Use OS file locks in validator client (#1958)

## Issue Addressed

Closes #1823

## Proposed Changes

* Use OS-level file locking for validator keystores, eliminating problems with lockfiles lingering after ungraceful shutdowns (`SIGKILL`, power outage). I'm using the `fs2` crate because it's cross-platform (unlike `file-lock`), and it seems to have the most downloads on crates.io.
* Deprecate + disable `--delete-lockfiles` CLI param, it's no longer necessary
* Delete the `validator_dir::Manager`, as it was mostly dead code and was only used in the `validator list` command, which has been rewritten to read the validator definitions YAML instead.

## Additional Info

Tested on:

- [x] Linux
- [x] macOS
- [x] Docker Linux
- [x] Docker macOS
- [ ] Windows
This commit is contained in:
Michael Sproul
2020-11-26 11:25:46 +00:00
parent fc07cc3fdf
commit 3486d6a809
21 changed files with 282 additions and 411 deletions

View File

@@ -3,24 +3,22 @@ use crate::builder::{
WITHDRAWAL_KEYSTORE_FILE,
};
use deposit_contract::decode_eth1_tx_data;
use derivative::Derivative;
use eth2_keystore::{Error as KeystoreError, Keystore, PlainText};
use std::fs::{read, remove_file, write, OpenOptions};
use lockfile::{Lockfile, LockfileError};
use std::fs::{read, write, OpenOptions};
use std::io;
use std::path::{Path, PathBuf};
use tree_hash::TreeHash;
use types::{DepositData, Hash256, Keypair};
/// The file used for indicating if a directory is in-use by another process.
const LOCK_FILE: &str = ".lock";
/// The file used to save the Eth1 transaction hash from a deposit.
pub const ETH1_DEPOSIT_TX_HASH_FILE: &str = "eth1-deposit-tx-hash.txt";
#[derive(Debug)]
pub enum Error {
DirectoryDoesNotExist(PathBuf),
DirectoryLocked(PathBuf),
UnableToCreateLockfile(io::Error),
LockfileError(LockfileError),
UnableToOpenKeystore(io::Error),
UnableToReadKeystore(KeystoreError),
UnableToOpenPassword(io::Error),
@@ -58,19 +56,22 @@ pub struct Eth1DepositData {
/// Provides a wrapper around a directory containing validator information.
///
/// Creates/deletes a lockfile in `self.dir` to attempt to prevent concurrent access from multiple
/// Holds a lockfile in `self.dir` to attempt to prevent concurrent access from multiple
/// processes.
#[derive(Debug, PartialEq)]
#[derive(Debug, Derivative)]
#[derivative(PartialEq)]
pub struct ValidatorDir {
dir: PathBuf,
#[derivative(PartialEq = "ignore")]
lockfile: Lockfile,
}
impl ValidatorDir {
/// Open `dir`, creating a lockfile to prevent concurrent access.
/// Open `dir`, obtaining a lockfile to prevent concurrent access.
///
/// ## Errors
///
/// If there is a filesystem error or if a lockfile already exists.
/// If there is a filesystem error or if the lockfile is locked by another process.
pub fn open<P: AsRef<Path>>(dir: P) -> Result<Self, Error> {
let dir: &Path = dir.as_ref();
let dir: PathBuf = dir.into();
@@ -79,49 +80,12 @@ impl ValidatorDir {
return Err(Error::DirectoryDoesNotExist(dir));
}
let lockfile = dir.join(LOCK_FILE);
if lockfile.exists() {
return Err(Error::DirectoryLocked(dir));
} else {
OpenOptions::new()
.write(true)
.create_new(true)
.open(lockfile)
.map_err(Error::UnableToCreateLockfile)?;
}
// Lock the keystore file that *might* be in this directory.
// This is not ideal, see: https://github.com/sigp/lighthouse/issues/1978
let lockfile_path = dir.join(format!("{}.lock", VOTING_KEYSTORE_FILE));
let lockfile = Lockfile::new(lockfile_path).map_err(Error::LockfileError)?;
Ok(Self { dir })
}
/// Open `dir`, regardless or not if a lockfile exists.
///
/// Returns `(validator_dir, lockfile_existed)`, where `lockfile_existed == true` if a lockfile
/// was already present before opening. Creates a lockfile if one did not already exist.
///
/// ## Errors
///
/// If there is a filesystem error.
pub fn force_open<P: AsRef<Path>>(dir: P) -> Result<(Self, bool), Error> {
let dir: &Path = dir.as_ref();
let dir: PathBuf = dir.into();
if !dir.exists() {
return Err(Error::DirectoryDoesNotExist(dir));
}
let lockfile = dir.join(LOCK_FILE);
let lockfile_exists = lockfile.exists();
if !lockfile_exists {
OpenOptions::new()
.write(true)
.create_new(true)
.open(lockfile)
.map_err(Error::UnableToCreateLockfile)?;
}
Ok((Self { dir }, lockfile_exists))
Ok(Self { dir, lockfile })
}
/// Returns the `dir` provided to `Self::open`.
@@ -238,18 +202,6 @@ impl ValidatorDir {
}
}
impl Drop for ValidatorDir {
fn drop(&mut self) {
let lockfile = self.dir.clone().join(LOCK_FILE);
if let Err(e) = remove_file(&lockfile) {
eprintln!(
"Unable to remove validator lockfile {:?}: {:?}",
lockfile, e
);
}
}
}
/// Attempts to load and decrypt a Keypair given path to the keystore.
pub fn unlock_keypair<P: AsRef<Path>>(
keystore_path: &PathBuf,