Line data Source code
1 : use std::collections::{HashMap, HashSet};
2 :
3 : use futures_util::{StreamExt, TryStreamExt};
4 : use pageserver::tenant::remote_timeline_client::remote_layer_path;
5 : use pageserver_api::controller_api::MetadataHealthUpdateRequest;
6 : use pageserver_api::shard::TenantShardId;
7 : use remote_storage::GenericRemoteStorage;
8 : use serde::Serialize;
9 : use tracing::{Instrument, info_span};
10 : use utils::id::TenantId;
11 : use utils::shard::ShardCount;
12 :
13 : use crate::checks::{
14 : BlobDataParseResult, RemoteTimelineBlobData, TenantObjectListing, TimelineAnalysis,
15 : branch_cleanup_and_check_errors, list_timeline_blobs,
16 : };
17 : use crate::metadata_stream::{stream_tenant_timelines, stream_tenants};
18 : use crate::{BucketConfig, NodeKind, RootTarget, TenantShardTimelineId, init_remote};
19 :
20 : #[derive(Serialize, Default)]
21 : pub struct MetadataSummary {
22 : tenant_count: usize,
23 : timeline_count: usize,
24 : timeline_shard_count: usize,
25 : /// Tenant-shard timeline (key) mapping to errors. The key has to be a string because it will be serialized to a JSON.
26 : /// The key is generated using `TenantShardTimelineId::to_string()`.
27 : with_errors: HashMap<String, Vec<String>>,
28 : /// Tenant-shard timeline (key) mapping to warnings. The key has to be a string because it will be serialized to a JSON.
29 : /// The key is generated using `TenantShardTimelineId::to_string()`.
30 : with_warnings: HashMap<String, Vec<String>>,
31 : with_orphans: HashSet<TenantShardTimelineId>,
32 : indices_by_version: HashMap<usize, usize>,
33 :
34 : #[serde(skip)]
35 : pub(crate) healthy_tenant_shards: HashSet<TenantShardId>,
36 : #[serde(skip)]
37 : pub(crate) unhealthy_tenant_shards: HashSet<TenantShardId>,
38 : }
39 :
40 : impl MetadataSummary {
41 0 : fn new() -> Self {
42 0 : Self::default()
43 0 : }
44 :
45 0 : fn update_data(&mut self, data: &RemoteTimelineBlobData) {
46 0 : self.timeline_shard_count += 1;
47 : if let BlobDataParseResult::Parsed {
48 0 : index_part,
49 : index_part_generation: _,
50 : s3_layers: _,
51 : index_part_last_modified_time: _,
52 : index_part_snapshot_time: _,
53 0 : } = &data.blob_data
54 0 : {
55 0 : *self
56 0 : .indices_by_version
57 0 : .entry(index_part.version())
58 0 : .or_insert(0) += 1;
59 0 : }
60 0 : }
61 :
62 0 : fn update_analysis(
63 0 : &mut self,
64 0 : id: &TenantShardTimelineId,
65 0 : analysis: &TimelineAnalysis,
66 0 : verbose: bool,
67 0 : ) {
68 0 : if analysis.is_healthy() {
69 0 : self.healthy_tenant_shards.insert(id.tenant_shard_id);
70 0 : } else {
71 0 : self.healthy_tenant_shards.remove(&id.tenant_shard_id);
72 0 : self.unhealthy_tenant_shards.insert(id.tenant_shard_id);
73 0 : }
74 :
75 0 : if !analysis.errors.is_empty() {
76 0 : let entry = self.with_errors.entry(id.to_string()).or_default();
77 0 : if verbose {
78 0 : entry.extend(analysis.errors.iter().cloned());
79 0 : }
80 0 : }
81 :
82 0 : if !analysis.warnings.is_empty() {
83 0 : let entry = self.with_warnings.entry(id.to_string()).or_default();
84 0 : if verbose {
85 0 : entry.extend(analysis.warnings.iter().cloned());
86 0 : }
87 0 : }
88 0 : }
89 :
90 0 : fn notify_timeline_orphan(&mut self, ttid: &TenantShardTimelineId) {
91 0 : self.with_orphans.insert(*ttid);
92 0 : }
93 :
94 : /// Long-form output for printing at end of a scan
95 0 : pub fn summary_string(&self) -> String {
96 0 : let version_summary: String = itertools::join(
97 0 : self.indices_by_version
98 0 : .iter()
99 0 : .map(|(k, v)| format!("{k}: {v}")),
100 0 : ", ",
101 0 : );
102 0 :
103 0 : format!(
104 0 : "Tenants: {}
105 0 : Timelines: {}
106 0 : Timeline-shards: {}
107 0 : With errors: {}
108 0 : With warnings: {}
109 0 : With orphan layers: {}
110 0 : Index versions: {version_summary}
111 0 : ",
112 0 : self.tenant_count,
113 0 : self.timeline_count,
114 0 : self.timeline_shard_count,
115 0 : self.with_errors.len(),
116 0 : self.with_warnings.len(),
117 0 : self.with_orphans.len(),
118 0 : )
119 0 : }
120 :
121 0 : pub fn is_fatal(&self) -> bool {
122 0 : !self.with_errors.is_empty()
123 0 : }
124 :
125 0 : pub fn is_empty(&self) -> bool {
126 0 : self.timeline_shard_count == 0
127 0 : }
128 :
129 0 : pub fn build_health_update_request(&self) -> MetadataHealthUpdateRequest {
130 0 : MetadataHealthUpdateRequest {
131 0 : healthy_tenant_shards: self.healthy_tenant_shards.clone(),
132 0 : unhealthy_tenant_shards: self.unhealthy_tenant_shards.clone(),
133 0 : }
134 0 : }
135 : }
136 :
137 : /// Scan the pageserver metadata in an S3 bucket, reporting errors and statistics.
138 0 : pub async fn scan_pageserver_metadata(
139 0 : bucket_config: BucketConfig,
140 0 : tenant_ids: Vec<TenantShardId>,
141 0 : verbose: bool,
142 0 : ) -> anyhow::Result<MetadataSummary> {
143 0 : let (remote_client, target) = init_remote(bucket_config, NodeKind::Pageserver).await?;
144 :
145 0 : let tenants = if tenant_ids.is_empty() {
146 0 : futures::future::Either::Left(stream_tenants(&remote_client, &target))
147 : } else {
148 0 : futures::future::Either::Right(futures::stream::iter(tenant_ids.into_iter().map(Ok)))
149 : };
150 :
151 : // How many tenants to process in parallel. We need to be mindful of pageservers
152 : // accessing the same per tenant prefixes, so use a lower setting than pageservers.
153 : const CONCURRENCY: usize = 32;
154 :
155 : // Generate a stream of TenantTimelineId
156 0 : let timelines = tenants.map_ok(|t| stream_tenant_timelines(&remote_client, &target, t));
157 0 : let timelines = timelines.try_buffered(CONCURRENCY);
158 0 : let timelines = timelines.try_flatten();
159 :
160 : // Generate a stream of S3TimelineBlobData
161 0 : async fn report_on_timeline(
162 0 : remote_client: &GenericRemoteStorage,
163 0 : target: &RootTarget,
164 0 : ttid: TenantShardTimelineId,
165 0 : ) -> anyhow::Result<(TenantShardTimelineId, RemoteTimelineBlobData)> {
166 0 : let data = list_timeline_blobs(remote_client, ttid, target).await?;
167 0 : Ok((ttid, data))
168 0 : }
169 0 : let timelines = timelines.map_ok(|ttid| report_on_timeline(&remote_client, &target, ttid));
170 0 : let mut timelines = std::pin::pin!(timelines.try_buffered(CONCURRENCY));
171 0 :
172 0 : // We must gather all the TenantShardTimelineId->S3TimelineBlobData for each tenant, because different
173 0 : // shards in the same tenant might refer to one anothers' keys if a shard split has happened.
174 0 :
175 0 : let mut tenant_id = None;
176 0 : let mut tenant_objects = TenantObjectListing::default();
177 0 : let mut tenant_timeline_results = Vec::new();
178 :
179 0 : async fn analyze_tenant(
180 0 : remote_client: &GenericRemoteStorage,
181 0 : tenant_id: TenantId,
182 0 : summary: &mut MetadataSummary,
183 0 : mut tenant_objects: TenantObjectListing,
184 0 : timelines: Vec<(TenantShardTimelineId, RemoteTimelineBlobData)>,
185 0 : highest_shard_count: ShardCount,
186 0 : verbose: bool,
187 0 : ) {
188 0 : summary.tenant_count += 1;
189 0 :
190 0 : let mut timeline_ids = HashSet::new();
191 0 : let mut timeline_generations = HashMap::new();
192 0 : for (ttid, data) in timelines {
193 0 : async {
194 0 : if ttid.tenant_shard_id.shard_count == highest_shard_count {
195 : // Only analyze `TenantShardId`s with highest shard count.
196 :
197 : // Stash the generation of each timeline, for later use identifying orphan layers
198 : if let BlobDataParseResult::Parsed {
199 0 : index_part,
200 0 : index_part_generation,
201 : s3_layers: _,
202 : index_part_last_modified_time: _,
203 : index_part_snapshot_time: _,
204 0 : } = &data.blob_data
205 : {
206 0 : if index_part.deleted_at.is_some() {
207 : // skip deleted timeline.
208 0 : tracing::info!(
209 0 : "Skip analysis of {} b/c timeline is already deleted",
210 : ttid
211 : );
212 0 : return;
213 0 : }
214 0 : timeline_generations.insert(ttid, *index_part_generation);
215 0 : }
216 :
217 : // Apply checks to this timeline shard's metadata, and in the process update `tenant_objects`
218 : // reference counts for layers across the tenant.
219 0 : let analysis = branch_cleanup_and_check_errors(
220 0 : remote_client,
221 0 : &ttid,
222 0 : &mut tenant_objects,
223 0 : None,
224 0 : None,
225 0 : Some(data),
226 0 : )
227 0 : .await;
228 0 : summary.update_analysis(&ttid, &analysis, verbose);
229 0 :
230 0 : timeline_ids.insert(ttid.timeline_id);
231 : } else {
232 0 : tracing::info!(
233 0 : "Skip analysis of {} b/c a lower shard count than {}",
234 : ttid,
235 : highest_shard_count.0,
236 : );
237 : }
238 0 : }
239 : .instrument(
240 0 : info_span!("analyze-timeline", shard = %ttid.tenant_shard_id.shard_slug(), timeline = %ttid.timeline_id),
241 : )
242 0 : .await
243 : }
244 :
245 0 : summary.timeline_count += timeline_ids.len();
246 :
247 : // Identifying orphan layers must be done on a tenant-wide basis, because individual
248 : // shards' layers may be referenced by other shards.
249 : //
250 : // Orphan layers are not a corruption, and not an indication of a problem. They are just
251 : // consuming some space in remote storage, and may be cleaned up at leisure.
252 0 : for (shard_index, timeline_id, layer_file, generation) in tenant_objects.get_orphans() {
253 0 : let ttid = TenantShardTimelineId {
254 0 : tenant_shard_id: TenantShardId {
255 0 : tenant_id,
256 0 : shard_count: shard_index.shard_count,
257 0 : shard_number: shard_index.shard_number,
258 0 : },
259 0 : timeline_id,
260 0 : };
261 :
262 0 : if let Some(timeline_generation) = timeline_generations.get(&ttid) {
263 0 : if &generation >= timeline_generation {
264 : // Candidate orphan layer is in the current or future generation relative
265 : // to the index we read for this timeline shard, so its absence from the index
266 : // doesn't make it an orphan: more likely, it is a case where the layer was
267 : // uploaded, but the index referencing the layer wasn't written yet.
268 0 : continue;
269 0 : }
270 0 : }
271 :
272 0 : let orphan_path = remote_layer_path(
273 0 : &tenant_id,
274 0 : &timeline_id,
275 0 : shard_index,
276 0 : &layer_file,
277 0 : generation,
278 0 : );
279 0 :
280 0 : tracing::info!("Orphan layer detected: {orphan_path}");
281 :
282 0 : summary.notify_timeline_orphan(&ttid);
283 : }
284 0 : }
285 :
286 : // Iterate through all the timeline results. These are in key-order, so
287 : // all results for the same tenant will be adjacent. We accumulate these,
288 : // and then call `analyze_tenant` to flush, when we see the next tenant ID.
289 0 : let mut summary = MetadataSummary::new();
290 0 : let mut highest_shard_count = ShardCount::MIN;
291 0 : while let Some(i) = timelines.next().await {
292 0 : let (ttid, data) = i?;
293 0 : summary.update_data(&data);
294 0 :
295 0 : match tenant_id {
296 0 : Some(prev_tenant_id) => {
297 0 : if prev_tenant_id != ttid.tenant_shard_id.tenant_id {
298 : // New tenant: analyze this tenant's timelines, clear accumulated tenant_timeline_results
299 0 : let tenant_objects = std::mem::take(&mut tenant_objects);
300 0 : let timelines = std::mem::take(&mut tenant_timeline_results);
301 0 : analyze_tenant(
302 0 : &remote_client,
303 0 : prev_tenant_id,
304 0 : &mut summary,
305 0 : tenant_objects,
306 0 : timelines,
307 0 : highest_shard_count,
308 0 : verbose,
309 0 : )
310 0 : .instrument(info_span!("analyze-tenant", tenant = %prev_tenant_id))
311 0 : .await;
312 0 : tenant_id = Some(ttid.tenant_shard_id.tenant_id);
313 0 : highest_shard_count = ttid.tenant_shard_id.shard_count;
314 0 : } else {
315 0 : highest_shard_count = highest_shard_count.max(ttid.tenant_shard_id.shard_count);
316 0 : }
317 : }
318 0 : None => {
319 0 : tenant_id = Some(ttid.tenant_shard_id.tenant_id);
320 0 : highest_shard_count = highest_shard_count.max(ttid.tenant_shard_id.shard_count);
321 0 : }
322 : }
323 :
324 0 : match &data.blob_data {
325 : BlobDataParseResult::Parsed {
326 : index_part: _,
327 0 : index_part_generation: _index_part_generation,
328 0 : s3_layers,
329 0 : index_part_last_modified_time: _,
330 0 : index_part_snapshot_time: _,
331 0 : } => {
332 0 : tenant_objects.push(ttid, s3_layers.clone());
333 0 : }
334 0 : BlobDataParseResult::Relic => (),
335 : BlobDataParseResult::Incorrect {
336 : errors: _,
337 0 : s3_layers,
338 0 : } => {
339 0 : tenant_objects.push(ttid, s3_layers.clone());
340 0 : }
341 : }
342 0 : tenant_timeline_results.push((ttid, data));
343 : }
344 :
345 0 : if !tenant_timeline_results.is_empty() {
346 0 : let tenant_id = tenant_id.expect("Must be set if results are present");
347 0 : analyze_tenant(
348 0 : &remote_client,
349 0 : tenant_id,
350 0 : &mut summary,
351 0 : tenant_objects,
352 0 : tenant_timeline_results,
353 0 : highest_shard_count,
354 0 : verbose,
355 0 : )
356 0 : .instrument(info_span!("analyze-tenant", tenant = %tenant_id))
357 0 : .await;
358 0 : }
359 :
360 0 : Ok(summary)
361 0 : }
|
