInfoService;
use crate::pathinfoservice::redb::RedbPathInfoService; use crate::pathinfoservice::redb::RedbPathInfoService;
use crate::pathinfoservice::MemoryPathInfoService; use crate::pathinfoservice::MemoryPathInfoService;
use crate::pathinfoservice::SledPathInfoService;
use crate::proto::PathInfo; use crate::proto::PathInfo;
use crate::tests::fixtures::DUMMY_PATH_DIGEST; use crate::tests::fixtures::DUMMY_PATH_DIGEST;
use tvix_castore::proto as castorepb; use tvix_castore::proto as castorepb;
@ -29,7 +28,6 @@ use self::utils::make_bigtable_path_info_service;
let (_, _, svc) = make_grpc_path_info_service_client().await; let (_, _, svc) = make_grpc_path_info_service_client().await;
svc svc
})] })]
#[case::sled(SledPathInfoService::new_temporary().unwrap())]
#[case::redb(RedbPathInfoService::new_temporary().unwrap())] #[case::redb(RedbPathInfoService::new_temporary().unwrap())]
#[case::signing(test_signing_service())] #[case::signing(test_signing_service())]
#[cfg_attr(all(feature = "cloud",feature="integration"), case::bigtable(make_bigtable_path_info_service().await))] #[cfg_attr(all(feature = "cloud",feature="integration"), case::bigtable(make_bigtable_path_info_service().await))]