use of org.apache.accumulo.core.tabletserver.thrift.TooManyFilesException in project accumulo by apache.
the class ThriftScanner method getBatchFromServer.
public static boolean getBatchFromServer(ClientContext context, Range range, KeyExtent extent, String server, SortedMap<Key, Value> results, SortedSet<Column> fetchedColumns, List<IterInfo> serverSideIteratorList, Map<String, Map<String, String>> serverSideIteratorOptions, int size, Authorizations authorizations, boolean retry, long batchTimeOut, String classLoaderContext) throws AccumuloException, AccumuloSecurityException, NotServingTabletException {
if (server == null)
throw new AccumuloException(new IOException());
final HostAndPort parsedServer = HostAndPort.fromString(server);
try {
TInfo tinfo = Tracer.traceInfo();
TabletClientService.Client client = ThriftUtil.getTServerClient(parsedServer, context);
try {
// not reading whole rows (or stopping on row boundries) so there is no need to enable isolation below
ScanState scanState = new ScanState(context, extent.getTableId(), authorizations, range, fetchedColumns, size, serverSideIteratorList, serverSideIteratorOptions, false, Constants.SCANNER_DEFAULT_READAHEAD_THRESHOLD, null, batchTimeOut, classLoaderContext);
TabletType ttype = TabletType.type(extent);
boolean waitForWrites = !serversWaitedForWrites.get(ttype).contains(server);
InitialScan isr = client.startScan(tinfo, scanState.context.rpcCreds(), extent.toThrift(), scanState.range.toThrift(), Translator.translate(scanState.columns, Translators.CT), scanState.size, scanState.serverSideIteratorList, scanState.serverSideIteratorOptions, scanState.authorizations.getAuthorizationsBB(), waitForWrites, scanState.isolated, scanState.readaheadThreshold, null, scanState.batchTimeOut, classLoaderContext);
if (waitForWrites)
serversWaitedForWrites.get(ttype).add(server);
Key.decompress(isr.result.results);
for (TKeyValue kv : isr.result.results) results.put(new Key(kv.key), new Value(kv.value));
client.closeScan(tinfo, isr.scanID);
return isr.result.more;
} finally {
ThriftUtil.returnClient(client);
}
} catch (TApplicationException tae) {
throw new AccumuloServerException(server, tae);
} catch (TooManyFilesException e) {
log.debug("Tablet ({}) has too many files {} : {}", extent, server, e.getMessage());
} catch (ThriftSecurityException e) {
log.warn("Security Violation in scan request to {}: {}", server, e.getMessage());
throw new AccumuloSecurityException(e.user, e.code, e);
} catch (TException e) {
log.debug("Error getting transport to {}: {}", server, e.getMessage());
}
throw new AccumuloException("getBatchFromServer: failed");
}
use of org.apache.accumulo.core.tabletserver.thrift.TooManyFilesException in project accumulo by apache.
the class ThriftScanner method scan.
public static List<KeyValue> scan(ClientContext context, ScanState scanState, int timeOut) throws ScanTimedOutException, AccumuloException, AccumuloSecurityException, TableNotFoundException {
TabletLocation loc = null;
Instance instance = context.getInstance();
long startTime = System.currentTimeMillis();
String lastError = null;
String error = null;
int tooManyFilesCount = 0;
long sleepMillis = 100;
final long maxSleepTime = context.getConfiguration().getTimeInMillis(Property.GENERAL_MAX_SCANNER_RETRY_PERIOD);
List<KeyValue> results = null;
Span span = Trace.start("scan");
try {
while (results == null && !scanState.finished) {
if (Thread.currentThread().isInterrupted()) {
throw new AccumuloException("Thread interrupted");
}
if ((System.currentTimeMillis() - startTime) / 1000.0 > timeOut)
throw new ScanTimedOutException();
while (loc == null) {
long currentTime = System.currentTimeMillis();
if ((currentTime - startTime) / 1000.0 > timeOut)
throw new ScanTimedOutException();
Span locateSpan = Trace.start("scan:locateTablet");
try {
loc = TabletLocator.getLocator(context, scanState.tableId).locateTablet(context, scanState.startRow, scanState.skipStartRow, false);
if (loc == null) {
if (!Tables.exists(instance, scanState.tableId))
throw new TableDeletedException(scanState.tableId.canonicalID());
else if (Tables.getTableState(instance, scanState.tableId) == TableState.OFFLINE)
throw new TableOfflineException(instance, scanState.tableId.canonicalID());
error = "Failed to locate tablet for table : " + scanState.tableId + " row : " + scanState.startRow;
if (!error.equals(lastError))
log.debug("{}", error);
else if (log.isTraceEnabled())
log.trace("{}", error);
lastError = error;
sleepMillis = pause(sleepMillis, maxSleepTime);
} else {
// when a tablet splits we do want to continue scanning the low child
// of the split if we are already passed it
Range dataRange = loc.tablet_extent.toDataRange();
if (scanState.range.getStartKey() != null && dataRange.afterEndKey(scanState.range.getStartKey())) {
// go to the next tablet
scanState.startRow = loc.tablet_extent.getEndRow();
scanState.skipStartRow = true;
loc = null;
} else if (scanState.range.getEndKey() != null && dataRange.beforeStartKey(scanState.range.getEndKey())) {
// should not happen
throw new RuntimeException("Unexpected tablet, extent : " + loc.tablet_extent + " range : " + scanState.range + " startRow : " + scanState.startRow);
}
}
} catch (AccumuloServerException e) {
log.debug("Scan failed, server side exception : {}", e.getMessage());
throw e;
} catch (AccumuloException e) {
error = "exception from tablet loc " + e.getMessage();
if (!error.equals(lastError))
log.debug("{}", error);
else if (log.isTraceEnabled())
log.trace("{}", error);
lastError = error;
sleepMillis = pause(sleepMillis, maxSleepTime);
} finally {
locateSpan.stop();
}
}
Span scanLocation = Trace.start("scan:location");
scanLocation.data("tserver", loc.tablet_location);
try {
results = scan(loc, scanState, context);
} catch (AccumuloSecurityException e) {
Tables.clearCache(instance);
if (!Tables.exists(instance, scanState.tableId))
throw new TableDeletedException(scanState.tableId.canonicalID());
e.setTableInfo(Tables.getPrintableTableInfoFromId(instance, scanState.tableId));
throw e;
} catch (TApplicationException tae) {
throw new AccumuloServerException(loc.tablet_location, tae);
} catch (TSampleNotPresentException tsnpe) {
String message = "Table " + Tables.getPrintableTableInfoFromId(instance, scanState.tableId) + " does not have sampling configured or built";
throw new SampleNotPresentException(message, tsnpe);
} catch (NotServingTabletException e) {
error = "Scan failed, not serving tablet " + loc;
if (!error.equals(lastError))
log.debug("{}", error);
else if (log.isTraceEnabled())
log.trace("{}", error);
lastError = error;
TabletLocator.getLocator(context, scanState.tableId).invalidateCache(loc.tablet_extent);
loc = null;
// no need to try the current scan id somewhere else
scanState.scanID = null;
if (scanState.isolated)
throw new IsolationException();
sleepMillis = pause(sleepMillis, maxSleepTime);
} catch (NoSuchScanIDException e) {
error = "Scan failed, no such scan id " + scanState.scanID + " " + loc;
if (!error.equals(lastError))
log.debug("{}", error);
else if (log.isTraceEnabled())
log.trace("{}", error);
lastError = error;
if (scanState.isolated)
throw new IsolationException();
scanState.scanID = null;
} catch (TooManyFilesException e) {
error = "Tablet has too many files " + loc + " retrying...";
if (!error.equals(lastError)) {
log.debug("{}", error);
tooManyFilesCount = 0;
} else {
tooManyFilesCount++;
if (tooManyFilesCount == 300)
log.warn("{}", error);
else if (log.isTraceEnabled())
log.trace("{}", error);
}
lastError = error;
// not sure what state the scan session on the server side is
// in after this occurs, so lets be cautious and start a new
// scan session
scanState.scanID = null;
if (scanState.isolated)
throw new IsolationException();
sleepMillis = pause(sleepMillis, maxSleepTime);
} catch (TException e) {
TabletLocator.getLocator(context, scanState.tableId).invalidateCache(context.getInstance(), loc.tablet_location);
error = "Scan failed, thrift error " + e.getClass().getName() + " " + e.getMessage() + " " + loc;
if (!error.equals(lastError))
log.debug("{}", error);
else if (log.isTraceEnabled())
log.trace("{}", error);
lastError = error;
loc = null;
// do not want to continue using the same scan id, if a timeout occurred could cause a batch to be skipped
// because a thread on the server side may still be processing the timed out continue scan
scanState.scanID = null;
if (scanState.isolated)
throw new IsolationException();
sleepMillis = pause(sleepMillis, maxSleepTime);
} finally {
scanLocation.stop();
}
}
if (results != null && results.size() == 0 && scanState.finished) {
results = null;
}
return results;
} catch (InterruptedException ex) {
throw new AccumuloException(ex);
} finally {
span.stop();
}
}
Aggregations