Examples with FailureContext org.apache.ignite.failure.FailureContext used on opensource projects

Example 26 with FailureContext

use of org.apache.ignite.failure.FailureContext in project ignite by apache.

the class FsyncFileWriteHandle method writeBuffer.

/**
 * @param pos Position in file to start write from. May be checked against actual position to wait previous writes
 * to complete.
 * @param buf Buffer to write to file.
 * @throws StorageException If failed.
 */
@SuppressWarnings("TooBroadScope")
private void writeBuffer(long pos, ByteBuffer buf) throws StorageException {
    boolean interrupted = false;
    lock.lock();
    try {
        assert fileIO != null : "Writing to a closed segment.";
        checkNode();
        long lastLogged = U.currentTimeMillis();
        long logBackoff = 2_000;
        // If we were too fast, need to wait previous writes to complete.
        while (written != pos) {
            // No one can write further than we are now.
            assert written < pos : "written = " + written + ", pos = " + pos;
            // Permutation occurred between blocks write operations.
            // Order of acquiring lock is not the same as order of write.
            long now = U.currentTimeMillis();
            if (now - lastLogged >= logBackoff) {
                if (logBackoff < 60 * 60_000)
                    logBackoff *= 2;
                U.warn(log, "Still waiting for a concurrent write to complete [written=" + written + ", pos=" + pos + ", lastFsyncPos=" + lastFsyncPos + ", stop=" + stop.get() + ", actualPos=" + safePosition() + ']');
                lastLogged = now;
            }
            try {
                writeComplete.await(2, TimeUnit.SECONDS);
            } catch (InterruptedException ignore) {
                interrupted = true;
            }
            checkNode();
        }
        // Do the write.
        int size = buf.remaining();
        assert size > 0 : size;
        try {
            assert written == fileIO.position();
            fileIO.writeFully(buf);
            written += size;
            metrics.onWalBytesWritten(size);
            assert written == fileIO.position();
        } catch (IOException e) {
            StorageException se = new StorageException("Unable to write", e);
            cctx.kernalContext().failure().process(new FailureContext(CRITICAL_ERROR, se));
            throw se;
        }
    } finally {
        writeComplete.signalAll();
        lock.unlock();
        if (interrupted)
            Thread.currentThread().interrupt();
    }
}

Example 27 with FailureContext

use of org.apache.ignite.failure.FailureContext in project ignite by apache.

the class GridJobWorker method execute0.

/**
 * @param skipNtf {@code True} to skip job processor {@code onUnheld()}
 *      notification (only from {@link #body()}).
 */
private void execute0(boolean skipNtf) {
    // Make sure flag is not set for current thread.
    HOLD.set(false);
    SqlFieldsQuery.setThreadedQueryInitiatorId("task:" + ses.getTaskName() + ":" + getJobId());
    try (OperationSecurityContext ignored = ctx.security().withContext(secCtx)) {
        if (partsReservation != null) {
            try {
                if (!partsReservation.reserve()) {
                    finishJob(null, null, true, true);
                    return;
                }
            } catch (Exception e) {
                IgniteException ex = new IgniteException("Failed to lock partitions " + "[jobId=" + ses.getJobId() + ", ses=" + ses + ']', e);
                U.error(log, "Failed to lock partitions [jobId=" + ses.getJobId() + ", ses=" + ses + ']', e);
                ;
                finishJob(null, ex, true);
                return;
            }
        }
        if (isCancelled())
            // If job was cancelled prior to assigning runner to it?
            super.cancel();
        if (!skipNtf) {
            if (holdLsnr.onUnheld(this)) {
                if (held.decrementAndGet() == 0)
                    status = RUNNING;
            } else {
                if (log.isDebugEnabled())
                    log.debug("Ignoring job execution (job was not held).");
                return;
            }
        }
        boolean sndRes = true;
        Object res = null;
        IgniteException ex = null;
        try {
            ctx.job().currentTaskSession(ses);
            if (reqTopVer != null)
                GridQueryProcessor.setRequestAffinityTopologyVersion(reqTopVer);
            // avoid computation altogether.
            if (isTimedOut())
                sndRes = false;
            else {
                res = U.wrapThreadLoader(dep.classLoader(), new Callable<Object>() {

                    @Nullable
                    @Override
                    public Object call() {
                        try {
                            if (internal && ctx.config().isPeerClassLoadingEnabled())
                                ctx.job().internal(true);
                            return job.execute();
                        } finally {
                            if (internal && ctx.config().isPeerClassLoadingEnabled())
                                ctx.job().internal(false);
                        }
                    }
                });
                if (log.isDebugEnabled()) {
                    log.debug(S.toString("Job execution has successfully finished", "job", job, false, "res", res, true));
                }
            }
        } catch (IgniteException e) {
            if (sysStopping && e.hasCause(IgniteInterruptedCheckedException.class, InterruptedException.class)) {
                ex = handleThrowable(e);
                assert ex != null;
            } else {
                if (X.hasCause(e, InterruptedException.class)) {
                    if (log.isDebugEnabled()) {
                        U.error(log, "Job was cancelled [jobId=" + ses.getJobId() + ", ses=" + ses + ']', e);
                    }
                } else if (X.hasCause(e, GridServiceNotFoundException.class) || X.hasCause(e, ClusterTopologyCheckedException.class)) {
                    if (log.isDebugEnabled()) {
                        // Should be throttled, because GridServiceProxy continuously retry getting service.
                        LT.error(log, e, "Failed to execute job [jobId=" + ses.getJobId() + ", ses=" + ses + ']');
                    }
                } else {
                    String msg = "Failed to execute job [jobId=" + ses.getJobId() + ", ses=" + ses + ']';
                    if (X.hasCause(e, OutOfMemoryError.class)) {
                        U.error(log, msg, e);
                        ctx.failure().process(new FailureContext(FailureType.CRITICAL_ERROR, e));
                    } else if (log.isDebugEnabled())
                        U.error(log, msg, e);
                }
                ex = e;
            }
        }// InterruptedException if job is being cancelled.
         catch (Throwable e) {
            ex = handleThrowable(e);
            assert ex != null;
            if (e instanceof Error)
                throw (Error) e;
        } finally {
            // Finish here only if not held by this thread.
            if (!HOLD.get())
                finishJob(res, ex, sndRes);
            else
                // Make sure flag is not set for current thread.
                // This may happen in case of nested internal task call with continuation.
                HOLD.set(false);
            ctx.job().currentTaskSession(null);
            if (reqTopVer != null)
                GridQueryProcessor.setRequestAffinityTopologyVersion(null);
        }
    } finally {
        SqlFieldsQuery.resetThreadedQueryInitiatorId();
        if (partsReservation != null)
            partsReservation.release();
    }
}

Example 28 with FailureContext

use of org.apache.ignite.failure.FailureContext in project ignite by apache.

the class FilePageStoreManager method storeCacheData.

/**
 * {@inheritDoc}
 */
@Override
public void storeCacheData(StoredCacheData cacheData, boolean overwrite) throws IgniteCheckedException {
    CacheConfiguration<?, ?> ccfg = cacheData.config();
    File cacheWorkDir = cacheWorkDir(ccfg);
    checkAndInitCacheWorkDir(cacheWorkDir);
    assert cacheWorkDir.exists() : "Work directory does not exist: " + cacheWorkDir;
    File file = cacheConfigurationFile(ccfg);
    Path filePath = file.toPath();
    chgLock.readLock().lock();
    try {
        if (overwrite || !Files.exists(filePath) || Files.size(filePath) == 0) {
            File tmp = new File(file.getParent(), file.getName() + TMP_SUFFIX);
            if (tmp.exists() && !tmp.delete()) {
                log.warning("Failed to delete temporary cache config file" + "(make sure Ignite process has enough rights):" + file.getName());
            }
            // Pre-existing file will be truncated upon stream open.
            try (OutputStream stream = new BufferedOutputStream(new FileOutputStream(tmp))) {
                marshaller.marshal(cacheData, stream);
            }
            if (Files.exists(filePath) && Files.size(filePath) > 0) {
                for (BiConsumer<String, File> lsnr : lsnrs) lsnr.accept(ccfg.getName(), file);
            }
            Files.move(tmp.toPath(), file.toPath(), StandardCopyOption.REPLACE_EXISTING);
        }
    } catch (IOException ex) {
        cctx.kernalContext().failure().process(new FailureContext(FailureType.CRITICAL_ERROR, ex));
        throw new IgniteCheckedException("Failed to persist cache configuration: " + ccfg.getName(), ex);
    } finally {
        chgLock.readLock().unlock();
    }
}

Example 29 with FailureContext

use of org.apache.ignite.failure.FailureContext in project ignite by apache.

the class FilePageStoreManager method initDir.

/**
 * @param cacheWorkDir Work directory.
 * @param grpId Group ID.
 * @param partitions Number of partitions.
 * @param pageMetrics Page metrics.
 * @param encrypted {@code True} if this cache encrypted.
 * @return Cache store holder.
 * @throws IgniteCheckedException If failed.
 */
private CacheStoreHolder initDir(File cacheWorkDir, int grpId, int partitions, PageMetrics pageMetrics, boolean encrypted) throws IgniteCheckedException {
    try {
        boolean dirExisted = checkAndInitCacheWorkDir(cacheWorkDir);
        if (dirExisted) {
            MaintenanceRegistry mntcReg = cctx.kernalContext().maintenanceRegistry();
            if (!mntcReg.isMaintenanceMode())
                DefragmentationFileUtils.beforeInitPageStores(cacheWorkDir, log);
        }
        File idxFile = new File(cacheWorkDir, INDEX_FILE_NAME);
        if (dirExisted && !idxFile.exists())
            grpsWithoutIdx.add(grpId);
        FileVersionCheckingFactory pageStoreFactory = getPageStoreFactory(grpId, encrypted);
        PageStore idxStore = pageStoreFactory.createPageStore(PageStore.TYPE_IDX, idxFile, pageMetrics.totalPages()::add);
        PageStore[] partStores = new PageStore[partitions];
        for (int partId = 0; partId < partStores.length; partId++) {
            final int p = partId;
            PageStore partStore = pageStoreFactory.createPageStore(PageStore.TYPE_DATA, () -> getPartitionFilePath(cacheWorkDir, p), pageMetrics.totalPages()::add);
            partStores[partId] = partStore;
        }
        return new CacheStoreHolder(idxStore, partStores);
    } catch (IgniteCheckedException e) {
        if (X.hasCause(e, StorageException.class, IOException.class))
            cctx.kernalContext().failure().process(new FailureContext(FailureType.CRITICAL_ERROR, e));
        throw e;
    }
}

Example 30 with FailureContext

use of org.apache.ignite.failure.FailureContext in project ignite by apache.

the class InlineIndexImpl method remove.

/**
 */
private void remove(CacheDataRow row) throws IgniteCheckedException {
    try {
        int segment = segmentForRow(row);
        IndexRowImpl idxRow = new IndexRowImpl(rowHnd, row);
        idxRow.prepareCache();
        segments[segment].removex(idxRow);
    } catch (Throwable t) {
        cctx.kernalContext().failure().process(new FailureContext(CRITICAL_ERROR, t));
        throw t;
    }
}