meilisearch/meilisearch-http/src/index_controller/dump_actor/mod.rs

235 lines
6.9 KiB
Rust
Raw Normal View History

2021-05-11 02:25:09 +08:00
mod actor;
mod handle_impl;
2021-05-11 02:25:09 +08:00
mod message;
mod v1;
mod v2;
2021-05-11 02:25:09 +08:00
use std::{fs::File, path::Path, sync::Arc};
2021-05-11 02:25:09 +08:00
2021-04-28 22:43:49 +08:00
use anyhow::bail;
use heed::EnvOpenOptions;
use log::{error, info};
use milli::update::{IndexDocumentsMethod, UpdateBuilder, UpdateFormat};
#[cfg(test)]
use mockall::automock;
2021-05-11 02:23:12 +08:00
use serde::{Deserialize, Serialize};
2021-04-28 22:43:49 +08:00
use tempfile::TempDir;
use thiserror::Error;
use uuid::Uuid;
2021-04-28 22:43:49 +08:00
2021-05-11 02:25:09 +08:00
use super::IndexMetadata;
use crate::helpers::compression;
2021-04-28 22:43:49 +08:00
use crate::index::Index;
use crate::index_controller::uuid_resolver;
2021-05-11 02:25:09 +08:00
pub use actor::DumpActor;
pub use handle_impl::*;
2021-05-11 02:25:09 +08:00
pub use message::DumpMsg;
pub type DumpResult<T> = std::result::Result<T, DumpError>;
#[derive(Error, Debug)]
pub enum DumpError {
#[error("error with index: {0}")]
Error(#[from] anyhow::Error),
#[error("Heed error: {0}")]
HeedError(#[from] heed::Error),
#[error("dump already running")]
DumpAlreadyRunning,
#[error("dump `{0}` does not exist")]
DumpDoesNotExist(String),
}
2021-04-28 22:43:49 +08:00
#[derive(Debug, Serialize, Deserialize, Copy, Clone)]
enum DumpVersion {
V1,
2021-04-27 16:27:43 +08:00
V2,
2021-04-28 22:43:49 +08:00
}
impl DumpVersion {
2021-05-05 20:11:56 +08:00
const CURRENT: Self = Self::V2;
2021-04-27 16:27:43 +08:00
/// Select the good importation function from the `DumpVersion` of metadata
2021-05-11 02:25:09 +08:00
pub fn import_index(
self,
size: usize,
uuid: Uuid,
2021-05-11 02:25:09 +08:00
dump_path: &Path,
db_path: &Path,
2021-05-07 00:44:16 +08:00
primary_key: Option<&str>,
2021-05-11 02:25:09 +08:00
) -> anyhow::Result<()> {
2021-04-27 16:27:43 +08:00
match self {
Self::V1 => v1::import_index(size, uuid, dump_path, db_path, primary_key),
Self::V2 => v2::import_index(size, uuid, dump_path, db_path, primary_key),
2021-04-27 16:27:43 +08:00
}
}
2021-04-28 22:43:49 +08:00
}
2021-05-11 02:25:09 +08:00
#[async_trait::async_trait]
#[cfg_attr(test, automock)]
pub trait DumpActorHandle {
/// Start the creation of a dump
/// Implementation: [handle_impl::DumpActorHandleImpl::create_dump]
async fn create_dump(&self) -> DumpResult<DumpInfo>;
/// Return the status of an already created dump
/// Implementation: [handle_impl::DumpActorHandleImpl::dump_status]
async fn dump_info(&self, uid: String) -> DumpResult<DumpInfo>;
}
2021-04-28 22:43:49 +08:00
#[derive(Debug, Serialize, Deserialize)]
#[serde(rename_all = "camelCase")]
2021-04-29 20:45:08 +08:00
pub struct Metadata {
2021-04-28 22:43:49 +08:00
indexes: Vec<IndexMetadata>,
db_version: String,
dump_version: DumpVersion,
}
2021-04-29 20:45:08 +08:00
impl Metadata {
/// Create a Metadata with the current dump version of meilisearch.
2021-04-28 22:43:49 +08:00
pub fn new(indexes: Vec<IndexMetadata>, db_version: String) -> Self {
2021-04-29 20:45:08 +08:00
Metadata {
2021-04-28 22:43:49 +08:00
indexes,
db_version,
dump_version: DumpVersion::CURRENT,
}
}
2021-04-29 20:45:08 +08:00
/// Extract Metadata from `metadata.json` file present at provided `dir_path`
2021-04-28 22:43:49 +08:00
fn from_path(dir_path: &Path) -> anyhow::Result<Self> {
let path = dir_path.join("metadata.json");
let file = File::open(path)?;
let reader = std::io::BufReader::new(file);
let metadata = serde_json::from_reader(reader)?;
Ok(metadata)
}
2021-04-29 20:45:08 +08:00
/// Write Metadata in `metadata.json` file at provided `dir_path`
2021-05-05 20:11:56 +08:00
pub async fn to_path(&self, dir_path: &Path) -> anyhow::Result<()> {
2021-04-28 22:43:49 +08:00
let path = dir_path.join("metadata.json");
2021-05-05 20:11:56 +08:00
tokio::fs::write(path, serde_json::to_string(self)?).await?;
2021-04-28 22:43:49 +08:00
Ok(())
}
}
2021-05-11 02:25:09 +08:00
#[derive(Debug, Serialize, Deserialize, PartialEq, Clone)]
#[serde(rename_all = "snake_case")]
pub enum DumpStatus {
Done,
InProgress,
Failed,
2021-04-28 22:43:49 +08:00
}
2021-05-11 02:25:09 +08:00
#[derive(Debug, Serialize, Clone)]
#[serde(rename_all = "camelCase")]
pub struct DumpInfo {
pub uid: String,
pub status: DumpStatus,
2021-05-25 16:48:57 +08:00
#[serde(skip_serializing_if = "Option::is_none")]
pub error: Option<String>,
2021-05-11 02:25:09 +08:00
}
2021-04-28 22:43:49 +08:00
2021-05-11 02:25:09 +08:00
impl DumpInfo {
pub fn new(uid: String, status: DumpStatus) -> Self {
Self {
uid,
status,
error: None,
}
2021-05-05 20:11:56 +08:00
}
2021-04-28 22:43:49 +08:00
2021-05-11 02:25:09 +08:00
pub fn with_error(&mut self, error: String) {
self.status = DumpStatus::Failed;
self.error = Some(error);
2021-05-05 20:11:56 +08:00
}
2021-04-28 22:43:49 +08:00
2021-05-11 02:25:09 +08:00
pub fn done(&mut self) {
self.status = DumpStatus::Done;
}
2021-04-28 22:43:49 +08:00
2021-05-11 02:25:09 +08:00
pub fn dump_already_in_progress(&self) -> bool {
self.status == DumpStatus::InProgress
}
2021-04-28 22:43:49 +08:00
}
2021-05-11 02:23:12 +08:00
pub fn load_dump(
2021-04-28 22:43:49 +08:00
db_path: impl AsRef<Path>,
dump_path: impl AsRef<Path>,
size: usize,
) -> anyhow::Result<()> {
info!("Importing dump from {}...", dump_path.as_ref().display());
let db_path = db_path.as_ref();
let dump_path = dump_path.as_ref();
2021-05-11 02:23:12 +08:00
let uuid_resolver = uuid_resolver::HeedUuidStore::new(&db_path)?;
2021-04-28 22:43:49 +08:00
// extract the dump in a temporary directory
2021-04-29 20:45:08 +08:00
let tmp_dir = TempDir::new_in(db_path)?;
2021-04-28 22:43:49 +08:00
let tmp_dir_path = tmp_dir.path();
compression::from_tar_gz(dump_path, tmp_dir_path)?;
// read dump metadata
2021-04-29 20:45:08 +08:00
let metadata = Metadata::from_path(&tmp_dir_path)?;
2021-04-28 22:43:49 +08:00
// remove indexes which have same `uuid` than indexes to import and create empty indexes
2021-05-11 02:23:12 +08:00
let existing_index_uids = uuid_resolver.list()?;
2021-04-28 22:43:49 +08:00
2021-04-27 16:27:43 +08:00
info!("Deleting indexes already present in the db and provided in the dump...");
2021-04-28 22:43:49 +08:00
for idx in &metadata.indexes {
if let Some((_, uuid)) = existing_index_uids.iter().find(|(s, _)| s == &idx.uid) {
// if we find the index in the `uuid_resolver` it's supposed to exist on the file system
// and we want to delete it
let path = db_path.join(&format!("indexes/index-{}", uuid));
info!("Deleting {}", path.display());
use std::io::ErrorKind::*;
match std::fs::remove_dir_all(path) {
Ok(()) => (),
// if an index was present in the metadata but missing of the fs we can ignore the
// problem because we are going to create it later
Err(e) if e.kind() == NotFound => (),
Err(e) => bail!(e),
}
} else {
// if the index does not exist in the `uuid_resolver` we create it
2021-05-11 02:23:12 +08:00
uuid_resolver.create_uuid(idx.uid.clone(), false)?;
2021-04-28 22:43:49 +08:00
}
}
// import each indexes content
for idx in metadata.indexes {
let dump_path = tmp_dir_path.join(&idx.uid);
2021-05-11 02:23:12 +08:00
// this cannot fail since we created all the missing uuid in the previous loop
let uuid = uuid_resolver.get_uuid(idx.uid)?.unwrap();
2021-05-07 00:44:16 +08:00
info!(
"Importing dump from {} into {}...",
dump_path.display(),
db_path.display()
);
metadata.dump_version.import_index(
size,
uuid,
&dump_path,
&db_path,
idx.meta.primary_key.as_ref().map(|s| s.as_ref()),
)?;
info!("Dump importation from {} succeed", dump_path.display());
2021-04-28 22:43:49 +08:00
}
// finally we can move all the unprocessed update file into our new DB
// this directory may not exists
let update_path = tmp_dir_path.join("update_files");
let db_update_path = db_path.join("updates/update_files");
if update_path.exists() {
let _ = std::fs::remove_dir_all(db_update_path);
std::fs::rename(
tmp_dir_path.join("update_files"),
db_path.join("updates/update_files"),
)?;
}
2021-05-07 00:44:16 +08:00
2021-04-28 22:43:49 +08:00
info!("Dump importation from {} succeed", dump_path.display());
Ok(())
}