use of org.apache.nifi.controller.repository.claim.ContentClaim in project nifi by apache.
the class StandardProcessSession method handleContentNotFound.
private void handleContentNotFound(final ContentNotFoundException nfe, final StandardRepositoryRecord suspectRecord) {
final ContentClaim registeredClaim = suspectRecord.getOriginalClaim();
final ContentClaim transientClaim = suspectRecord.getWorkingClaim();
final ContentClaim missingClaim = nfe.getMissingClaim();
final ProvenanceEventRecord dropEvent = provenanceReporter.drop(suspectRecord.getCurrent(), nfe.getMessage() == null ? "Content Not Found" : nfe.getMessage());
if (dropEvent != null) {
context.getProvenanceRepository().registerEvent(dropEvent);
}
if (missingClaim == registeredClaim) {
suspectRecord.markForAbort();
rollback();
throw new MissingFlowFileException("Unable to find content for FlowFile", nfe);
}
if (missingClaim == transientClaim) {
rollback();
throw new MissingFlowFileException("Unable to find content for FlowFile", nfe);
}
}
use of org.apache.nifi.controller.repository.claim.ContentClaim in project nifi by apache.
the class StandardProcessSession method write.
@Override
public OutputStream write(FlowFile source) {
verifyTaskActive();
source = validateRecordState(source);
final StandardRepositoryRecord record = records.get(source);
ContentClaim newClaim = null;
try {
newClaim = claimCache.getContentClaim();
claimLog.debug("Creating ContentClaim {} for 'write' for {}", newClaim, source);
ensureNotAppending(newClaim);
final OutputStream rawStream = claimCache.write(newClaim);
final OutputStream disableOnClose = new DisableOnCloseOutputStream(rawStream);
final ByteCountingOutputStream countingOut = new ByteCountingOutputStream(disableOnClose);
final FlowFile sourceFlowFile = source;
final ContentClaim updatedClaim = newClaim;
final OutputStream errorHandlingOutputStream = new OutputStream() {
private boolean closed = false;
@Override
public void write(final int b) throws IOException {
try {
countingOut.write(b);
} catch (final IOException ioe) {
LOG.error("Failed to write content to " + sourceFlowFile + "; rolling back session", ioe);
rollback(true);
close();
throw new FlowFileAccessException("Failed to write to Content Repository for " + sourceFlowFile, ioe);
}
}
@Override
public void write(final byte[] b) throws IOException {
try {
countingOut.write(b);
} catch (final IOException ioe) {
LOG.error("Failed to write content to " + sourceFlowFile + "; rolling back session", ioe);
rollback(true);
close();
throw new FlowFileAccessException("Failed to write to Content Repository for " + sourceFlowFile, ioe);
}
}
@Override
public void write(final byte[] b, final int off, final int len) throws IOException {
try {
countingOut.write(b, off, len);
} catch (final IOException ioe) {
LOG.error("Failed to write content to " + sourceFlowFile + "; rolling back session", ioe);
rollback(true);
close();
throw new FlowFileAccessException("Failed to write to Content Repository for " + sourceFlowFile, ioe);
}
}
@Override
public void flush() throws IOException {
try {
countingOut.flush();
} catch (final IOException ioe) {
LOG.error("Failed to write content to " + sourceFlowFile + "; rolling back session", ioe);
rollback(true);
close();
throw new FlowFileAccessException("Failed to write to Content Repository for " + sourceFlowFile, ioe);
}
}
@Override
public void close() throws IOException {
if (closed) {
return;
}
closed = true;
writeRecursionSet.remove(sourceFlowFile);
final long bytesWritten = countingOut.getBytesWritten();
StandardProcessSession.this.bytesWritten += bytesWritten;
final OutputStream removed = openOutputStreams.remove(sourceFlowFile);
if (removed == null) {
LOG.error("Closed Session's OutputStream but there was no entry for it in the map; sourceFlowFile={}; map={}", sourceFlowFile, openOutputStreams);
}
flush();
removeTemporaryClaim(record);
final FlowFileRecord newFile = new StandardFlowFileRecord.Builder().fromFlowFile(record.getCurrent()).contentClaim(updatedClaim).contentClaimOffset(Math.max(0, updatedClaim.getLength() - bytesWritten)).size(bytesWritten).build();
record.setWorking(newFile);
}
};
writeRecursionSet.add(source);
openOutputStreams.put(source, errorHandlingOutputStream);
return createTaskTerminationStream(errorHandlingOutputStream);
} catch (final ContentNotFoundException nfe) {
// need to reset write claim before we can remove the claim
resetWriteClaims();
destroyContent(newClaim);
handleContentNotFound(nfe, record);
throw nfe;
} catch (final FlowFileAccessException ffae) {
// need to reset write claim before we can remove the claim
resetWriteClaims();
destroyContent(newClaim);
throw ffae;
} catch (final IOException ioe) {
// need to reset write claim before we can remove the claim
resetWriteClaims();
destroyContent(newClaim);
throw new ProcessException("IOException thrown from " + connectableDescription + ": " + ioe.toString(), ioe);
} catch (final Throwable t) {
// need to reset write claim before we can remove the claim
resetWriteClaims();
destroyContent(newClaim);
throw t;
}
}
use of org.apache.nifi.controller.repository.claim.ContentClaim in project nifi by apache.
the class StandardProcessSession method enrich.
@Override
public StandardProvenanceEventRecord enrich(final ProvenanceEventRecord rawEvent, final FlowFile flowFile) {
verifyTaskActive();
final StandardRepositoryRecord repoRecord = records.get(flowFile);
if (repoRecord == null) {
throw new FlowFileHandlingException(flowFile + " is not known in this session (" + toString() + ")");
}
final StandardProvenanceEventRecord.Builder recordBuilder = new StandardProvenanceEventRecord.Builder().fromEvent(rawEvent);
if (repoRecord.getCurrent() != null && repoRecord.getCurrentClaim() != null) {
final ContentClaim currentClaim = repoRecord.getCurrentClaim();
final long currentOffset = repoRecord.getCurrentClaimOffset();
final long size = flowFile.getSize();
final ResourceClaim resourceClaim = currentClaim.getResourceClaim();
recordBuilder.setCurrentContentClaim(resourceClaim.getContainer(), resourceClaim.getSection(), resourceClaim.getId(), currentOffset + currentClaim.getOffset(), size);
}
if (repoRecord.getOriginal() != null && repoRecord.getOriginalClaim() != null) {
final ContentClaim originalClaim = repoRecord.getOriginalClaim();
final long originalOffset = repoRecord.getOriginal().getContentClaimOffset();
final long originalSize = repoRecord.getOriginal().getSize();
final ResourceClaim resourceClaim = originalClaim.getResourceClaim();
recordBuilder.setPreviousContentClaim(resourceClaim.getContainer(), resourceClaim.getSection(), resourceClaim.getId(), originalOffset + originalClaim.getOffset(), originalSize);
}
final FlowFileQueue originalQueue = repoRecord.getOriginalQueue();
if (originalQueue != null) {
recordBuilder.setSourceQueueIdentifier(originalQueue.getIdentifier());
}
recordBuilder.setAttributes(repoRecord.getOriginalAttributes(), repoRecord.getUpdatedAttributes());
return recordBuilder.build();
}
use of org.apache.nifi.controller.repository.claim.ContentClaim in project nifi by apache.
the class StandardProcessSession method migrate.
private void migrate(final StandardProcessSession newOwner, Collection<FlowFile> flowFiles) {
// We don't call validateRecordState() here because we want to allow migration of FlowFiles that have already been marked as removed or transferred, etc.
flowFiles = flowFiles.stream().map(this::getMostRecent).collect(Collectors.toList());
for (final FlowFile flowFile : flowFiles) {
if (openInputStreams.containsKey(flowFile)) {
throw new IllegalStateException(flowFile + " cannot be migrated to a new Process Session because this session currently " + "has an open InputStream for the FlowFile, created by calling ProcessSession.read(FlowFile)");
}
if (openOutputStreams.containsKey(flowFile)) {
throw new IllegalStateException(flowFile + " cannot be migrated to a new Process Session because this session currently " + "has an open OutputStream for the FlowFile, created by calling ProcessSession.write(FlowFile)");
}
if (readRecursionSet.containsKey(flowFile)) {
throw new IllegalStateException(flowFile + " already in use for an active callback or InputStream created by ProcessSession.read(FlowFile) has not been closed");
}
if (writeRecursionSet.contains(flowFile)) {
throw new IllegalStateException(flowFile + " already in use for an active callback or OutputStream created by ProcessSession.write(FlowFile) has not been closed");
}
final StandardRepositoryRecord record = records.get(flowFile);
if (record == null) {
throw new FlowFileHandlingException(flowFile + " is not known in this session (" + toString() + ")");
}
}
// If we have a FORK event for one of the given FlowFiles, then all children must also be migrated. Otherwise, we
// could have a case where we have FlowFile A transferred and eventually exiting the flow and later the 'newOwner'
// ProcessSession is committed, claiming to have created FlowFiles from the parent, which is no longer even in
// the flow. This would be very confusing when looking at the provenance for the FlowFile, so it is best to avoid this.
final Set<String> flowFileIds = flowFiles.stream().map(ff -> ff.getAttribute(CoreAttributes.UUID.key())).collect(Collectors.toSet());
for (final Map.Entry<FlowFile, ProvenanceEventBuilder> entry : forkEventBuilders.entrySet()) {
final FlowFile eventFlowFile = entry.getKey();
if (flowFiles.contains(eventFlowFile)) {
final ProvenanceEventBuilder eventBuilder = entry.getValue();
for (final String childId : eventBuilder.getChildFlowFileIds()) {
if (!flowFileIds.contains(childId)) {
throw new IllegalStateException("Cannot migrate " + eventFlowFile + " to a new session because it was forked to create " + eventBuilder.getChildFlowFileIds().size() + " children and not all children are being migrated. If any FlowFile is forked, all of its children must also be migrated at the same time as the forked FlowFile");
}
}
}
}
// event builder for the new owner of the FlowFile and remove the child from our fork event builder.
for (final Map.Entry<FlowFile, ProvenanceEventBuilder> entry : forkEventBuilders.entrySet()) {
final FlowFile eventFlowFile = entry.getKey();
final ProvenanceEventBuilder eventBuilder = entry.getValue();
final Set<String> childrenIds = new HashSet<>(eventBuilder.getChildFlowFileIds());
ProvenanceEventBuilder copy = null;
for (final FlowFile flowFile : flowFiles) {
final String flowFileId = flowFile.getAttribute(CoreAttributes.UUID.key());
if (childrenIds.contains(flowFileId)) {
eventBuilder.removeChildFlowFile(flowFile);
if (copy == null) {
copy = eventBuilder.copy();
copy.getChildFlowFileIds().clear();
}
copy.addChildFlowFile(flowFileId);
}
}
if (copy != null) {
newOwner.forkEventBuilders.put(eventFlowFile, copy);
}
}
newOwner.processingStartTime = Math.min(newOwner.processingStartTime, processingStartTime);
for (final FlowFile flowFile : flowFiles) {
final FlowFileRecord flowFileRecord = (FlowFileRecord) flowFile;
final StandardRepositoryRecord repoRecord = this.records.remove(flowFile);
newOwner.records.put(flowFileRecord, repoRecord);
// Adjust the counts for Connections for each FlowFile that was pulled from a Connection.
// We do not have to worry about accounting for 'input counts' on connections because those
// are incremented only during a checkpoint, and anything that's been checkpointed has
// also been committed above.
final FlowFileQueue inputQueue = repoRecord.getOriginalQueue();
if (inputQueue != null) {
final String connectionId = inputQueue.getIdentifier();
incrementConnectionOutputCounts(connectionId, -1, -repoRecord.getOriginal().getSize());
newOwner.incrementConnectionOutputCounts(connectionId, 1, repoRecord.getOriginal().getSize());
unacknowledgedFlowFiles.get(inputQueue).remove(flowFile);
newOwner.unacknowledgedFlowFiles.computeIfAbsent(inputQueue, queue -> new HashSet<>()).add(flowFileRecord);
flowFilesIn--;
contentSizeIn -= flowFile.getSize();
newOwner.flowFilesIn++;
newOwner.contentSizeIn += flowFile.getSize();
}
final String flowFileId = flowFile.getAttribute(CoreAttributes.UUID.key());
if (removedFlowFiles.remove(flowFileId)) {
newOwner.removedFlowFiles.add(flowFileId);
newOwner.removedCount++;
newOwner.removedBytes += flowFile.getSize();
removedCount--;
removedBytes -= flowFile.getSize();
}
if (createdFlowFiles.remove(flowFileId)) {
newOwner.createdFlowFiles.add(flowFileId);
}
if (repoRecord.getTransferRelationship() != null) {
flowFilesOut--;
contentSizeOut -= flowFile.getSize();
newOwner.flowFilesOut++;
newOwner.contentSizeOut += flowFile.getSize();
}
final List<ProvenanceEventRecord> events = generatedProvenanceEvents.remove(flowFile);
if (events != null) {
newOwner.generatedProvenanceEvents.put(flowFile, events);
}
final ContentClaim currentClaim = repoRecord.getCurrentClaim();
if (currentClaim != null) {
final ByteCountingOutputStream appendableStream = appendableStreams.remove(currentClaim);
if (appendableStream != null) {
newOwner.appendableStreams.put(currentClaim, appendableStream);
}
}
final Path toDelete = deleteOnCommit.remove(flowFile);
if (toDelete != null) {
newOwner.deleteOnCommit.put(flowFile, toDelete);
}
}
provenanceReporter.migrate(newOwner.provenanceReporter, flowFileIds);
}
use of org.apache.nifi.controller.repository.claim.ContentClaim in project nifi by apache.
the class StandardProcessSession method summarizeEvents.
private String summarizeEvents(final Checkpoint checkpoint) {
// relationship to flowfile ID's
final Map<Relationship, Set<String>> transferMap = new HashMap<>();
final Set<String> modifiedFlowFileIds = new HashSet<>();
int largestTransferSetSize = 0;
for (final Map.Entry<FlowFileRecord, StandardRepositoryRecord> entry : checkpoint.records.entrySet()) {
final FlowFile flowFile = entry.getKey();
final StandardRepositoryRecord record = entry.getValue();
final Relationship relationship = record.getTransferRelationship();
if (Relationship.SELF.equals(relationship)) {
continue;
}
Set<String> transferIds = transferMap.get(relationship);
if (transferIds == null) {
transferIds = new HashSet<>();
transferMap.put(relationship, transferIds);
}
transferIds.add(flowFile.getAttribute(CoreAttributes.UUID.key()));
largestTransferSetSize = Math.max(largestTransferSetSize, transferIds.size());
final ContentClaim workingClaim = record.getWorkingClaim();
if (workingClaim != null && workingClaim != record.getOriginalClaim() && record.getTransferRelationship() != null) {
modifiedFlowFileIds.add(flowFile.getAttribute(CoreAttributes.UUID.key()));
}
}
final int numRemoved = checkpoint.removedFlowFiles.size();
final int numModified = modifiedFlowFileIds.size();
final int numCreated = checkpoint.createdFlowFiles.size();
final StringBuilder sb = new StringBuilder(512);
if (!LOG.isDebugEnabled() && (largestTransferSetSize > VERBOSE_LOG_THRESHOLD || numModified > VERBOSE_LOG_THRESHOLD || numCreated > VERBOSE_LOG_THRESHOLD || numRemoved > VERBOSE_LOG_THRESHOLD)) {
if (numCreated > 0) {
sb.append("created ").append(numCreated).append(" FlowFiles, ");
}
if (numModified > 0) {
sb.append("modified ").append(modifiedFlowFileIds.size()).append(" FlowFiles, ");
}
if (numRemoved > 0) {
sb.append("removed ").append(numRemoved).append(" FlowFiles, ");
}
for (final Map.Entry<Relationship, Set<String>> entry : transferMap.entrySet()) {
if (entry.getKey() != null) {
sb.append("Transferred ").append(entry.getValue().size()).append(" FlowFiles");
final Relationship relationship = entry.getKey();
if (relationship != Relationship.ANONYMOUS) {
sb.append(" to '").append(relationship.getName()).append("', ");
}
}
}
} else {
if (numCreated > 0) {
sb.append("created FlowFiles ").append(checkpoint.createdFlowFiles).append(", ");
}
if (numModified > 0) {
sb.append("modified FlowFiles ").append(modifiedFlowFileIds).append(", ");
}
if (numRemoved > 0) {
sb.append("removed FlowFiles ").append(checkpoint.removedFlowFiles).append(", ");
}
for (final Map.Entry<Relationship, Set<String>> entry : transferMap.entrySet()) {
if (entry.getKey() != null) {
sb.append("Transferred FlowFiles ").append(entry.getValue());
final Relationship relationship = entry.getKey();
if (relationship != Relationship.ANONYMOUS) {
sb.append(" to '").append(relationship.getName()).append("', ");
}
}
}
}
if (sb.length() > 2 && sb.subSequence(sb.length() - 2, sb.length()).equals(", ")) {
sb.delete(sb.length() - 2, sb.length());
}
// don't add processing time if we did nothing, because we don't log the summary anyway
if (sb.length() > 0) {
final long processingNanos = checkpoint.processingTime;
sb.append(", Processing Time = ");
formatNanos(processingNanos, sb);
}
return sb.toString();
}
Aggregations