use std::collections::BTreeSet; use std::fs::File; use std::io::{BufRead, BufReader}; use std::os::unix::io::AsRawFd; use std::path::{Path, PathBuf}; use std::sync::{Arc, RwLock}; use std::thread::spawn; use std::time::SystemTime; use anyhow::{bail, format_err, Error}; use crossbeam_channel::{bounded, TryRecvError}; use proxmox_sys::fs::{create_path, CreateOptions}; use crate::rrd::{CF, DST, RRA, RRD}; mod journal; use journal::*; mod rrd_map; use rrd_map::*; /// RRD cache - keep RRD data in RAM, but write updates to disk /// /// This cache is designed to run as single instance (no concurrent /// access from other processes). pub struct RRDCache { config: Arc, state: Arc>, rrd_map: Arc>, } pub(crate) struct CacheConfig { apply_interval: f64, basedir: PathBuf, file_options: CreateOptions, dir_options: CreateOptions, } impl RRDCache { /// Creates a new instance /// /// `basedir`: All files are stored relative to this path. /// /// `file_options`: Files are created with this options. /// /// `dir_options`: Directories are created with this options. /// /// `apply_interval`: Commit journal after `apply_interval` seconds. /// /// `load_rrd_cb`; The callback function is used to load RRD files, /// and should return a newly generated RRD if the file does not /// exists (or is unreadable). This may generate RRDs with /// different configurations (dependent on `rel_path`). pub fn new>( basedir: P, file_options: Option, dir_options: Option, apply_interval: f64, load_rrd_cb: fn(path: &Path, rel_path: &str, dst: DST) -> RRD, ) -> Result { let basedir = basedir.as_ref().to_owned(); let file_options = file_options.unwrap_or_else(CreateOptions::new); let dir_options = dir_options.unwrap_or_else(CreateOptions::new); create_path( &basedir, Some(dir_options.clone()), Some(dir_options.clone()), ) .map_err(|err: Error| format_err!("unable to create rrdb stat dir - {}", err))?; let config = Arc::new(CacheConfig { basedir, file_options, dir_options, apply_interval, }); let state = JournalState::new(Arc::clone(&config))?; let rrd_map = RRDMap::new(Arc::clone(&config), load_rrd_cb); Ok(Self { config: Arc::clone(&config), state: Arc::new(RwLock::new(state)), rrd_map: Arc::new(RwLock::new(rrd_map)), }) } /// Create a new RRD as used by the proxmox backup server /// /// It contains the following RRAs: /// /// * cf=average,r=60,n=1440 => 1day /// * cf=maximum,r=60,n=1440 => 1day /// * cf=average,r=30*60,n=1440 => 1month /// * cf=maximum,r=30*60,n=1440 => 1month /// * cf=average,r=6*3600,n=1440 => 1year /// * cf=maximum,r=6*3600,n=1440 => 1year /// * cf=average,r=7*86400,n=570 => 10years /// * cf=maximum,r=7*86400,n=570 => 10year /// /// The resultion data file size is about 80KB. pub fn create_proxmox_backup_default_rrd(dst: DST) -> RRD { let rra_list = vec![ // 1 min * 1440 => 1 day RRA::new(CF::Average, 60, 1440), RRA::new(CF::Maximum, 60, 1440), // 30 min * 1440 => 30 days ~ 1 month RRA::new(CF::Average, 30 * 60, 1440), RRA::new(CF::Maximum, 30 * 60, 1440), // 6 h * 1440 => 360 days ~ 1 year RRA::new(CF::Average, 6 * 3600, 1440), RRA::new(CF::Maximum, 6 * 3600, 1440), // 1 week * 570 => 10 years RRA::new(CF::Average, 7 * 86400, 570), RRA::new(CF::Maximum, 7 * 86400, 570), ]; RRD::new(dst, rra_list) } /// Sync the journal data to disk (using `fdatasync` syscall) pub fn sync_journal(&self) -> Result<(), Error> { self.state.read().unwrap().sync_journal() } /// Apply and commit the journal. Should be used at server startup. pub fn apply_journal(&self) -> Result { let config = Arc::clone(&self.config); let state = Arc::clone(&self.state); let rrd_map = Arc::clone(&self.rrd_map); let mut state_guard = self.state.write().unwrap(); let journal_applied = state_guard.journal_applied; if let Some(ref recv) = state_guard.apply_thread_result { match recv.try_recv() { Ok(Ok(())) => { // finished without errors, OK state_guard.apply_thread_result = None; } Ok(Err(err)) => { // finished with errors, log them log::error!("{}", err); state_guard.apply_thread_result = None; } Err(TryRecvError::Empty) => { // still running return Ok(journal_applied); } Err(TryRecvError::Disconnected) => { // crashed, start again log::error!("apply journal thread crashed - try again"); state_guard.apply_thread_result = None; } } } let now = proxmox_time::epoch_f64(); let wants_commit = (now - state_guard.last_journal_flush) > self.config.apply_interval; if journal_applied && !wants_commit { return Ok(journal_applied); } state_guard.last_journal_flush = proxmox_time::epoch_f64(); let (sender, receiver) = bounded(1); state_guard.apply_thread_result = Some(receiver); spawn(move || { let result = apply_and_commit_journal_thread(config, state, rrd_map, journal_applied) .map_err(|err| err.to_string()); sender.send(result).unwrap(); }); Ok(journal_applied) } /// Update data in RAM and write file back to disk (journal) pub fn update_value( &self, rel_path: &str, time: f64, value: f64, dst: DST, ) -> Result<(), Error> { let journal_applied = self.apply_journal()?; self.state .write() .unwrap() .append_journal_entry(time, value, dst, rel_path)?; if journal_applied { self.rrd_map .write() .unwrap() .update(rel_path, time, value, dst, false)?; } Ok(()) } /// Extract data from cached RRD /// /// `start`: Start time. If not sepecified, we simply extract 10 data points. /// /// `end`: End time. Default is to use the current time. pub fn extract_cached_data( &self, base: &str, name: &str, cf: CF, resolution: u64, start: Option, end: Option, ) -> Result>)>, Error> { self.rrd_map .read() .unwrap() .extract_cached_data(base, name, cf, resolution, start, end) } } fn apply_and_commit_journal_thread( config: Arc, state: Arc>, rrd_map: Arc>, commit_only: bool, ) -> Result<(), Error> { if commit_only { state.write().unwrap().rotate_journal()?; // start new journal, keep old one } else { let start_time = SystemTime::now(); log::debug!("applying rrd journal"); match apply_journal_impl(Arc::clone(&state), Arc::clone(&rrd_map)) { Ok(entries) => { let elapsed = start_time.elapsed().unwrap().as_secs_f64(); log::info!( "applied rrd journal ({} entries in {:.3} seconds)", entries, elapsed ); } Err(err) => bail!("apply rrd journal failed - {}", err), } } let start_time = SystemTime::now(); log::debug!("commit rrd journal"); match commit_journal_impl(config, state, rrd_map) { Ok(rrd_file_count) => { let elapsed = start_time.elapsed().unwrap().as_secs_f64(); log::info!( "rrd journal successfully committed ({} files in {:.3} seconds)", rrd_file_count, elapsed ); } Err(err) => bail!("rrd journal commit failed: {}", err), } Ok(()) } fn apply_journal_lines( state: Arc>, rrd_map: Arc>, journal_name: &str, // used for logging reader: &mut BufReader, lock_read_line: bool, ) -> Result { let mut linenr = 0; loop { linenr += 1; let mut line = String::new(); let len = if lock_read_line { let _lock = state.read().unwrap(); // make sure we read entire lines reader.read_line(&mut line)? } else { reader.read_line(&mut line)? }; if len == 0 { break; } let entry: JournalEntry = match line.parse() { Ok(entry) => entry, Err(err) => { log::warn!( "unable to parse rrd journal '{}' line {} (skip) - {}", journal_name, linenr, err, ); continue; // skip unparsable lines } }; rrd_map.write().unwrap().update( &entry.rel_path, entry.time, entry.value, entry.dst, true, )?; } Ok(linenr) } fn apply_journal_impl( state: Arc>, rrd_map: Arc>, ) -> Result { let mut lines = 0; // Apply old journals first let journal_list = state.read().unwrap().list_old_journals()?; for entry in journal_list { log::info!("apply old journal log {}", entry.name); let file = std::fs::OpenOptions::new().read(true).open(&entry.path)?; let mut reader = BufReader::new(file); lines += apply_journal_lines( Arc::clone(&state), Arc::clone(&rrd_map), &entry.name, &mut reader, false, )?; } let mut journal = state.read().unwrap().open_journal_reader()?; lines += apply_journal_lines( Arc::clone(&state), Arc::clone(&rrd_map), "rrd.journal", &mut journal, true, )?; { let mut state_guard = state.write().unwrap(); // block other writers lines += apply_journal_lines( Arc::clone(&state), Arc::clone(&rrd_map), "rrd.journal", &mut journal, false, )?; state_guard.rotate_journal()?; // start new journal, keep old one // We need to apply the journal only once, because further updates // are always directly applied. state_guard.journal_applied = true; } Ok(lines) } fn fsync_file_or_dir(path: &Path) -> Result<(), Error> { let file = std::fs::File::open(path)?; nix::unistd::fsync(file.as_raw_fd())?; Ok(()) } pub(crate) fn fsync_file_and_parent(path: &Path) -> Result<(), Error> { let file = std::fs::File::open(path)?; nix::unistd::fsync(file.as_raw_fd())?; if let Some(parent) = path.parent() { fsync_file_or_dir(parent)?; } Ok(()) } fn rrd_parent_dir(basedir: &Path, rel_path: &str) -> PathBuf { let mut path = basedir.to_owned(); let rel_path = Path::new(rel_path); if let Some(parent) = rel_path.parent() { path.push(parent); } path } fn commit_journal_impl( config: Arc, state: Arc>, rrd_map: Arc>, ) -> Result { let files = rrd_map.read().unwrap().file_list(); let mut rrd_file_count = 0; let mut errors = 0; let mut dir_set = BTreeSet::new(); log::info!("write rrd data back to disk"); // save all RRDs - we only need a read lock here // Note: no fsync here (we do it afterwards) for rel_path in files.iter() { let parent_dir = rrd_parent_dir(&config.basedir, rel_path); dir_set.insert(parent_dir); rrd_file_count += 1; if let Err(err) = rrd_map.read().unwrap().flush_rrd_file(rel_path) { errors += 1; log::error!("unable to save rrd {}: {}", rel_path, err); } } if errors != 0 { bail!("errors during rrd flush - unable to commit rrd journal"); } // Important: We fsync files after writing all data! This increase // the likelihood that files are already synced, so this is // much faster (although we need to re-open the files). log::info!("starting rrd data sync"); for rel_path in files.iter() { let mut path = config.basedir.clone(); path.push(&rel_path); fsync_file_or_dir(&path) .map_err(|err| format_err!("fsync rrd file {} failed - {}", rel_path, err))?; } // also fsync directories for dir_path in dir_set { fsync_file_or_dir(&dir_path) .map_err(|err| format_err!("fsync rrd dir {:?} failed - {}", dir_path, err))?; } // if everything went ok, remove the old journal files state.write().unwrap().remove_old_journals()?; Ok(rrd_file_count) }