Examples with ByteCountingOutputStream org.apache.nifi.stream.io.ByteCountingOutputStream used on opensource projects

Example 6 with ByteCountingOutputStream

use of org.apache.nifi.stream.io.ByteCountingOutputStream 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;
    }
}

Example 7 with ByteCountingOutputStream

use of org.apache.nifi.stream.io.ByteCountingOutputStream 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);
}

Example 8 with ByteCountingOutputStream

use of org.apache.nifi.stream.io.ByteCountingOutputStream in project nifi by apache.

the class StandardProcessSession method ensureNotAppending.

private void ensureNotAppending(final ContentClaim claim) throws IOException {
    if (claim == null) {
        return;
    }
    final ByteCountingOutputStream outStream = appendableStreams.remove(claim);
    if (outStream == null) {
        return;
    }
    outStream.flush();
    outStream.close();
}

Example 9 with ByteCountingOutputStream

use of org.apache.nifi.stream.io.ByteCountingOutputStream in project nifi by apache.

the class StandardProcessSession method write.

@Override
public FlowFile write(FlowFile source, final StreamCallback writer) {
    verifyTaskActive();
    source = validateRecordState(source);
    final StandardRepositoryRecord record = records.get(source);
    final ContentClaim currClaim = record.getCurrentClaim();
    long writtenToFlowFile = 0L;
    ContentClaim newClaim = null;
    try {
        newClaim = claimCache.getContentClaim();
        claimLog.debug("Creating ContentClaim {} for 'write' for {}", newClaim, source);
        ensureNotAppending(newClaim);
        if (currClaim != null) {
            claimCache.flush(currClaim.getResourceClaim());
        }
        try (final InputStream is = getInputStream(source, currClaim, record.getCurrentClaimOffset(), true);
            final InputStream limitedIn = new LimitedInputStream(is, source.getSize());
            final InputStream disableOnCloseIn = new DisableOnCloseInputStream(limitedIn);
            final ByteCountingInputStream countingIn = new ByteCountingInputStream(disableOnCloseIn, bytesRead);
            final OutputStream os = claimCache.write(newClaim);
            final OutputStream disableOnCloseOut = new DisableOnCloseOutputStream(os);
            final ByteCountingOutputStream countingOut = new ByteCountingOutputStream(disableOnCloseOut)) {
            writeRecursionSet.add(source);
            // We want to differentiate between IOExceptions thrown by the repository and IOExceptions thrown from
            // Processor code. As a result, as have the FlowFileAccessInputStream that catches IOException from the repository
            // and translates into either FlowFileAccessException or ContentNotFoundException. We keep track of any
            // ContentNotFoundException because if it is thrown, the Processor code may catch it and do something else with it
            // but in reality, if it is thrown, we want to know about it and handle it, even if the Processor code catches it.
            final FlowFileAccessInputStream ffais = new FlowFileAccessInputStream(countingIn, source, currClaim);
            final FlowFileAccessOutputStream ffaos = new FlowFileAccessOutputStream(countingOut, source);
            boolean cnfeThrown = false;
            try {
                writer.process(createTaskTerminationStream(ffais), createTaskTerminationStream(ffaos));
            } catch (final ContentNotFoundException cnfe) {
                cnfeThrown = true;
                throw cnfe;
            } finally {
                writtenToFlowFile = countingOut.getBytesWritten();
                this.bytesWritten += writtenToFlowFile;
                this.bytesRead += countingIn.getBytesRead();
                writeRecursionSet.remove(source);
                // if cnfeThrown is true, we don't need to re-thrown the Exception; it will propagate.
                if (!cnfeThrown && ffais.getContentNotFoundException() != null) {
                    throw ffais.getContentNotFoundException();
                }
            }
        }
    } catch (final ContentNotFoundException nfe) {
        destroyContent(newClaim);
        handleContentNotFound(nfe, record);
    } catch (final IOException ioe) {
        destroyContent(newClaim);
        throw new ProcessException("IOException thrown from " + connectableDescription + ": " + ioe.toString(), ioe);
    } catch (final FlowFileAccessException ffae) {
        destroyContent(newClaim);
        throw ffae;
    } catch (final Throwable t) {
        destroyContent(newClaim);
        throw t;
    }
    removeTemporaryClaim(record);
    final FlowFileRecord newFile = new StandardFlowFileRecord.Builder().fromFlowFile(record.getCurrent()).contentClaim(newClaim).contentClaimOffset(Math.max(0L, newClaim.getLength() - writtenToFlowFile)).size(writtenToFlowFile).build();
    record.setWorking(newFile);
    return newFile;
}

Example 10 with ByteCountingOutputStream

use of org.apache.nifi.stream.io.ByteCountingOutputStream in project nifi by apache.

the class StandardProcessSession method append.

@Override
public FlowFile append(FlowFile source, final OutputStreamCallback writer) {
    verifyTaskActive();
    source = validateRecordState(source);
    final StandardRepositoryRecord record = records.get(source);
    long newSize = 0L;
    // Get the current Content Claim from the record and see if we already have
    // an OutputStream that we can append to.
    final ContentClaim oldClaim = record.getCurrentClaim();
    ByteCountingOutputStream outStream = oldClaim == null ? null : appendableStreams.get(oldClaim);
    long originalByteWrittenCount = 0;
    ContentClaim newClaim = null;
    try {
        if (outStream == null) {
            claimCache.flush(oldClaim);
            try (final InputStream oldClaimIn = context.getContentRepository().read(oldClaim)) {
                newClaim = context.getContentRepository().create(context.getConnectable().isLossTolerant());
                claimLog.debug("Creating ContentClaim {} for 'append' for {}", newClaim, source);
                final OutputStream rawOutStream = context.getContentRepository().write(newClaim);
                final OutputStream bufferedOutStream = new BufferedOutputStream(rawOutStream);
                outStream = new ByteCountingOutputStream(bufferedOutStream);
                originalByteWrittenCount = 0;
                appendableStreams.put(newClaim, outStream);
                // We need to copy all of the data from the old claim to the new claim
                StreamUtils.copy(oldClaimIn, outStream);
                // wrap our OutputStreams so that the processor cannot close it
                try (final OutputStream disableOnClose = new DisableOnCloseOutputStream(outStream)) {
                    writeRecursionSet.add(source);
                    writer.process(new FlowFileAccessOutputStream(disableOnClose, source));
                } finally {
                    writeRecursionSet.remove(source);
                }
            }
        } else {
            newClaim = oldClaim;
            originalByteWrittenCount = outStream.getBytesWritten();
            // wrap our OutputStreams so that the processor cannot close it
            try (final OutputStream disableOnClose = new DisableOnCloseOutputStream(outStream);
                final OutputStream flowFileAccessOutStream = new FlowFileAccessOutputStream(disableOnClose, source)) {
                writeRecursionSet.add(source);
                writer.process(flowFileAccessOutStream);
            } finally {
                writeRecursionSet.remove(source);
            }
        }
        // update the newSize to reflect the number of bytes written
        newSize = outStream.getBytesWritten();
    } catch (final ContentNotFoundException nfe) {
        // need to reset write claim before we can remove the claim
        resetWriteClaims();
        // it here also, we would be decrementing the claimant count twice!
        if (newClaim != oldClaim) {
            destroyContent(newClaim);
        }
        handleContentNotFound(nfe, record);
    } catch (final IOException ioe) {
        // need to reset write claim before we can remove the claim
        resetWriteClaims();
        // See above explanation for why this is done only if newClaim != oldClaim
        if (newClaim != oldClaim) {
            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();
        // See above explanation for why this is done only if newClaim != oldClaim
        if (newClaim != oldClaim) {
            destroyContent(newClaim);
        }
        throw t;
    } finally {
        if (outStream != null) {
            final long bytesWrittenThisIteration = outStream.getBytesWritten() - originalByteWrittenCount;
            bytesWritten += bytesWrittenThisIteration;
        }
    }
    // the FlowFile was written to, via #write() and then append() was called.
    if (newClaim != oldClaim) {
        removeTemporaryClaim(record);
    }
    final FlowFileRecord newFile = new StandardFlowFileRecord.Builder().fromFlowFile(record.getCurrent()).contentClaim(newClaim).contentClaimOffset(0).size(newSize).build();
    record.setWorking(newFile);
    return newFile;
}