Vladsz83 commented on code in PR #11391: URL: https://github.com/apache/ignite/pull/11391#discussion_r1679400040
########## modules/core/src/main/java/org/apache/ignite/internal/processors/cache/persistence/snapshot/SnapshotCheckProcess.java: ########## @@ -0,0 +1,549 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.ignite.internal.processors.cache.persistence.snapshot; + +import java.io.File; +import java.io.Serializable; +import java.util.ArrayList; +import java.util.Collection; +import java.util.Collections; +import java.util.HashMap; +import java.util.List; +import java.util.Map; +import java.util.Set; +import java.util.UUID; +import java.util.concurrent.ConcurrentHashMap; +import java.util.concurrent.ThreadLocalRandom; +import java.util.function.Function; +import java.util.stream.Collectors; +import org.apache.ignite.IgniteException; +import org.apache.ignite.IgniteLogger; +import org.apache.ignite.cluster.ClusterNode; +import org.apache.ignite.events.DiscoveryEvent; +import org.apache.ignite.internal.GridKernalContext; +import org.apache.ignite.internal.IgniteInternalFuture; +import org.apache.ignite.internal.cluster.ClusterTopologyCheckedException; +import org.apache.ignite.internal.management.cache.IdleVerifyResultV2; +import org.apache.ignite.internal.management.cache.PartitionKeyV2; +import org.apache.ignite.internal.processors.cache.verify.PartitionHashRecordV2; +import org.apache.ignite.internal.util.distributed.DistributedProcess; +import org.apache.ignite.internal.util.future.GridFinishedFuture; +import org.apache.ignite.internal.util.future.GridFutureAdapter; +import org.apache.ignite.internal.util.typedef.F; +import org.apache.ignite.internal.util.typedef.internal.CU; +import org.jetbrains.annotations.Nullable; + +import static org.apache.ignite.events.EventType.EVT_NODE_FAILED; +import static org.apache.ignite.events.EventType.EVT_NODE_LEFT; +import static org.apache.ignite.internal.util.distributed.DistributedProcess.DistributedProcessType.SNAPSHOT_CHECK_METAS; +import static org.apache.ignite.internal.util.distributed.DistributedProcess.DistributedProcessType.SNAPSHOT_VALIDATE_PARTS; + +/** Distributed process of snapshot checking (with the partition hashes). */ +public class SnapshotCheckProcess { + /** */ + private final IgniteLogger log; + + /** */ + private final GridKernalContext kctx; + + /** Snapshot check requests per snapshot on every node. */ + private final Map<String, SnapshotCheckProcessRequest> requests = new ConcurrentHashMap<>(); + + /** Cluster-wide operation futures per snapshot called from current node. */ + private final Map<UUID, GridFutureAdapter<SnapshotPartitionsVerifyTaskResult>> clusterOpFuts = new ConcurrentHashMap<>(); + + /** Check metas first phase subprocess. */ + private final DistributedProcess<SnapshotCheckProcessRequest, CheckResultDTO> phase1CheckMetas; + + /** Partition hashes second phase subprocess. */ + private final DistributedProcess<SnapshotCheckProcessRequest, CheckResultDTO> phase2PartsHashes; + + /** */ + public SnapshotCheckProcess(GridKernalContext kctx) { + this.kctx = kctx; + + log = kctx.log(getClass()); + + phase1CheckMetas = new DistributedProcess<>(kctx, SNAPSHOT_CHECK_METAS, this::prepareAndCheckMetas, + this::reducePreparationAndMetasCheck); + + phase2PartsHashes = new DistributedProcess<>(kctx, SNAPSHOT_VALIDATE_PARTS, this::validateParts, + this::reduceValidatePartsAndFinish); + + kctx.event().addLocalEventListener((evt) -> nodeLeft(((DiscoveryEvent)evt).eventNode()), EVT_NODE_FAILED, EVT_NODE_LEFT); + } + + /** */ + Map<String, SnapshotCheckProcessRequest> requests() { + return Collections.unmodifiableMap(requests); + } + + /** + * Stops the process with the passed exception. + * + * @param th The interrupt reason. + * @param rqFilter If not {@code null}, used to filter which requests/process to stop. If {@code null}, stops all the validations. + */ + void interrupt(Throwable th, @Nullable Function<SnapshotCheckProcessRequest, Boolean> rqFilter) { + requests.values().forEach(rq -> { + if (rqFilter == null || rqFilter.apply(rq)) { + rq.error(th); + + clean(rq.requestId(), th, null, null); + } + }); + } + + /** Stops the related validation if the node is a mandatory one. */ + private void nodeLeft(ClusterNode node) { + if (node.isClient() || requests.isEmpty()) + return; + + interrupt( + new ClusterTopologyCheckedException("Snapshot checking stopped. A node left the cluster: " + node + '.'), + rq -> rq.nodes().contains(node.id()) + ); + } + + /** Phase 2 and process finish. */ + private IgniteInternalFuture<?> reduceValidatePartsAndFinish( + UUID procId, + Map<UUID, CheckResultDTO> results, + Map<UUID, Throwable> errors + ) { + clean(procId, null, results, errors); + + return new GridFinishedFuture<>(); + } + + /** Phase 2 beginning. */ + private IgniteInternalFuture<CheckResultDTO> validateParts(SnapshotCheckProcessRequest req) { + SnapshotCheckProcessRequest locReq; + + if (stopAndCleanOnError(req, null) || (locReq = requests.get(req.snapshotName())) == null) + return new GridFinishedFuture<>(); + + assert locReq.equals(req); + + // Store metas to collect cluster operation result later. + GridFutureAdapter<SnapshotPartitionsVerifyTaskResult> clusterOpFut = clusterOpFuts.get(req.requestId()); + + if (clusterOpFut != null) + locReq.metas = req.metas; + + // Local meta might be null if current node started after the snapshot creation or placement. + if (!req.nodes.contains(kctx.localNodeId()) || locReq.meta() == null) + return new GridFinishedFuture<>(); + + GridFutureAdapter<CheckResultDTO> locPartsChkFut = new GridFutureAdapter<>(); + + stopFutureOnAnyFailure(locPartsChkFut, () -> { + locReq.fut(locPartsChkFut); + + // An error can occure when the local future is still null. + Throwable lorReqErr = locReq.error(); + + if (lorReqErr != null) + locPartsChkFut.onDone(lorReqErr); + + if (!locPartsChkFut.isDone()) { + File snpDir = kctx.cache().context().snapshotMgr().snapshotLocalDir(locReq.snapshotName(), locReq.snapshotPath()); + + kctx.cache().context().snapshotMgr().checker().checkPartitions(locReq.meta(), snpDir, locReq.groups(), false, true, false) + .whenComplete((res, err) -> { + if (err != null) + locPartsChkFut.onDone(err); + else + locPartsChkFut.onDone(new CheckResultDTO(res)); + }); + } + }); + + return locPartsChkFut; + } + + /** + * If required, stops and clean related validation if errors occured + * + * @return {@code True} if the validation was stopped and cleaned. {@code False} otherwise. + */ + private boolean stopAndCleanOnError(SnapshotCheckProcessRequest req, @Nullable Map<UUID, Throwable> occurredErrors) { + assert req != null; + + assert F.isEmpty(occurredErrors) || req.nodes().containsAll(occurredErrors.keySet()); + + if (F.isEmpty(occurredErrors) && req.error() == null) + return false; + + clean(req.requestId(), req.error(), null, occurredErrors); + + return true; + } + + /** Cleans certain snapshot validation. */ + private void clean( + UUID rqId, + @Nullable Throwable opErr, + @Nullable Map<UUID, CheckResultDTO> results, + @Nullable Map<UUID, Throwable> errors + ) { + GridFutureAdapter<SnapshotPartitionsVerifyTaskResult> clusterOpFut = clusterOpFuts.remove(rqId); + + stopFutureOnAnyFailure(clusterOpFut, () -> { + SnapshotCheckProcessRequest locRq = currentRequest(null, rqId); + + Throwable err = opErr; + + if (locRq != null) + err = stopAndCleanLocRequest(locRq, err); + + if (clusterOpFut == null || clusterOpFut.isDone()) + return; + + boolean finished; + + // Nodes' results and errors reducing is available on a required node where the process request must be registered. + assert (F.isEmpty(errors) && F.isEmpty(results)) || locRq != null; + + Map<ClusterNode, Exception> errors0 = locRq == null || F.isEmpty(errors) + ? Collections.emptyMap() + : collectErrors(errors, locRq.nodes()); + + if (err == null && !F.isEmpty(results)) { + assert results.values().stream().noneMatch(res -> res != null && res.metas != null); + + Map<ClusterNode, Map<PartitionKeyV2, PartitionHashRecordV2>> results0 = collectPartsHashes(results, locRq.nodes()); + + IdleVerifyResultV2 chkRes = SnapshotChecker.reduceHashesResults(results0, errors0); + + finished = clusterOpFut.onDone(new SnapshotPartitionsVerifyTaskResult(locRq.metas, chkRes)); + } + else + finished = finishClusterFutureWithErr(clusterOpFut, err, errors0); + + if (finished && log.isInfoEnabled()) + log.info("Snapshot validation process finished, req: " + locRq + '.'); + }); + } + + /** */ + private Throwable stopAndCleanLocRequest(SnapshotCheckProcessRequest locRq, @Nullable Throwable err) { + if (err == null && locRq.error() != null) + err = locRq.error(); + + GridFutureAdapter<?> locWorkingFut = locRq.fut(); + + boolean finished = false; + + // Try to stop local working future ASAP. + if (locWorkingFut != null) + finished = err == null ? locWorkingFut.onDone() : locWorkingFut.onDone(err); + + requests.remove(locRq.snapshotName()); + + if (finished && log.isInfoEnabled()) + log.info("Finished snapshot local validation, req: " + locRq + '.'); + + return err; + } + + /** */ + private Map<ClusterNode, Exception> collectErrors(Map<UUID, Throwable> errors, Set<UUID> requiredNodes) { + if (errors.isEmpty()) + return Collections.emptyMap(); + + return errors.entrySet().stream().filter(e -> requiredNodes.contains(e.getKey()) && e.getValue() != null) + .collect(Collectors.toMap(e -> kctx.cluster().get().node(e.getKey()), e -> asException(e.getValue()))); + } + + /** */ + private Map<ClusterNode, Map<PartitionKeyV2, PartitionHashRecordV2>> collectPartsHashes( + @Nullable Map<UUID, CheckResultDTO> results, + Collection<UUID> requiredNodes + ) { + if (F.isEmpty(results)) + return Collections.emptyMap(); + + return results.entrySet().stream() + .filter(e -> requiredNodes.contains(e.getKey()) && e.getValue() != null) + .collect(Collectors.toMap(e -> kctx.cluster().get().node(e.getKey()), e -> e.getValue().partsHashes)); + } + + /** + * @param snpName Snapshot name of the validation process. If {@code null}, ignored. + * @param procId If {@code snpName} is {@code null}, is used to find the operation request. + * @return Current snapshot checking request by {@code snpName} or {@code procId}. + */ + private @Nullable SnapshotCheckProcessRequest currentRequest(@Nullable String snpName, UUID procId) { + return snpName == null + ? requests.values().stream().filter(rq -> rq.requestId().equals(procId)).findFirst().orElse(null) + : requests.get(snpName); + } + + /** Phase 1 beginning: prepare, collect and check local metas. */ + private IgniteInternalFuture<CheckResultDTO> prepareAndCheckMetas(SnapshotCheckProcessRequest extReq) { + SnapshotCheckProcessRequest locReq = requests.computeIfAbsent(extReq.snapshotName(), snpName -> extReq); + + if (!locReq.equals(extReq)) { + Throwable err = new IllegalStateException("Validation of snapshot '" + extReq.snapshotName() + + "' has already started. Request=" + locReq + '.'); + + clean(extReq.requestId(), err, null, null); Review Comment: Yes, we could. I failed to do it quickly. Error processing is one of the huges parts of code and tests. Also It doesn't become simplier. Why changing? Repeating other process as much as possible was my goal at all. We do lesser serialization here, exist asap. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
