Examples with FILES org.apache.accumulo.core.metadata.schema.TabletMetadata.ColumnType.FILES used on opensource projects

Example 1 with FILES

use of org.apache.accumulo.core.metadata.schema.TabletMetadata.ColumnType.FILES in project accumulo by apache.

the class BulkNewIT method verifyMetadata.

private void verifyMetadata(AccumuloClient client, String tableName, Map<String, Set<String>> expectedHashes) {
    Set<String> endRowsSeen = new HashSet<>();
    String id = client.tableOperations().tableIdMap().get(tableName);
    try (TabletsMetadata tablets = TabletsMetadata.builder(client).forTable(TableId.of(id)).fetch(FILES, LOADED, PREV_ROW).build()) {
        for (TabletMetadata tablet : tablets) {
            assertTrue(tablet.getLoaded().isEmpty());
            Set<String> fileHashes = tablet.getFiles().stream().map(f -> hash(f.getMetaUpdateDelete())).collect(Collectors.toSet());
            String endRow = tablet.getEndRow() == null ? "null" : tablet.getEndRow().toString();
            assertEquals(expectedHashes.get(endRow), fileHashes);
            endRowsSeen.add(endRow);
        }
        assertEquals(expectedHashes.keySet(), endRowsSeen);
    }
}

Example 2 with FILES

use of org.apache.accumulo.core.metadata.schema.TabletMetadata.ColumnType.FILES in project accumulo by apache.

the class Merge method getSizeIterator.

protected Iterator<Size> getSizeIterator(AccumuloClient client, String tablename, Text start, Text end) throws MergeException {
    // open up metadata, walk through the tablets.
    TableId tableId;
    TabletsMetadata tablets;
    try {
        ClientContext context = (ClientContext) client;
        tableId = context.getTableId(tablename);
        tablets = TabletsMetadata.builder(context).scanMetadataTable().overRange(new KeyExtent(tableId, end, start).toMetaRange()).fetch(FILES, PREV_ROW).build();
    } catch (Exception e) {
        throw new MergeException(e);
    }
    return tablets.stream().map(tm -> {
        long size = tm.getFilesMap().values().stream().mapToLong(DataFileValue::getSize).sum();
        return new Size(tm.getExtent(), size);
    }).iterator();
}

Example 3 with FILES

use of org.apache.accumulo.core.metadata.schema.TabletMetadata.ColumnType.FILES in project accumulo by apache.

the class Gatherer method getFilesGroupedByLocation.

/**
 * @param fileSelector
 *          only returns files that match this predicate
 * @return A map of the form : {@code map<tserver location, map<path, list<range>>} . The ranges
 *         associated with a file represent the tablets that use the file.
 */
private Map<String, Map<TabletFile, List<TRowRange>>> getFilesGroupedByLocation(Predicate<TabletFile> fileSelector) {
    Iterable<TabletMetadata> tmi = TabletsMetadata.builder(ctx).forTable(tableId).overlapping(startRow, endRow).fetch(FILES, LOCATION, LAST, PREV_ROW).build();
    // get a subset of files
    Map<TabletFile, List<TabletMetadata>> files = new HashMap<>();
    for (TabletMetadata tm : tmi) {
        for (TabletFile file : tm.getFiles()) {
            if (fileSelector.test(file)) {
                // TODO push this filtering to server side and possibly use batch scanner
                files.computeIfAbsent(file, s -> new ArrayList<>()).add(tm);
            }
        }
    }
    // group by location, then file
    Map<String, Map<TabletFile, List<TRowRange>>> locations = new HashMap<>();
    List<String> tservers = null;
    for (Entry<TabletFile, List<TabletMetadata>> entry : files.entrySet()) {
        String location = // filter
        entry.getValue().stream().filter(tm -> tm.getLocation() != null).map(// convert to host:port strings
        tm -> tm.getLocation().getHostPort()).min(// find minimum host:port
        String::compareTo).orElse(// if no locations,
        entry.getValue().stream().filter(tm -> tm.getLast() != null).map(// convert to host:port strings
        tm -> tm.getLast().getHostPort()).min(String::compareTo).orElse(// find minimum last location or return null
        null));
        if (location == null) {
            if (tservers == null) {
                tservers = ctx.instanceOperations().getTabletServers();
                Collections.sort(tservers);
            }
            // When no location, the approach below will consistently choose the same tserver for the
            // same file (as long as the set of tservers is stable).
            int idx = Math.abs(Hashing.murmur3_32_fixed().hashString(entry.getKey().getPathStr(), UTF_8).asInt()) % tservers.size();
            location = tservers.get(idx);
        }
        // merge contiguous ranges
        List<Range> merged = Range.mergeOverlapping(Lists.transform(entry.getValue(), tm -> tm.getExtent().toDataRange()));
        List<TRowRange> ranges = // clip
        merged.stream().map(r -> toClippedExtent(r).toThrift()).collect(Collectors.toList());
        // ranges
        // to
        // queried
        // range
        locations.computeIfAbsent(location, s -> new HashMap<>()).put(entry.getKey(), ranges);
    }
    return locations;
}