about summary refs log tree commit diff
path: root/tvix/castore/src/blobservice/sled.rs
diff options
context:
space:
mode:
authorFlorian Klink <flokli@flokli.de>2023-09-21T19·32+0300
committerclbot <clbot@tvl.fyi>2023-09-22T12·51+0000
commit32f41458c0a0f62bf906021ef096c465ccc45581 (patch)
tree3aaab8c453871f39c46fb43f8278aa933b24519d /tvix/castore/src/blobservice/sled.rs
parentd8ef0cfb4a859af7e33828b013356412d02532da (diff)
refactor(tvix): move castore into tvix-castore crate r/6629
This splits the pure content-addressed layers from tvix-store into a
`castore` crate, and only leaves PathInfo related things, as well as the
CLI entrypoint in the tvix-store crate.

Notable changes:
 - `fixtures` and `utils` had to be moved out of the `test` cfg, so they
   can be imported from tvix-store.
 - Some ad-hoc fixtures in the test were moved to proper fixtures in the
   same step.
 - The protos are now created by a (more static) recipe in the protos/
   directory.

The (now two) golang targets are commented out, as it's not possible to
update them properly in the same CL. This will be done by a followup CL
once this is merged (and whitby deployed)

Bug: https://b.tvl.fyi/issues/301

Change-Id: I8d675d4bf1fb697eb7d479747c1b1e3635718107
Reviewed-on: https://cl.tvl.fyi/c/depot/+/9370
Reviewed-by: tazjin <tazjin@tvl.su>
Reviewed-by: flokli <flokli@flokli.de>
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Diffstat (limited to 'tvix/castore/src/blobservice/sled.rs')
-rw-r--r--tvix/castore/src/blobservice/sled.rs249
1 files changed, 249 insertions, 0 deletions
diff --git a/tvix/castore/src/blobservice/sled.rs b/tvix/castore/src/blobservice/sled.rs
new file mode 100644
index 000000000000..209f0b76fc7a
--- /dev/null
+++ b/tvix/castore/src/blobservice/sled.rs
@@ -0,0 +1,249 @@
+use super::{BlobReader, BlobService, BlobWriter};
+use crate::{B3Digest, Error};
+use std::{
+    io::{self, Cursor, Write},
+    path::PathBuf,
+    task::Poll,
+};
+use tonic::async_trait;
+use tracing::instrument;
+
+#[derive(Clone)]
+pub struct SledBlobService {
+    db: sled::Db,
+}
+
+impl SledBlobService {
+    pub fn new(p: PathBuf) -> Result<Self, sled::Error> {
+        let config = sled::Config::default().use_compression(true).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 BlobService for SledBlobService {
+    /// Constructs a [SledBlobService] from the passed [url::Url]:
+    /// - scheme has to be `sled://`
+    /// - there may not be a host.
+    /// - a path to the sled needs to be provided (which may not be `/`).
+    fn from_url(url: &url::Url) -> Result<Self, Error> {
+        if url.scheme() != "sled" {
+            return Err(crate::Error::StorageError("invalid scheme".to_string()));
+        }
+
+        if url.has_host() {
+            return Err(crate::Error::StorageError(format!(
+                "invalid host: {}",
+                url.host().unwrap()
+            )));
+        }
+
+        // TODO: expose compression and other parameters as URL parameters, drop new and new_temporary?
+        if url.path().is_empty() {
+            Self::new_temporary().map_err(|e| Error::StorageError(e.to_string()))
+        } else if url.path() == "/" {
+            Err(crate::Error::StorageError(
+                "cowardly refusing to open / with sled".to_string(),
+            ))
+        } else {
+            Self::new(url.path().into()).map_err(|e| Error::StorageError(e.to_string()))
+        }
+    }
+
+    #[instrument(skip(self), fields(blob.digest=%digest))]
+    async fn has(&self, digest: &B3Digest) -> Result<bool, Error> {
+        match self.db.contains_key(digest.to_vec()) {
+            Ok(has) => Ok(has),
+            Err(e) => Err(Error::StorageError(e.to_string())),
+        }
+    }
+
+    #[instrument(skip(self), fields(blob.digest=%digest))]
+    async fn open_read(&self, digest: &B3Digest) -> Result<Option<Box<dyn BlobReader>>, Error> {
+        match self.db.get(digest.to_vec()) {
+            Ok(None) => Ok(None),
+            Ok(Some(data)) => Ok(Some(Box::new(Cursor::new(data[..].to_vec())))),
+            Err(e) => Err(Error::StorageError(e.to_string())),
+        }
+    }
+
+    #[instrument(skip(self))]
+    async fn open_write(&self) -> Box<dyn BlobWriter> {
+        Box::new(SledBlobWriter::new(self.db.clone()))
+    }
+}
+
+pub struct SledBlobWriter {
+    db: sled::Db,
+
+    /// Contains the buffer Vec and hasher, or None if already closed
+    writers: Option<(Vec<u8>, blake3::Hasher)>,
+
+    /// The digest that has been returned, if we successfully closed.
+    digest: Option<B3Digest>,
+}
+
+impl SledBlobWriter {
+    pub fn new(db: sled::Db) -> Self {
+        Self {
+            db,
+            writers: Some((Vec::new(), blake3::Hasher::new())),
+            digest: None,
+        }
+    }
+}
+
+impl tokio::io::AsyncWrite for SledBlobWriter {
+    fn poll_write(
+        mut self: std::pin::Pin<&mut Self>,
+        _cx: &mut std::task::Context<'_>,
+        b: &[u8],
+    ) -> std::task::Poll<Result<usize, io::Error>> {
+        Poll::Ready(match &mut self.writers {
+            None => Err(io::Error::new(
+                io::ErrorKind::NotConnected,
+                "already closed",
+            )),
+            Some((ref mut buf, ref mut hasher)) => {
+                let bytes_written = buf.write(b)?;
+                hasher.write(&b[..bytes_written])
+            }
+        })
+    }
+
+    fn poll_flush(
+        mut self: std::pin::Pin<&mut Self>,
+        _cx: &mut std::task::Context<'_>,
+    ) -> std::task::Poll<Result<(), io::Error>> {
+        Poll::Ready(match &mut self.writers {
+            None => Err(io::Error::new(
+                io::ErrorKind::NotConnected,
+                "already closed",
+            )),
+            Some(_) => Ok(()),
+        })
+    }
+
+    fn poll_shutdown(
+        self: std::pin::Pin<&mut Self>,
+        _cx: &mut std::task::Context<'_>,
+    ) -> std::task::Poll<Result<(), io::Error>> {
+        // shutdown is "instantaneous", we only write to a Vec<u8> as buffer.
+        Poll::Ready(Ok(()))
+    }
+}
+
+#[async_trait]
+impl BlobWriter for SledBlobWriter {
+    async fn close(&mut self) -> Result<B3Digest, Error> {
+        if self.writers.is_none() {
+            match &self.digest {
+                Some(digest) => Ok(digest.clone()),
+                None => Err(crate::Error::StorageError(
+                    "previously closed with error".to_string(),
+                )),
+            }
+        } else {
+            let (buf, hasher) = self.writers.take().unwrap();
+
+            let digest: B3Digest = hasher.finalize().as_bytes().into();
+
+            // Only insert if the blob doesn't already exist.
+            if !self.db.contains_key(digest.to_vec()).map_err(|e| {
+                Error::StorageError(format!("Unable to check if we have blob {}: {}", digest, e))
+            })? {
+                // put buf in there. This will move buf out.
+                self.db
+                    .insert(digest.to_vec(), buf)
+                    .map_err(|e| Error::StorageError(format!("unable to insert blob: {}", e)))?;
+            }
+
+            self.digest = Some(digest.clone());
+
+            Ok(digest)
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use tempfile::TempDir;
+
+    use super::BlobService;
+    use super::SledBlobService;
+
+    /// This uses a wrong scheme.
+    #[test]
+    fn test_invalid_scheme() {
+        let url = url::Url::parse("http://foo.example/test").expect("must parse");
+
+        assert!(SledBlobService::from_url(&url).is_err());
+    }
+
+    /// This uses the correct scheme, and doesn't specify a path (temporary sled).
+    #[test]
+    fn test_valid_scheme_temporary() {
+        let url = url::Url::parse("sled://").expect("must parse");
+
+        assert!(SledBlobService::from_url(&url).is_ok());
+    }
+
+    /// This sets the path to a location that doesn't exist, which should fail (as sled doesn't mkdir -p)
+    #[test]
+    fn test_nonexistent_path() {
+        let tmpdir = TempDir::new().unwrap();
+
+        let mut url = url::Url::parse("sled://foo.example").expect("must parse");
+        url.set_path(tmpdir.path().join("foo").join("bar").to_str().unwrap());
+
+        assert!(SledBlobService::from_url(&url).is_err());
+    }
+
+    /// This uses the correct scheme, and specifies / as path (which should fail
+    // for obvious reasons)
+    #[test]
+    fn test_invalid_path_root() {
+        let url = url::Url::parse("sled:///").expect("must parse");
+
+        assert!(SledBlobService::from_url(&url).is_err());
+    }
+
+    /// This uses the correct scheme, and sets a tempdir as location.
+    #[test]
+    fn test_valid_scheme_path() {
+        let tmpdir = TempDir::new().unwrap();
+
+        let mut url = url::Url::parse("sled://").expect("must parse");
+        url.set_path(tmpdir.path().to_str().unwrap());
+
+        assert!(SledBlobService::from_url(&url).is_ok());
+    }
+
+    /// This sets a host, rather than a path, which should fail.
+    #[test]
+    fn test_invalid_host() {
+        let url = url::Url::parse("sled://foo.example").expect("must parse");
+
+        assert!(SledBlobService::from_url(&url).is_err());
+    }
+
+    /// This sets a host AND a valid path, which should fail
+    #[test]
+    fn test_invalid_host_and_path() {
+        let tmpdir = TempDir::new().unwrap();
+
+        let mut url = url::Url::parse("sled://foo.example").expect("must parse");
+        url.set_path(tmpdir.path().to_str().unwrap());
+
+        assert!(SledBlobService::from_url(&url).is_err());
+    }
+}