]>
Commit | Line | Data |
---|---|---|
42c2b5be | 1 | use std::collections::{HashMap, HashSet}; |
54552dda | 2 | use std::io::{self, Write}; |
4c7cc5b3 | 3 | use std::os::unix::io::AsRawFd; |
367f002e | 4 | use std::path::{Path, PathBuf}; |
cb4b721c | 5 | use std::str::FromStr; |
42c2b5be | 6 | use std::sync::{Arc, Mutex}; |
367f002e | 7 | |
f7d4e4b5 | 8 | use anyhow::{bail, format_err, Error}; |
2c32fdde | 9 | use lazy_static::lazy_static; |
4c7cc5b3 | 10 | use nix::unistd::{unlinkat, UnlinkatFlags}; |
e4439025 | 11 | |
fef61684 DC |
12 | use proxmox_schema::ApiType; |
13 | ||
42c2b5be TL |
14 | use proxmox_sys::fs::{file_read_optional_string, replace_file, CreateOptions}; |
15 | use proxmox_sys::fs::{lock_dir_noblock, DirLockGuard}; | |
d5790a9f | 16 | use proxmox_sys::process_locker::ProcessLockSharedGuard; |
25877d05 | 17 | use proxmox_sys::WorkerTaskContext; |
d5790a9f | 18 | use proxmox_sys::{task_log, task_warn}; |
529de6c7 | 19 | |
fef61684 | 20 | use pbs_api_types::{ |
8c74349b | 21 | Authid, BackupNamespace, BackupType, ChunkOrder, DataStoreConfig, DatastoreTuning, |
118e9849 | 22 | GarbageCollectionStatus, HumanByte, Operation, UPID, |
fef61684 | 23 | }; |
529de6c7 | 24 | |
42c2b5be | 25 | use crate::backup_info::{BackupDir, BackupGroup}; |
6d5d305d DM |
26 | use crate::chunk_store::ChunkStore; |
27 | use crate::dynamic_index::{DynamicIndexReader, DynamicIndexWriter}; | |
28 | use crate::fixed_index::{FixedIndexReader, FixedIndexWriter}; | |
5444b914 | 29 | use crate::hierarchy::{ListGroups, ListGroupsType, ListNamespaces, ListNamespacesRecursive}; |
6d5d305d | 30 | use crate::index::IndexFile; |
9ccf933b | 31 | use crate::manifest::{archive_type, ArchiveType}; |
4bc84a65 | 32 | use crate::task_tracking::update_active_operations; |
42c2b5be | 33 | use crate::DataBlob; |
6d5d305d | 34 | |
367f002e | 35 | lazy_static! { |
42c2b5be TL |
36 | static ref DATASTORE_MAP: Mutex<HashMap<String, Arc<DataStoreImpl>>> = |
37 | Mutex::new(HashMap::new()); | |
b3483782 | 38 | } |
ff3d3100 | 39 | |
9751ef4b DC |
40 | /// checks if auth_id is owner, or, if owner is a token, if |
41 | /// auth_id is the user of the token | |
42c2b5be TL |
42 | pub fn check_backup_owner(owner: &Authid, auth_id: &Authid) -> Result<(), Error> { |
43 | let correct_owner = | |
44 | owner == auth_id || (owner.is_token() && &Authid::from(owner.user().clone()) == auth_id); | |
9751ef4b DC |
45 | if !correct_owner { |
46 | bail!("backup owner check failed ({} != {})", auth_id, owner); | |
47 | } | |
48 | Ok(()) | |
49 | } | |
50 | ||
e5064ba6 DM |
51 | /// Datastore Management |
52 | /// | |
53 | /// A Datastore can store severals backups, and provides the | |
54 | /// management interface for backup. | |
4bc84a65 | 55 | pub struct DataStoreImpl { |
1629d2ad | 56 | chunk_store: Arc<ChunkStore>, |
81b2a872 | 57 | gc_mutex: Mutex<()>, |
f2b99c34 | 58 | last_gc_status: Mutex<GarbageCollectionStatus>, |
0698f78d | 59 | verify_new: bool, |
fef61684 | 60 | chunk_order: ChunkOrder, |
51d900d1 | 61 | last_digest: Option<[u8; 32]>, |
529de6c7 DM |
62 | } |
63 | ||
6da20161 WB |
64 | impl DataStoreImpl { |
65 | // This one just panics on everything | |
66 | #[doc(hidden)] | |
615a50c1 | 67 | pub(crate) unsafe fn new_test() -> Arc<Self> { |
6da20161 WB |
68 | Arc::new(Self { |
69 | chunk_store: Arc::new(unsafe { ChunkStore::panic_store() }), | |
70 | gc_mutex: Mutex::new(()), | |
71 | last_gc_status: Mutex::new(GarbageCollectionStatus::default()), | |
72 | verify_new: false, | |
73 | chunk_order: ChunkOrder::None, | |
51d900d1 | 74 | last_digest: None, |
6da20161 WB |
75 | }) |
76 | } | |
77 | } | |
78 | ||
4bc84a65 HL |
79 | pub struct DataStore { |
80 | inner: Arc<DataStoreImpl>, | |
81 | operation: Option<Operation>, | |
82 | } | |
83 | ||
84 | impl Clone for DataStore { | |
85 | fn clone(&self) -> Self { | |
86 | let mut new_operation = self.operation; | |
87 | if let Some(operation) = self.operation { | |
88 | if let Err(e) = update_active_operations(self.name(), operation, 1) { | |
89 | log::error!("could not update active operations - {}", e); | |
90 | new_operation = None; | |
91 | } | |
92 | } | |
93 | ||
94 | DataStore { | |
95 | inner: self.inner.clone(), | |
96 | operation: new_operation, | |
97 | } | |
98 | } | |
99 | } | |
100 | ||
101 | impl Drop for DataStore { | |
102 | fn drop(&mut self) { | |
103 | if let Some(operation) = self.operation { | |
104 | if let Err(e) = update_active_operations(self.name(), operation, -1) { | |
105 | log::error!("could not update active operations - {}", e); | |
106 | } | |
107 | } | |
108 | } | |
109 | } | |
110 | ||
529de6c7 | 111 | impl DataStore { |
6da20161 WB |
112 | // This one just panics on everything |
113 | #[doc(hidden)] | |
615a50c1 | 114 | pub(crate) unsafe fn new_test() -> Arc<Self> { |
6da20161 WB |
115 | Arc::new(Self { |
116 | inner: unsafe { DataStoreImpl::new_test() }, | |
117 | operation: None, | |
118 | }) | |
119 | } | |
120 | ||
e9d2fc93 HL |
121 | pub fn lookup_datastore( |
122 | name: &str, | |
123 | operation: Option<Operation>, | |
124 | ) -> Result<Arc<DataStore>, Error> { | |
51d900d1 TL |
125 | // we could use the ConfigVersionCache's generation for staleness detection, but we load |
126 | // the config anyway -> just use digest, additional benefit: manual changes get detected | |
127 | let (config, digest) = pbs_config::datastore::config()?; | |
e9d2fc93 | 128 | let config: DataStoreConfig = config.lookup("datastore", name)?; |
e9d2fc93 HL |
129 | |
130 | if let Some(maintenance_mode) = config.get_maintenance_mode() { | |
131 | if let Err(error) = maintenance_mode.check(operation) { | |
0408f60b | 132 | bail!("datastore '{name}' is in {error}"); |
e9d2fc93 HL |
133 | } |
134 | } | |
135 | ||
4bc84a65 HL |
136 | if let Some(operation) = operation { |
137 | update_active_operations(name, operation, 1)?; | |
138 | } | |
139 | ||
33a1ef7a TL |
140 | let mut datastore_cache = DATASTORE_MAP.lock().unwrap(); |
141 | let entry = datastore_cache.get(name); | |
2c32fdde | 142 | |
c7f7236b | 143 | // reuse chunk store so that we keep using the same process locker instance! |
0bd9c870 | 144 | let chunk_store = if let Some(datastore) = &entry { |
51d900d1 TL |
145 | let last_digest = datastore.last_digest.as_ref(); |
146 | if let Some(true) = last_digest.map(|last_digest| last_digest == &digest) { | |
4bc84a65 HL |
147 | return Ok(Arc::new(Self { |
148 | inner: Arc::clone(datastore), | |
149 | operation, | |
150 | })); | |
2c32fdde | 151 | } |
0bd9c870 DC |
152 | Arc::clone(&datastore.chunk_store) |
153 | } else { | |
154 | Arc::new(ChunkStore::open(name, &config.path)?) | |
155 | }; | |
2c32fdde | 156 | |
51d900d1 | 157 | let datastore = DataStore::with_store_and_config(chunk_store, config, Some(digest))?; |
f0a61124 DM |
158 | |
159 | let datastore = Arc::new(datastore); | |
33a1ef7a | 160 | datastore_cache.insert(name.to_string(), datastore.clone()); |
2c32fdde | 161 | |
4bc84a65 HL |
162 | Ok(Arc::new(Self { |
163 | inner: datastore, | |
164 | operation, | |
165 | })) | |
2c32fdde DM |
166 | } |
167 | ||
062cf75c | 168 | /// removes all datastores that are not configured anymore |
42c2b5be | 169 | pub fn remove_unused_datastores() -> Result<(), Error> { |
e7d4be9d | 170 | let (config, _digest) = pbs_config::datastore::config()?; |
062cf75c DC |
171 | |
172 | let mut map = DATASTORE_MAP.lock().unwrap(); | |
173 | // removes all elements that are not in the config | |
42c2b5be | 174 | map.retain(|key, _| config.sections.contains_key(key)); |
062cf75c DC |
175 | Ok(()) |
176 | } | |
177 | ||
6da20161 | 178 | /// Open a raw database given a name and a path. |
c66fa32c TL |
179 | /// |
180 | /// # Safety | |
181 | /// See the safety section in `open_from_config` | |
519ca9d0 | 182 | pub unsafe fn open_path( |
6da20161 WB |
183 | name: &str, |
184 | path: impl AsRef<Path>, | |
185 | operation: Option<Operation>, | |
186 | ) -> Result<Arc<Self>, Error> { | |
187 | let path = path | |
188 | .as_ref() | |
189 | .to_str() | |
190 | .ok_or_else(|| format_err!("non-utf8 paths not supported"))? | |
191 | .to_owned(); | |
192 | unsafe { Self::open_from_config(DataStoreConfig::new(name.to_owned(), path), operation) } | |
193 | } | |
194 | ||
195 | /// Open a datastore given a raw configuration. | |
c66fa32c TL |
196 | /// |
197 | /// # Safety | |
74cad4a8 | 198 | /// There's no memory safety implication, but as this is opening a new ChunkStore it will |
c66fa32c TL |
199 | /// create a new process locker instance, potentially on the same path as existing safely |
200 | /// created ones. This is dangerous as dropping the reference of this and thus the underlying | |
201 | /// chunkstore's process locker will close all locks from our process on the config.path, | |
202 | /// breaking guarantees we need to uphold for safe long backup + GC interaction on newer/older | |
203 | /// process instances (from package update). | |
519ca9d0 | 204 | unsafe fn open_from_config( |
6da20161 WB |
205 | config: DataStoreConfig, |
206 | operation: Option<Operation>, | |
207 | ) -> Result<Arc<Self>, Error> { | |
208 | let name = config.name.clone(); | |
209 | ||
210 | let chunk_store = ChunkStore::open(&name, &config.path)?; | |
0bd9c870 DC |
211 | let inner = Arc::new(Self::with_store_and_config( |
212 | Arc::new(chunk_store), | |
213 | config, | |
51d900d1 | 214 | None, |
0bd9c870 | 215 | )?); |
6da20161 WB |
216 | |
217 | if let Some(operation) = operation { | |
218 | update_active_operations(&name, operation, 1)?; | |
219 | } | |
220 | ||
221 | Ok(Arc::new(Self { inner, operation })) | |
222 | } | |
223 | ||
224 | fn with_store_and_config( | |
0bd9c870 | 225 | chunk_store: Arc<ChunkStore>, |
118deb4d | 226 | config: DataStoreConfig, |
51d900d1 | 227 | last_digest: Option<[u8; 32]>, |
4bc84a65 | 228 | ) -> Result<DataStoreImpl, Error> { |
b683fd58 DC |
229 | let mut gc_status_path = chunk_store.base_path(); |
230 | gc_status_path.push(".gc-status"); | |
231 | ||
232 | let gc_status = if let Some(state) = file_read_optional_string(gc_status_path)? { | |
233 | match serde_json::from_str(&state) { | |
234 | Ok(state) => state, | |
235 | Err(err) => { | |
dce4b540 | 236 | log::error!("error reading gc-status: {}", err); |
b683fd58 DC |
237 | GarbageCollectionStatus::default() |
238 | } | |
239 | } | |
240 | } else { | |
241 | GarbageCollectionStatus::default() | |
242 | }; | |
f2b99c34 | 243 | |
fef61684 | 244 | let tuning: DatastoreTuning = serde_json::from_value( |
42c2b5be TL |
245 | DatastoreTuning::API_SCHEMA |
246 | .parse_property_string(config.tuning.as_deref().unwrap_or(""))?, | |
fef61684 DC |
247 | )?; |
248 | let chunk_order = tuning.chunk_order.unwrap_or(ChunkOrder::Inode); | |
249 | ||
4bc84a65 | 250 | Ok(DataStoreImpl { |
0bd9c870 | 251 | chunk_store, |
81b2a872 | 252 | gc_mutex: Mutex::new(()), |
f2b99c34 | 253 | last_gc_status: Mutex::new(gc_status), |
0698f78d | 254 | verify_new: config.verify_new.unwrap_or(false), |
fef61684 | 255 | chunk_order, |
51d900d1 | 256 | last_digest, |
529de6c7 DM |
257 | }) |
258 | } | |
259 | ||
d59397e6 WB |
260 | pub fn get_chunk_iterator( |
261 | &self, | |
262 | ) -> Result< | |
25877d05 | 263 | impl Iterator<Item = (Result<proxmox_sys::fs::ReadDirEntry, Error>, usize, bool)>, |
42c2b5be | 264 | Error, |
d59397e6 | 265 | > { |
4bc84a65 | 266 | self.inner.chunk_store.get_chunk_iterator() |
d59397e6 WB |
267 | } |
268 | ||
42c2b5be TL |
269 | pub fn create_fixed_writer<P: AsRef<Path>>( |
270 | &self, | |
271 | filename: P, | |
272 | size: usize, | |
273 | chunk_size: usize, | |
274 | ) -> Result<FixedIndexWriter, Error> { | |
275 | let index = FixedIndexWriter::create( | |
276 | self.inner.chunk_store.clone(), | |
277 | filename.as_ref(), | |
278 | size, | |
279 | chunk_size, | |
280 | )?; | |
529de6c7 DM |
281 | |
282 | Ok(index) | |
283 | } | |
284 | ||
42c2b5be TL |
285 | pub fn open_fixed_reader<P: AsRef<Path>>( |
286 | &self, | |
287 | filename: P, | |
288 | ) -> Result<FixedIndexReader, Error> { | |
289 | let full_path = self.inner.chunk_store.relative_path(filename.as_ref()); | |
a7c72ad9 DM |
290 | |
291 | let index = FixedIndexReader::open(&full_path)?; | |
529de6c7 DM |
292 | |
293 | Ok(index) | |
294 | } | |
3d5c11e5 | 295 | |
93d5d779 | 296 | pub fn create_dynamic_writer<P: AsRef<Path>>( |
42c2b5be TL |
297 | &self, |
298 | filename: P, | |
93d5d779 | 299 | ) -> Result<DynamicIndexWriter, Error> { |
42c2b5be | 300 | let index = DynamicIndexWriter::create(self.inner.chunk_store.clone(), filename.as_ref())?; |
0433db19 DM |
301 | |
302 | Ok(index) | |
303 | } | |
ff3d3100 | 304 | |
42c2b5be TL |
305 | pub fn open_dynamic_reader<P: AsRef<Path>>( |
306 | &self, | |
307 | filename: P, | |
308 | ) -> Result<DynamicIndexReader, Error> { | |
309 | let full_path = self.inner.chunk_store.relative_path(filename.as_ref()); | |
d48a9955 DM |
310 | |
311 | let index = DynamicIndexReader::open(&full_path)?; | |
77703d95 DM |
312 | |
313 | Ok(index) | |
314 | } | |
315 | ||
5de2bced WB |
316 | pub fn open_index<P>(&self, filename: P) -> Result<Box<dyn IndexFile + Send>, Error> |
317 | where | |
318 | P: AsRef<Path>, | |
319 | { | |
320 | let filename = filename.as_ref(); | |
42c2b5be TL |
321 | let out: Box<dyn IndexFile + Send> = match archive_type(filename)? { |
322 | ArchiveType::DynamicIndex => Box::new(self.open_dynamic_reader(filename)?), | |
323 | ArchiveType::FixedIndex => Box::new(self.open_fixed_reader(filename)?), | |
324 | _ => bail!("cannot open index file of unknown type: {:?}", filename), | |
325 | }; | |
5de2bced WB |
326 | Ok(out) |
327 | } | |
328 | ||
1369bcdb | 329 | /// Fast index verification - only check if chunks exists |
28570d19 DM |
330 | pub fn fast_index_verification( |
331 | &self, | |
332 | index: &dyn IndexFile, | |
42c2b5be | 333 | checked: &mut HashSet<[u8; 32]>, |
28570d19 | 334 | ) -> Result<(), Error> { |
1369bcdb DM |
335 | for pos in 0..index.index_count() { |
336 | let info = index.chunk_info(pos).unwrap(); | |
28570d19 DM |
337 | if checked.contains(&info.digest) { |
338 | continue; | |
339 | } | |
340 | ||
42c2b5be TL |
341 | self.stat_chunk(&info.digest).map_err(|err| { |
342 | format_err!( | |
343 | "fast_index_verification error, stat_chunk {} failed - {}", | |
344 | hex::encode(&info.digest), | |
345 | err, | |
346 | ) | |
347 | })?; | |
28570d19 DM |
348 | |
349 | checked.insert(info.digest); | |
1369bcdb DM |
350 | } |
351 | ||
352 | Ok(()) | |
353 | } | |
354 | ||
60f9a6ea | 355 | pub fn name(&self) -> &str { |
4bc84a65 | 356 | self.inner.chunk_store.name() |
60f9a6ea DM |
357 | } |
358 | ||
ff3d3100 | 359 | pub fn base_path(&self) -> PathBuf { |
4bc84a65 | 360 | self.inner.chunk_store.base_path() |
ff3d3100 DM |
361 | } |
362 | ||
133d718f | 363 | /// Returns the absolute path for a backup namespace on this datastore |
8c74349b WB |
364 | pub fn namespace_path(&self, ns: &BackupNamespace) -> PathBuf { |
365 | let mut path = self.base_path(); | |
366 | path.reserve(ns.path_len()); | |
367 | for part in ns.components() { | |
368 | path.push("ns"); | |
369 | path.push(part); | |
370 | } | |
371 | path | |
372 | } | |
373 | ||
5444b914 WB |
374 | /// Returns the absolute path for a backup_type |
375 | pub fn type_path(&self, ns: &BackupNamespace, backup_type: BackupType) -> PathBuf { | |
376 | let mut full_path = self.namespace_path(ns); | |
377 | full_path.push(backup_type.to_string()); | |
378 | full_path | |
379 | } | |
380 | ||
41b373ec | 381 | /// Returns the absolute path for a backup_group |
133d718f WB |
382 | pub fn group_path( |
383 | &self, | |
384 | ns: &BackupNamespace, | |
385 | backup_group: &pbs_api_types::BackupGroup, | |
386 | ) -> PathBuf { | |
387 | let mut full_path = self.namespace_path(ns); | |
db87d93e | 388 | full_path.push(backup_group.to_string()); |
41b373ec DM |
389 | full_path |
390 | } | |
391 | ||
392 | /// Returns the absolute path for backup_dir | |
133d718f WB |
393 | pub fn snapshot_path( |
394 | &self, | |
395 | ns: &BackupNamespace, | |
396 | backup_dir: &pbs_api_types::BackupDir, | |
397 | ) -> PathBuf { | |
398 | let mut full_path = self.namespace_path(ns); | |
db87d93e | 399 | full_path.push(backup_dir.to_string()); |
41b373ec DM |
400 | full_path |
401 | } | |
402 | ||
dc3d716b TL |
403 | /// Create a backup namespace. |
404 | pub fn create_namespace( | |
405 | self: &Arc<Self>, | |
406 | parent: &BackupNamespace, | |
407 | name: String, | |
408 | ) -> Result<BackupNamespace, Error> { | |
fc99c279 | 409 | if !self.namespace_exists(parent) { |
dc3d716b TL |
410 | bail!("cannot create new namespace, parent {parent} doesn't already exists"); |
411 | } | |
412 | ||
413 | // construct ns before mkdir to enforce max-depth and name validity | |
414 | let ns = BackupNamespace::from_parent_ns(parent, name)?; | |
415 | ||
ca3f8757 | 416 | let mut ns_full_path = self.base_path(); |
dc3d716b TL |
417 | ns_full_path.push(ns.path()); |
418 | ||
419 | std::fs::create_dir_all(ns_full_path)?; | |
420 | ||
421 | Ok(ns) | |
422 | } | |
423 | ||
fc99c279 DC |
424 | /// Returns if the given namespace exists on the datastore |
425 | pub fn namespace_exists(&self, ns: &BackupNamespace) -> bool { | |
ca3f8757 | 426 | let mut path = self.base_path(); |
fc99c279 DC |
427 | path.push(ns.path()); |
428 | path.exists() | |
429 | } | |
430 | ||
4c7cc5b3 TL |
431 | /// Remove all backup groups of a single namespace level but not the namespace itself. |
432 | /// | |
433 | /// Does *not* descends into child-namespaces and doesn't remoes the namespace itself either. | |
434 | /// | |
435 | /// Returns true if all the groups were removed, and false if some were protected. | |
436 | pub fn remove_namespace_groups(self: &Arc<Self>, ns: &BackupNamespace) -> Result<bool, Error> { | |
437 | // FIXME: locking? The single groups/snapshots are already protected, so may not be | |
74cad4a8 | 438 | // necessary (depends on what we all allow to do with namespaces) |
4c7cc5b3 TL |
439 | log::info!("removing all groups in namespace {}:/{ns}", self.name()); |
440 | ||
441 | let mut removed_all_groups = true; | |
442 | ||
443 | for group in self.iter_backup_groups(ns.to_owned())? { | |
444 | let removed_group = group?.destroy()?; | |
445 | removed_all_groups = removed_all_groups && removed_group; | |
446 | } | |
447 | ||
448 | let base_file = std::fs::File::open(self.base_path())?; | |
449 | let base_fd = base_file.as_raw_fd(); | |
450 | for ty in BackupType::iter() { | |
451 | let mut ty_dir = ns.path(); | |
452 | ty_dir.push(ty.to_string()); | |
453 | // best effort only, but we probably should log the error | |
454 | if let Err(err) = unlinkat(Some(base_fd), &ty_dir, UnlinkatFlags::RemoveDir) { | |
11ca8343 | 455 | if err != nix::errno::Errno::ENOENT { |
4c7cc5b3 TL |
456 | log::error!("failed to remove backup type {ty} in {ns} - {err}"); |
457 | } | |
458 | } | |
459 | } | |
460 | ||
461 | Ok(removed_all_groups) | |
462 | } | |
463 | ||
d1f9ccea TL |
464 | /// Remove a complete backup namespace optionally including all it's, and child namespaces', |
465 | /// groups. If `removed_groups` is false this only prunes empty namespaces. | |
4c7cc5b3 | 466 | /// |
d1f9ccea TL |
467 | /// Returns true if everything requested, and false if some groups were protected or if some |
468 | /// namespaces weren't empty even though all groups were deleted (race with new backup) | |
4c7cc5b3 TL |
469 | pub fn remove_namespace_recursive( |
470 | self: &Arc<Self>, | |
471 | ns: &BackupNamespace, | |
d1f9ccea | 472 | delete_groups: bool, |
4c7cc5b3 | 473 | ) -> Result<bool, Error> { |
d1f9ccea TL |
474 | let store = self.name(); |
475 | let mut removed_all_requested = true; | |
476 | if delete_groups { | |
477 | log::info!("removing whole namespace recursively below {store}:/{ns}",); | |
478 | for ns in self.recursive_iter_backup_ns(ns.to_owned())? { | |
479 | let removed_ns_groups = self.remove_namespace_groups(&ns?)?; | |
480 | removed_all_requested = removed_all_requested && removed_ns_groups; | |
481 | } | |
482 | } else { | |
483 | log::info!("pruning empty namespace recursively below {store}:/{ns}"); | |
4c7cc5b3 TL |
484 | } |
485 | ||
486 | // now try to delete the actual namespaces, bottom up so that we can use safe rmdir that | |
487 | // will choke if a new backup/group appeared in the meantime (but not on an new empty NS) | |
488 | let mut children = self | |
489 | .recursive_iter_backup_ns(ns.to_owned())? | |
490 | .collect::<Result<Vec<BackupNamespace>, Error>>()?; | |
491 | ||
bc001e12 | 492 | children.sort_by_key(|b| std::cmp::Reverse(b.depth())); |
4c7cc5b3 TL |
493 | |
494 | let base_file = std::fs::File::open(self.base_path())?; | |
495 | let base_fd = base_file.as_raw_fd(); | |
496 | ||
497 | for ns in children.iter() { | |
498 | let mut ns_dir = ns.path(); | |
499 | ns_dir.push("ns"); | |
500 | let _ = unlinkat(Some(base_fd), &ns_dir, UnlinkatFlags::RemoveDir); | |
501 | ||
502 | if !ns.is_root() { | |
503 | match unlinkat(Some(base_fd), &ns.path(), UnlinkatFlags::RemoveDir) { | |
d1f9ccea | 504 | Ok(()) => log::debug!("removed namespace {ns}"), |
11ca8343 | 505 | Err(nix::errno::Errno::ENOENT) => { |
d1f9ccea TL |
506 | log::debug!("namespace {ns} already removed") |
507 | } | |
11ca8343 | 508 | Err(nix::errno::Errno::ENOTEMPTY) if !delete_groups => { |
3aafa613 | 509 | removed_all_requested = false; |
d1f9ccea TL |
510 | log::debug!("skip removal of non-empty namespace {ns}") |
511 | } | |
512 | Err(err) => { | |
513 | removed_all_requested = false; | |
514 | log::warn!("failed to remove namespace {ns} - {err}") | |
515 | } | |
4c7cc5b3 TL |
516 | } |
517 | } | |
518 | } | |
519 | ||
d1f9ccea | 520 | Ok(removed_all_requested) |
4c7cc5b3 TL |
521 | } |
522 | ||
f03649b8 TL |
523 | /// Remove a complete backup group including all snapshots. |
524 | /// | |
525 | /// Returns true if all snapshots were removed, and false if some were protected | |
db87d93e | 526 | pub fn remove_backup_group( |
6da20161 | 527 | self: &Arc<Self>, |
133d718f | 528 | ns: &BackupNamespace, |
db87d93e WB |
529 | backup_group: &pbs_api_types::BackupGroup, |
530 | ) -> Result<bool, Error> { | |
133d718f | 531 | let backup_group = self.backup_group(ns.clone(), backup_group.clone()); |
db87d93e | 532 | |
f03649b8 | 533 | backup_group.destroy() |
4b4eba0b DM |
534 | } |
535 | ||
8f579717 | 536 | /// Remove a backup directory including all content |
db87d93e | 537 | pub fn remove_backup_dir( |
6da20161 | 538 | self: &Arc<Self>, |
133d718f | 539 | ns: &BackupNamespace, |
db87d93e WB |
540 | backup_dir: &pbs_api_types::BackupDir, |
541 | force: bool, | |
542 | ) -> Result<(), Error> { | |
133d718f | 543 | let backup_dir = self.backup_dir(ns.clone(), backup_dir.clone())?; |
db87d93e | 544 | |
f03649b8 | 545 | backup_dir.destroy(force) |
8f579717 DM |
546 | } |
547 | ||
41b373ec DM |
548 | /// Returns the time of the last successful backup |
549 | /// | |
550 | /// Or None if there is no backup in the group (or the group dir does not exist). | |
db87d93e | 551 | pub fn last_successful_backup( |
6da20161 | 552 | self: &Arc<Self>, |
133d718f | 553 | ns: &BackupNamespace, |
db87d93e WB |
554 | backup_group: &pbs_api_types::BackupGroup, |
555 | ) -> Result<Option<i64>, Error> { | |
133d718f | 556 | let backup_group = self.backup_group(ns.clone(), backup_group.clone()); |
db87d93e | 557 | |
4b77d300 | 558 | let group_path = backup_group.full_group_path(); |
41b373ec DM |
559 | |
560 | if group_path.exists() { | |
6da20161 | 561 | backup_group.last_successful_backup() |
41b373ec DM |
562 | } else { |
563 | Ok(None) | |
564 | } | |
565 | } | |
566 | ||
133d718f WB |
567 | /// Return the path of the 'owner' file. |
568 | fn owner_path(&self, ns: &BackupNamespace, group: &pbs_api_types::BackupGroup) -> PathBuf { | |
569 | self.group_path(ns, group).join("owner") | |
570 | } | |
571 | ||
54552dda DM |
572 | /// Returns the backup owner. |
573 | /// | |
e6dc35ac | 574 | /// The backup owner is the entity who first created the backup group. |
133d718f WB |
575 | pub fn get_owner( |
576 | &self, | |
577 | ns: &BackupNamespace, | |
578 | backup_group: &pbs_api_types::BackupGroup, | |
579 | ) -> Result<Authid, Error> { | |
580 | let full_path = self.owner_path(ns, backup_group); | |
25877d05 | 581 | let owner = proxmox_sys::fs::file_read_firstline(full_path)?; |
dcf5a0f6 | 582 | owner.trim_end().parse() // remove trailing newline |
54552dda DM |
583 | } |
584 | ||
db87d93e WB |
585 | pub fn owns_backup( |
586 | &self, | |
133d718f | 587 | ns: &BackupNamespace, |
db87d93e WB |
588 | backup_group: &pbs_api_types::BackupGroup, |
589 | auth_id: &Authid, | |
590 | ) -> Result<bool, Error> { | |
133d718f | 591 | let owner = self.get_owner(ns, backup_group)?; |
9751ef4b | 592 | |
8e0b852f | 593 | Ok(check_backup_owner(&owner, auth_id).is_ok()) |
9751ef4b DC |
594 | } |
595 | ||
54552dda | 596 | /// Set the backup owner. |
e7cb4dc5 WB |
597 | pub fn set_owner( |
598 | &self, | |
133d718f | 599 | ns: &BackupNamespace, |
db87d93e | 600 | backup_group: &pbs_api_types::BackupGroup, |
e6dc35ac | 601 | auth_id: &Authid, |
e7cb4dc5 WB |
602 | force: bool, |
603 | ) -> Result<(), Error> { | |
133d718f | 604 | let path = self.owner_path(ns, backup_group); |
54552dda DM |
605 | |
606 | let mut open_options = std::fs::OpenOptions::new(); | |
607 | open_options.write(true); | |
608 | open_options.truncate(true); | |
609 | ||
610 | if force { | |
611 | open_options.create(true); | |
612 | } else { | |
613 | open_options.create_new(true); | |
614 | } | |
615 | ||
42c2b5be TL |
616 | let mut file = open_options |
617 | .open(&path) | |
54552dda DM |
618 | .map_err(|err| format_err!("unable to create owner file {:?} - {}", path, err))?; |
619 | ||
e6dc35ac | 620 | writeln!(file, "{}", auth_id) |
54552dda DM |
621 | .map_err(|err| format_err!("unable to write owner file {:?} - {}", path, err))?; |
622 | ||
623 | Ok(()) | |
624 | } | |
625 | ||
1fc82c41 | 626 | /// Create (if it does not already exists) and lock a backup group |
54552dda DM |
627 | /// |
628 | /// And set the owner to 'userid'. If the group already exists, it returns the | |
629 | /// current owner (instead of setting the owner). | |
1fc82c41 | 630 | /// |
1ffe0301 | 631 | /// This also acquires an exclusive lock on the directory and returns the lock guard. |
e7cb4dc5 WB |
632 | pub fn create_locked_backup_group( |
633 | &self, | |
133d718f | 634 | ns: &BackupNamespace, |
db87d93e | 635 | backup_group: &pbs_api_types::BackupGroup, |
e6dc35ac FG |
636 | auth_id: &Authid, |
637 | ) -> Result<(Authid, DirLockGuard), Error> { | |
8731e40a | 638 | // create intermediate path first: |
44288184 | 639 | let mut full_path = self.base_path(); |
133d718f | 640 | for ns in ns.components() { |
8c74349b WB |
641 | full_path.push("ns"); |
642 | full_path.push(ns); | |
643 | } | |
db87d93e | 644 | full_path.push(backup_group.ty.as_str()); |
8731e40a WB |
645 | std::fs::create_dir_all(&full_path)?; |
646 | ||
db87d93e | 647 | full_path.push(&backup_group.id); |
54552dda DM |
648 | |
649 | // create the last component now | |
650 | match std::fs::create_dir(&full_path) { | |
651 | Ok(_) => { | |
42c2b5be TL |
652 | let guard = lock_dir_noblock( |
653 | &full_path, | |
654 | "backup group", | |
655 | "another backup is already running", | |
656 | )?; | |
133d718f WB |
657 | self.set_owner(ns, backup_group, auth_id, false)?; |
658 | let owner = self.get_owner(ns, backup_group)?; // just to be sure | |
1fc82c41 | 659 | Ok((owner, guard)) |
54552dda DM |
660 | } |
661 | Err(ref err) if err.kind() == io::ErrorKind::AlreadyExists => { | |
42c2b5be TL |
662 | let guard = lock_dir_noblock( |
663 | &full_path, | |
664 | "backup group", | |
665 | "another backup is already running", | |
666 | )?; | |
133d718f | 667 | let owner = self.get_owner(ns, backup_group)?; // just to be sure |
1fc82c41 | 668 | Ok((owner, guard)) |
54552dda DM |
669 | } |
670 | Err(err) => bail!("unable to create backup group {:?} - {}", full_path, err), | |
671 | } | |
672 | } | |
673 | ||
674 | /// Creates a new backup snapshot inside a BackupGroup | |
675 | /// | |
676 | /// The BackupGroup directory needs to exist. | |
42c2b5be TL |
677 | pub fn create_locked_backup_dir( |
678 | &self, | |
133d718f | 679 | ns: &BackupNamespace, |
db87d93e | 680 | backup_dir: &pbs_api_types::BackupDir, |
42c2b5be | 681 | ) -> Result<(PathBuf, bool, DirLockGuard), Error> { |
133d718f WB |
682 | let full_path = self.snapshot_path(ns, backup_dir); |
683 | let relative_path = full_path.strip_prefix(self.base_path()).map_err(|err| { | |
684 | format_err!( | |
685 | "failed to produce correct path for backup {backup_dir} in namespace {ns}: {err}" | |
686 | ) | |
687 | })?; | |
ff3d3100 | 688 | |
42c2b5be TL |
689 | let lock = || { |
690 | lock_dir_noblock( | |
691 | &full_path, | |
692 | "snapshot", | |
693 | "internal error - tried creating snapshot that's already in use", | |
694 | ) | |
695 | }; | |
f23f7543 | 696 | |
8731e40a | 697 | match std::fs::create_dir(&full_path) { |
133d718f | 698 | Ok(_) => Ok((relative_path.to_owned(), true, lock()?)), |
42c2b5be | 699 | Err(ref e) if e.kind() == io::ErrorKind::AlreadyExists => { |
133d718f | 700 | Ok((relative_path.to_owned(), false, lock()?)) |
42c2b5be TL |
701 | } |
702 | Err(e) => Err(e.into()), | |
8731e40a | 703 | } |
ff3d3100 DM |
704 | } |
705 | ||
90e38696 TL |
706 | /// Get a streaming iter over single-level backup namespaces of a datatstore |
707 | /// | |
708 | /// The iterated item is still a Result that can contain errors from rather unexptected FS or | |
709 | /// parsing errors. | |
710 | pub fn iter_backup_ns( | |
711 | self: &Arc<DataStore>, | |
712 | ns: BackupNamespace, | |
713 | ) -> Result<ListNamespaces, Error> { | |
714 | ListNamespaces::new(Arc::clone(self), ns) | |
715 | } | |
716 | ||
717 | /// Get a streaming iter over single-level backup namespaces of a datatstore, filtered by Ok | |
718 | /// | |
719 | /// The iterated item's result is already unwrapped, if it contained an error it will be | |
720 | /// logged. Can be useful in iterator chain commands | |
721 | pub fn iter_backup_ns_ok( | |
722 | self: &Arc<DataStore>, | |
723 | ns: BackupNamespace, | |
724 | ) -> Result<impl Iterator<Item = BackupNamespace> + 'static, Error> { | |
725 | let this = Arc::clone(self); | |
726 | Ok( | |
ca3f8757 | 727 | ListNamespaces::new(Arc::clone(self), ns)?.filter_map(move |ns| match ns { |
90e38696 TL |
728 | Ok(ns) => Some(ns), |
729 | Err(err) => { | |
730 | log::error!("list groups error on datastore {} - {}", this.name(), err); | |
731 | None | |
732 | } | |
733 | }), | |
734 | ) | |
735 | } | |
736 | ||
737 | /// Get a streaming iter over single-level backup namespaces of a datatstore | |
738 | /// | |
739 | /// The iterated item is still a Result that can contain errors from rather unexptected FS or | |
740 | /// parsing errors. | |
741 | pub fn recursive_iter_backup_ns( | |
742 | self: &Arc<DataStore>, | |
743 | ns: BackupNamespace, | |
744 | ) -> Result<ListNamespacesRecursive, Error> { | |
745 | ListNamespacesRecursive::new(Arc::clone(self), ns) | |
746 | } | |
747 | ||
748 | /// Get a streaming iter over single-level backup namespaces of a datatstore, filtered by Ok | |
749 | /// | |
750 | /// The iterated item's result is already unwrapped, if it contained an error it will be | |
751 | /// logged. Can be useful in iterator chain commands | |
752 | pub fn recursive_iter_backup_ns_ok( | |
753 | self: &Arc<DataStore>, | |
754 | ns: BackupNamespace, | |
15a92724 | 755 | max_depth: Option<usize>, |
90e38696 TL |
756 | ) -> Result<impl Iterator<Item = BackupNamespace> + 'static, Error> { |
757 | let this = Arc::clone(self); | |
15a92724 | 758 | Ok(if let Some(depth) = max_depth { |
ca3f8757 | 759 | ListNamespacesRecursive::new_max_depth(Arc::clone(self), ns, depth)? |
15a92724 | 760 | } else { |
ca3f8757 | 761 | ListNamespacesRecursive::new(Arc::clone(self), ns)? |
15a92724 TL |
762 | } |
763 | .filter_map(move |ns| match ns { | |
764 | Ok(ns) => Some(ns), | |
765 | Err(err) => { | |
766 | log::error!("list groups error on datastore {} - {}", this.name(), err); | |
767 | None | |
768 | } | |
769 | })) | |
90e38696 TL |
770 | } |
771 | ||
5444b914 WB |
772 | /// Get a streaming iter over top-level backup groups of a datatstore of a particular type. |
773 | /// | |
774 | /// The iterated item is still a Result that can contain errors from rather unexptected FS or | |
775 | /// parsing errors. | |
776 | pub fn iter_backup_type( | |
777 | self: &Arc<DataStore>, | |
778 | ns: BackupNamespace, | |
779 | ty: BackupType, | |
780 | ) -> Result<ListGroupsType, Error> { | |
781 | ListGroupsType::new(Arc::clone(self), ns, ty) | |
782 | } | |
783 | ||
784 | /// Get a streaming iter over top-level backup groups of a datatstore of a particular type, | |
785 | /// filtered by `Ok` results | |
786 | /// | |
787 | /// The iterated item's result is already unwrapped, if it contained an error it will be | |
788 | /// logged. Can be useful in iterator chain commands | |
789 | pub fn iter_backup_type_ok( | |
790 | self: &Arc<DataStore>, | |
791 | ns: BackupNamespace, | |
792 | ty: BackupType, | |
793 | ) -> Result<impl Iterator<Item = BackupGroup> + 'static, Error> { | |
794 | Ok(self.iter_backup_type(ns, ty)?.ok()) | |
795 | } | |
796 | ||
7b125de3 TL |
797 | /// Get a streaming iter over top-level backup groups of a datatstore |
798 | /// | |
799 | /// The iterated item is still a Result that can contain errors from rather unexptected FS or | |
800 | /// parsing errors. | |
8c74349b WB |
801 | pub fn iter_backup_groups( |
802 | self: &Arc<DataStore>, | |
803 | ns: BackupNamespace, | |
804 | ) -> Result<ListGroups, Error> { | |
805 | ListGroups::new(Arc::clone(self), ns) | |
7b125de3 TL |
806 | } |
807 | ||
808 | /// Get a streaming iter over top-level backup groups of a datatstore, filtered by Ok results | |
809 | /// | |
810 | /// The iterated item's result is already unwrapped, if it contained an error it will be | |
811 | /// logged. Can be useful in iterator chain commands | |
6da20161 WB |
812 | pub fn iter_backup_groups_ok( |
813 | self: &Arc<DataStore>, | |
8c74349b | 814 | ns: BackupNamespace, |
6da20161 | 815 | ) -> Result<impl Iterator<Item = BackupGroup> + 'static, Error> { |
5444b914 | 816 | Ok(self.iter_backup_groups(ns)?.ok()) |
7b125de3 TL |
817 | } |
818 | ||
c90dbb5c | 819 | /// Get a in-memory vector for all top-level backup groups of a datatstore |
7b125de3 TL |
820 | /// |
821 | /// NOTE: using the iterator directly is most often more efficient w.r.t. memory usage | |
8c74349b WB |
822 | pub fn list_backup_groups( |
823 | self: &Arc<DataStore>, | |
824 | ns: BackupNamespace, | |
825 | ) -> Result<Vec<BackupGroup>, Error> { | |
826 | ListGroups::new(Arc::clone(self), ns)?.collect() | |
c90dbb5c TL |
827 | } |
828 | ||
3d5c11e5 | 829 | pub fn list_images(&self) -> Result<Vec<PathBuf>, Error> { |
ff3d3100 | 830 | let base = self.base_path(); |
3d5c11e5 DM |
831 | |
832 | let mut list = vec![]; | |
833 | ||
95cea65b DM |
834 | use walkdir::WalkDir; |
835 | ||
84466003 | 836 | let walker = WalkDir::new(&base).into_iter(); |
95cea65b DM |
837 | |
838 | // make sure we skip .chunks (and other hidden files to keep it simple) | |
839 | fn is_hidden(entry: &walkdir::DirEntry) -> bool { | |
42c2b5be TL |
840 | entry |
841 | .file_name() | |
95cea65b | 842 | .to_str() |
d8d8af98 | 843 | .map(|s| s.starts_with('.')) |
95cea65b DM |
844 | .unwrap_or(false) |
845 | } | |
c3b090ac TL |
846 | let handle_entry_err = |err: walkdir::Error| { |
847 | if let Some(inner) = err.io_error() { | |
d08cff51 FG |
848 | if let Some(path) = err.path() { |
849 | if inner.kind() == io::ErrorKind::PermissionDenied { | |
c3b090ac TL |
850 | // only allow to skip ext4 fsck directory, avoid GC if, for example, |
851 | // a user got file permissions wrong on datastore rsync to new server | |
852 | if err.depth() > 1 || !path.ends_with("lost+found") { | |
d08cff51 | 853 | bail!("cannot continue garbage-collection safely, permission denied on: {:?}", path) |
c3b090ac | 854 | } |
d08cff51 | 855 | } else { |
42c2b5be TL |
856 | bail!( |
857 | "unexpected error on datastore traversal: {} - {:?}", | |
858 | inner, | |
859 | path | |
860 | ) | |
d08cff51 FG |
861 | } |
862 | } else { | |
863 | bail!("unexpected error on datastore traversal: {}", inner) | |
c3b090ac TL |
864 | } |
865 | } | |
866 | Ok(()) | |
867 | }; | |
95cea65b | 868 | for entry in walker.filter_entry(|e| !is_hidden(e)) { |
c3b090ac TL |
869 | let path = match entry { |
870 | Ok(entry) => entry.into_path(), | |
871 | Err(err) => { | |
872 | handle_entry_err(err)?; | |
42c2b5be TL |
873 | continue; |
874 | } | |
c3b090ac | 875 | }; |
1e8da0a7 | 876 | if let Ok(archive_type) = archive_type(&path) { |
42c2b5be TL |
877 | if archive_type == ArchiveType::FixedIndex |
878 | || archive_type == ArchiveType::DynamicIndex | |
879 | { | |
95cea65b | 880 | list.push(path); |
3d5c11e5 DM |
881 | } |
882 | } | |
883 | } | |
884 | ||
885 | Ok(list) | |
886 | } | |
887 | ||
a660978c DM |
888 | // mark chunks used by ``index`` as used |
889 | fn index_mark_used_chunks<I: IndexFile>( | |
890 | &self, | |
891 | index: I, | |
892 | file_name: &Path, // only used for error reporting | |
893 | status: &mut GarbageCollectionStatus, | |
c8449217 | 894 | worker: &dyn WorkerTaskContext, |
a660978c | 895 | ) -> Result<(), Error> { |
a660978c DM |
896 | status.index_file_count += 1; |
897 | status.index_data_bytes += index.index_bytes(); | |
898 | ||
899 | for pos in 0..index.index_count() { | |
f6b1d1cc | 900 | worker.check_abort()?; |
0fd55b08 | 901 | worker.fail_on_shutdown()?; |
a660978c | 902 | let digest = index.index_digest(pos).unwrap(); |
4bc84a65 | 903 | if !self.inner.chunk_store.cond_touch_chunk(digest, false)? { |
ca3f8757 | 904 | let hex = hex::encode(digest); |
c23192d3 | 905 | task_warn!( |
f6b1d1cc | 906 | worker, |
ca3f8757 | 907 | "warning: unable to access non-existent chunk {hex}, required by {file_name:?}" |
f6b1d1cc | 908 | ); |
fd192564 SR |
909 | |
910 | // touch any corresponding .bad files to keep them around, meaning if a chunk is | |
911 | // rewritten correctly they will be removed automatically, as well as if no index | |
912 | // file requires the chunk anymore (won't get to this loop then) | |
913 | for i in 0..=9 { | |
914 | let bad_ext = format!("{}.bad", i); | |
915 | let mut bad_path = PathBuf::new(); | |
916 | bad_path.push(self.chunk_path(digest).0); | |
917 | bad_path.set_extension(bad_ext); | |
4bc84a65 | 918 | self.inner.chunk_store.cond_touch_path(&bad_path, false)?; |
fd192564 | 919 | } |
a660978c DM |
920 | } |
921 | } | |
922 | Ok(()) | |
923 | } | |
924 | ||
f6b1d1cc WB |
925 | fn mark_used_chunks( |
926 | &self, | |
927 | status: &mut GarbageCollectionStatus, | |
c8449217 | 928 | worker: &dyn WorkerTaskContext, |
f6b1d1cc | 929 | ) -> Result<(), Error> { |
3d5c11e5 | 930 | let image_list = self.list_images()?; |
8317873c DM |
931 | let image_count = image_list.len(); |
932 | ||
8317873c DM |
933 | let mut last_percentage: usize = 0; |
934 | ||
cb4b721c FG |
935 | let mut strange_paths_count: u64 = 0; |
936 | ||
ea368a06 | 937 | for (i, img) in image_list.into_iter().enumerate() { |
f6b1d1cc | 938 | worker.check_abort()?; |
0fd55b08 | 939 | worker.fail_on_shutdown()?; |
92da93b2 | 940 | |
cb4b721c FG |
941 | if let Some(backup_dir_path) = img.parent() { |
942 | let backup_dir_path = backup_dir_path.strip_prefix(self.base_path())?; | |
943 | if let Some(backup_dir_str) = backup_dir_path.to_str() { | |
db87d93e | 944 | if pbs_api_types::BackupDir::from_str(backup_dir_str).is_err() { |
cb4b721c FG |
945 | strange_paths_count += 1; |
946 | } | |
947 | } | |
948 | } | |
949 | ||
efcac39d | 950 | match std::fs::File::open(&img) { |
e0762002 | 951 | Ok(file) => { |
788d82d9 | 952 | if let Ok(archive_type) = archive_type(&img) { |
e0762002 | 953 | if archive_type == ArchiveType::FixedIndex { |
788d82d9 | 954 | let index = FixedIndexReader::new(file).map_err(|e| { |
efcac39d | 955 | format_err!("can't read index '{}' - {}", img.to_string_lossy(), e) |
2f0b9235 | 956 | })?; |
788d82d9 | 957 | self.index_mark_used_chunks(index, &img, status, worker)?; |
e0762002 | 958 | } else if archive_type == ArchiveType::DynamicIndex { |
788d82d9 | 959 | let index = DynamicIndexReader::new(file).map_err(|e| { |
efcac39d | 960 | format_err!("can't read index '{}' - {}", img.to_string_lossy(), e) |
2f0b9235 | 961 | })?; |
788d82d9 | 962 | self.index_mark_used_chunks(index, &img, status, worker)?; |
e0762002 DM |
963 | } |
964 | } | |
965 | } | |
788d82d9 | 966 | Err(err) if err.kind() == io::ErrorKind::NotFound => (), // ignore vanished files |
efcac39d | 967 | Err(err) => bail!("can't open index {} - {}", img.to_string_lossy(), err), |
77703d95 | 968 | } |
8317873c | 969 | |
ea368a06 | 970 | let percentage = (i + 1) * 100 / image_count; |
8317873c | 971 | if percentage > last_percentage { |
c23192d3 | 972 | task_log!( |
f6b1d1cc | 973 | worker, |
7956877f | 974 | "marked {}% ({} of {} index files)", |
f6b1d1cc | 975 | percentage, |
ea368a06 | 976 | i + 1, |
f6b1d1cc WB |
977 | image_count, |
978 | ); | |
8317873c DM |
979 | last_percentage = percentage; |
980 | } | |
3d5c11e5 DM |
981 | } |
982 | ||
cb4b721c | 983 | if strange_paths_count > 0 { |
c23192d3 | 984 | task_log!( |
cb4b721c FG |
985 | worker, |
986 | "found (and marked) {} index files outside of expected directory scheme", | |
987 | strange_paths_count, | |
988 | ); | |
989 | } | |
990 | ||
3d5c11e5 | 991 | Ok(()) |
f2b99c34 DM |
992 | } |
993 | ||
994 | pub fn last_gc_status(&self) -> GarbageCollectionStatus { | |
4bc84a65 | 995 | self.inner.last_gc_status.lock().unwrap().clone() |
f2b99c34 | 996 | } |
3d5c11e5 | 997 | |
8545480a | 998 | pub fn garbage_collection_running(&self) -> bool { |
4bc84a65 | 999 | !matches!(self.inner.gc_mutex.try_lock(), Ok(_)) |
8545480a DM |
1000 | } |
1001 | ||
42c2b5be TL |
1002 | pub fn garbage_collection( |
1003 | &self, | |
1004 | worker: &dyn WorkerTaskContext, | |
1005 | upid: &UPID, | |
1006 | ) -> Result<(), Error> { | |
4bc84a65 | 1007 | if let Ok(ref mut _mutex) = self.inner.gc_mutex.try_lock() { |
c6772c92 TL |
1008 | // avoids that we run GC if an old daemon process has still a |
1009 | // running backup writer, which is not save as we have no "oldest | |
1010 | // writer" information and thus no safe atime cutoff | |
42c2b5be | 1011 | let _exclusive_lock = self.inner.chunk_store.try_exclusive_lock()?; |
43b13033 | 1012 | |
6ef1b649 | 1013 | let phase1_start_time = proxmox_time::epoch_i64(); |
42c2b5be TL |
1014 | let oldest_writer = self |
1015 | .inner | |
1016 | .chunk_store | |
1017 | .oldest_writer() | |
1018 | .unwrap_or(phase1_start_time); | |
11861a48 | 1019 | |
bc001e12 TL |
1020 | let mut gc_status = GarbageCollectionStatus { |
1021 | upid: Some(upid.to_string()), | |
1022 | ..Default::default() | |
1023 | }; | |
f6b1d1cc | 1024 | |
c23192d3 | 1025 | task_log!(worker, "Start GC phase1 (mark used chunks)"); |
f6b1d1cc WB |
1026 | |
1027 | self.mark_used_chunks(&mut gc_status, worker)?; | |
1028 | ||
c23192d3 | 1029 | task_log!(worker, "Start GC phase2 (sweep unused chunks)"); |
4bc84a65 | 1030 | self.inner.chunk_store.sweep_unused_chunks( |
f6b1d1cc WB |
1031 | oldest_writer, |
1032 | phase1_start_time, | |
1033 | &mut gc_status, | |
1034 | worker, | |
1035 | )?; | |
1036 | ||
c23192d3 | 1037 | task_log!( |
f6b1d1cc WB |
1038 | worker, |
1039 | "Removed garbage: {}", | |
1040 | HumanByte::from(gc_status.removed_bytes), | |
1041 | ); | |
c23192d3 | 1042 | task_log!(worker, "Removed chunks: {}", gc_status.removed_chunks); |
cf459b19 | 1043 | if gc_status.pending_bytes > 0 { |
c23192d3 | 1044 | task_log!( |
f6b1d1cc WB |
1045 | worker, |
1046 | "Pending removals: {} (in {} chunks)", | |
1047 | HumanByte::from(gc_status.pending_bytes), | |
1048 | gc_status.pending_chunks, | |
1049 | ); | |
cf459b19 | 1050 | } |
a9767cf7 | 1051 | if gc_status.removed_bad > 0 { |
c23192d3 | 1052 | task_log!(worker, "Removed bad chunks: {}", gc_status.removed_bad); |
a9767cf7 | 1053 | } |
cf459b19 | 1054 | |
b4fb2623 | 1055 | if gc_status.still_bad > 0 { |
c23192d3 | 1056 | task_log!(worker, "Leftover bad chunks: {}", gc_status.still_bad); |
b4fb2623 DM |
1057 | } |
1058 | ||
c23192d3 | 1059 | task_log!( |
f6b1d1cc WB |
1060 | worker, |
1061 | "Original data usage: {}", | |
1062 | HumanByte::from(gc_status.index_data_bytes), | |
1063 | ); | |
868c5852 DM |
1064 | |
1065 | if gc_status.index_data_bytes > 0 { | |
42c2b5be TL |
1066 | let comp_per = |
1067 | (gc_status.disk_bytes as f64 * 100.) / gc_status.index_data_bytes as f64; | |
c23192d3 | 1068 | task_log!( |
f6b1d1cc WB |
1069 | worker, |
1070 | "On-Disk usage: {} ({:.2}%)", | |
1071 | HumanByte::from(gc_status.disk_bytes), | |
1072 | comp_per, | |
1073 | ); | |
868c5852 DM |
1074 | } |
1075 | ||
c23192d3 | 1076 | task_log!(worker, "On-Disk chunks: {}", gc_status.disk_chunks); |
868c5852 | 1077 | |
d6373f35 | 1078 | let deduplication_factor = if gc_status.disk_bytes > 0 { |
42c2b5be | 1079 | (gc_status.index_data_bytes as f64) / (gc_status.disk_bytes as f64) |
d6373f35 DM |
1080 | } else { |
1081 | 1.0 | |
1082 | }; | |
1083 | ||
c23192d3 | 1084 | task_log!(worker, "Deduplication factor: {:.2}", deduplication_factor); |
d6373f35 | 1085 | |
868c5852 | 1086 | if gc_status.disk_chunks > 0 { |
42c2b5be | 1087 | let avg_chunk = gc_status.disk_bytes / (gc_status.disk_chunks as u64); |
c23192d3 | 1088 | task_log!(worker, "Average chunk size: {}", HumanByte::from(avg_chunk)); |
868c5852 | 1089 | } |
64e53b28 | 1090 | |
b683fd58 DC |
1091 | if let Ok(serialized) = serde_json::to_string(&gc_status) { |
1092 | let mut path = self.base_path(); | |
1093 | path.push(".gc-status"); | |
1094 | ||
21211748 | 1095 | let backup_user = pbs_config::backup_user()?; |
b683fd58 DC |
1096 | let mode = nix::sys::stat::Mode::from_bits_truncate(0o0644); |
1097 | // set the correct owner/group/permissions while saving file | |
1098 | // owner(rw) = backup, group(r)= backup | |
1099 | let options = CreateOptions::new() | |
1100 | .perm(mode) | |
1101 | .owner(backup_user.uid) | |
1102 | .group(backup_user.gid); | |
1103 | ||
1104 | // ignore errors | |
e0a19d33 | 1105 | let _ = replace_file(path, serialized.as_bytes(), options, false); |
b683fd58 DC |
1106 | } |
1107 | ||
4bc84a65 | 1108 | *self.inner.last_gc_status.lock().unwrap() = gc_status; |
64e53b28 | 1109 | } else { |
d4b59ae0 | 1110 | bail!("Start GC failed - (already running/locked)"); |
64e53b28 | 1111 | } |
3d5c11e5 DM |
1112 | |
1113 | Ok(()) | |
1114 | } | |
3b7ade9e | 1115 | |
ccc3896f | 1116 | pub fn try_shared_chunk_store_lock(&self) -> Result<ProcessLockSharedGuard, Error> { |
4bc84a65 | 1117 | self.inner.chunk_store.try_shared_lock() |
1cf5178a DM |
1118 | } |
1119 | ||
42c2b5be | 1120 | pub fn chunk_path(&self, digest: &[u8; 32]) -> (PathBuf, String) { |
4bc84a65 | 1121 | self.inner.chunk_store.chunk_path(digest) |
d48a9955 DM |
1122 | } |
1123 | ||
b298e9f1 | 1124 | pub fn cond_touch_chunk(&self, digest: &[u8; 32], assert_exists: bool) -> Result<bool, Error> { |
42c2b5be TL |
1125 | self.inner |
1126 | .chunk_store | |
b298e9f1 | 1127 | .cond_touch_chunk(digest, assert_exists) |
42c2b5be TL |
1128 | } |
1129 | ||
1130 | pub fn insert_chunk(&self, chunk: &DataBlob, digest: &[u8; 32]) -> Result<(bool, u64), Error> { | |
4bc84a65 | 1131 | self.inner.chunk_store.insert_chunk(chunk, digest) |
3b7ade9e | 1132 | } |
60f9a6ea | 1133 | |
7f394c80 | 1134 | pub fn stat_chunk(&self, digest: &[u8; 32]) -> Result<std::fs::Metadata, Error> { |
4bc84a65 | 1135 | let (chunk_path, _digest_str) = self.inner.chunk_store.chunk_path(digest); |
7f394c80 DC |
1136 | std::fs::metadata(chunk_path).map_err(Error::from) |
1137 | } | |
1138 | ||
39f18b30 | 1139 | pub fn load_chunk(&self, digest: &[u8; 32]) -> Result<DataBlob, Error> { |
4bc84a65 | 1140 | let (chunk_path, digest_str) = self.inner.chunk_store.chunk_path(digest); |
39f18b30 | 1141 | |
6ef1b649 | 1142 | proxmox_lang::try_block!({ |
39f18b30 DM |
1143 | let mut file = std::fs::File::open(&chunk_path)?; |
1144 | DataBlob::load_from_reader(&mut file) | |
42c2b5be TL |
1145 | }) |
1146 | .map_err(|err| { | |
1147 | format_err!( | |
1148 | "store '{}', unable to load chunk '{}' - {}", | |
1149 | self.name(), | |
1150 | digest_str, | |
1151 | err, | |
1152 | ) | |
1153 | }) | |
1a374fcf SR |
1154 | } |
1155 | ||
8292d3d2 | 1156 | /// Updates the protection status of the specified snapshot. |
42c2b5be | 1157 | pub fn update_protection(&self, backup_dir: &BackupDir, protection: bool) -> Result<(), Error> { |
6da20161 | 1158 | let full_path = backup_dir.full_path(); |
8292d3d2 DC |
1159 | |
1160 | let _guard = lock_dir_noblock(&full_path, "snapshot", "possibly running or in use")?; | |
1161 | ||
6da20161 | 1162 | let protected_path = backup_dir.protected_file(); |
8292d3d2 DC |
1163 | if protection { |
1164 | std::fs::File::create(protected_path) | |
1165 | .map_err(|err| format_err!("could not create protection file: {}", err))?; | |
1166 | } else if let Err(err) = std::fs::remove_file(protected_path) { | |
1167 | // ignore error for non-existing file | |
1168 | if err.kind() != std::io::ErrorKind::NotFound { | |
1169 | bail!("could not remove protection file: {}", err); | |
1170 | } | |
1171 | } | |
1172 | ||
1173 | Ok(()) | |
1174 | } | |
1175 | ||
0698f78d | 1176 | pub fn verify_new(&self) -> bool { |
4bc84a65 | 1177 | self.inner.verify_new |
0698f78d | 1178 | } |
4921a411 | 1179 | |
bc001e12 TL |
1180 | /// returns a list of chunks sorted by their inode number on disk chunks that couldn't get |
1181 | /// stat'ed are placed at the end of the list | |
4921a411 DC |
1182 | pub fn get_chunks_in_order<F, A>( |
1183 | &self, | |
1184 | index: &Box<dyn IndexFile + Send>, | |
1185 | skip_chunk: F, | |
1186 | check_abort: A, | |
1187 | ) -> Result<Vec<(usize, u64)>, Error> | |
1188 | where | |
1189 | F: Fn(&[u8; 32]) -> bool, | |
1190 | A: Fn(usize) -> Result<(), Error>, | |
1191 | { | |
1192 | let index_count = index.index_count(); | |
1193 | let mut chunk_list = Vec::with_capacity(index_count); | |
1194 | use std::os::unix::fs::MetadataExt; | |
1195 | for pos in 0..index_count { | |
1196 | check_abort(pos)?; | |
1197 | ||
1198 | let info = index.chunk_info(pos).unwrap(); | |
1199 | ||
1200 | if skip_chunk(&info.digest) { | |
1201 | continue; | |
1202 | } | |
1203 | ||
4bc84a65 | 1204 | let ino = match self.inner.chunk_order { |
fef61684 DC |
1205 | ChunkOrder::Inode => { |
1206 | match self.stat_chunk(&info.digest) { | |
1207 | Err(_) => u64::MAX, // could not stat, move to end of list | |
1208 | Ok(metadata) => metadata.ino(), | |
1209 | } | |
1210 | } | |
1211 | ChunkOrder::None => 0, | |
4921a411 DC |
1212 | }; |
1213 | ||
1214 | chunk_list.push((pos, ino)); | |
1215 | } | |
1216 | ||
4bc84a65 | 1217 | match self.inner.chunk_order { |
fef61684 DC |
1218 | // sorting by inode improves data locality, which makes it lots faster on spinners |
1219 | ChunkOrder::Inode => { | |
1220 | chunk_list.sort_unstable_by(|(_, ino_a), (_, ino_b)| ino_a.cmp(ino_b)) | |
1221 | } | |
1222 | ChunkOrder::None => {} | |
1223 | } | |
4921a411 DC |
1224 | |
1225 | Ok(chunk_list) | |
1226 | } | |
db87d93e | 1227 | |
6b0c6492 | 1228 | /// Open a backup group from this datastore. |
133d718f WB |
1229 | pub fn backup_group( |
1230 | self: &Arc<Self>, | |
1231 | ns: BackupNamespace, | |
1232 | group: pbs_api_types::BackupGroup, | |
1233 | ) -> BackupGroup { | |
ca3f8757 | 1234 | BackupGroup::new(Arc::clone(self), ns, group) |
db87d93e WB |
1235 | } |
1236 | ||
6b0c6492 | 1237 | /// Open a backup group from this datastore. |
8c74349b WB |
1238 | pub fn backup_group_from_parts<T>( |
1239 | self: &Arc<Self>, | |
1240 | ns: BackupNamespace, | |
1241 | ty: BackupType, | |
1242 | id: T, | |
1243 | ) -> BackupGroup | |
6b0c6492 WB |
1244 | where |
1245 | T: Into<String>, | |
1246 | { | |
133d718f | 1247 | self.backup_group(ns, (ty, id.into()).into()) |
6b0c6492 WB |
1248 | } |
1249 | ||
133d718f | 1250 | /* |
6b0c6492 WB |
1251 | /// Open a backup group from this datastore by backup group path such as `vm/100`. |
1252 | /// | |
1253 | /// Convenience method for `store.backup_group(path.parse()?)` | |
6da20161 | 1254 | pub fn backup_group_from_path(self: &Arc<Self>, path: &str) -> Result<BackupGroup, Error> { |
133d718f | 1255 | todo!("split out the namespace"); |
db87d93e | 1256 | } |
133d718f | 1257 | */ |
db87d93e | 1258 | |
6b0c6492 | 1259 | /// Open a snapshot (backup directory) from this datastore. |
133d718f WB |
1260 | pub fn backup_dir( |
1261 | self: &Arc<Self>, | |
1262 | ns: BackupNamespace, | |
1263 | dir: pbs_api_types::BackupDir, | |
1264 | ) -> Result<BackupDir, Error> { | |
1265 | BackupDir::with_group(self.backup_group(ns, dir.group), dir.time) | |
6b0c6492 WB |
1266 | } |
1267 | ||
1268 | /// Open a snapshot (backup directory) from this datastore. | |
db87d93e | 1269 | pub fn backup_dir_from_parts<T>( |
6da20161 | 1270 | self: &Arc<Self>, |
8c74349b | 1271 | ns: BackupNamespace, |
db87d93e WB |
1272 | ty: BackupType, |
1273 | id: T, | |
1274 | time: i64, | |
1275 | ) -> Result<BackupDir, Error> | |
1276 | where | |
1277 | T: Into<String>, | |
1278 | { | |
133d718f | 1279 | self.backup_dir(ns, (ty, id.into(), time).into()) |
db87d93e WB |
1280 | } |
1281 | ||
6b0c6492 | 1282 | /// Open a snapshot (backup directory) from this datastore with a cached rfc3339 time string. |
db87d93e | 1283 | pub fn backup_dir_with_rfc3339<T: Into<String>>( |
6da20161 | 1284 | self: &Arc<Self>, |
db87d93e WB |
1285 | group: BackupGroup, |
1286 | time_string: T, | |
1287 | ) -> Result<BackupDir, Error> { | |
1288 | BackupDir::with_rfc3339(group, time_string.into()) | |
1289 | } | |
1290 | ||
133d718f | 1291 | /* |
6b0c6492 | 1292 | /// Open a snapshot (backup directory) from this datastore by a snapshot path. |
6da20161 | 1293 | pub fn backup_dir_from_path(self: &Arc<Self>, path: &str) -> Result<BackupDir, Error> { |
133d718f | 1294 | todo!("split out the namespace"); |
db87d93e | 1295 | } |
133d718f | 1296 | */ |
529de6c7 | 1297 | } |