When you bulk load data to Couchbase Server, you can accidentally overwhelm available memory in the Couchbase cluster before it can store data on disk. If this happens, Couchbase Server will immediately send a response indicating the operation cannot be handled at the moment but can be handled later.

This is sometimes referred to as "handling Temp OOM", where where OOM means out of memory. Note though that the actual temporary failure could be sent back for reasons other than OOM. However, temporary OOM is the most common underlying cause for this error.

To handle this problem, you could perform an exponential backoff as part of your bulk load. The backoff essentially reduces the number of requests sent to Couchbase Server as it receives OOM errors:

package com.couchbase.sample.dataloader;

import com.couchbase.client.CouchbaseClient;
import java.io.IOException;
import java.net.URI;
import java.util.List;
import net.spy.memcached.internal.OperationFuture;
import net.spy.memcached.ops.OperationStatus;

/**
 *
   * The StoreHandler exists mainly to abstract the need to store things
   * to the Couchbase Cluster even in environments where we may receive
   * temporary failures.
 *
 * @author ingenthr
 */
public class StoreHandler {

  CouchbaseClient cbc;
  private final List<URI> baselist;
  private final String bucketname;
  private final String password;

  /**
   *
   * Create a new StoreHandler.  This will not be ready until it's initialized
   * with the init() call.
   *
   * @param baselist
   * @param bucketname
   * @param password
   */
  public StoreHandler(List<URI> baselist, String bucketname, String password) {
    this.baselist = baselist; // TODO: maybe copy this?
    this.bucketname = bucketname;
    this.password = password;


  }

  /**
   * Initialize this StoreHandler.
   *
   * This will build the connections for the StoreHandler and prepare it
   * for use.  Initialization is separated from creation to ensure we would
   * not throw exceptions from the constructor.
   *
   *
   * @return StoreHandler
   * @throws IOException
   */
  public StoreHandler init() throws IOException {
    // I prefer to avoid exceptions from constructors, a legacy we're kind
    // of stuck with, so wrapped here
    cbc = new CouchbaseClient(baselist, bucketname, password);
    return this;
  }

  /**
   *
   * Perform a regular, asynchronous set.
   *
   * @param key
   * @param exp
   * @param value
   * @return the OperationFuture<Boolean> that wraps this set operation
   */
  public OperationFuture<Boolean> set(String key, int exp, Object value) {
    return cbc.set(key, exp, cbc);
  }

  /**
   * Continuously try a set with exponential backoff until number of tries or
   * successful.  The exponential backoff will wait a maximum of 1 second, or
   * whatever
   *
   * @param key
   * @param exp
   * @param value
   * @param tries number of tries before giving up
   * @return the OperationFuture<Boolean> that wraps this set operation
   */
  public OperationFuture<Boolean> contSet(String key, int exp, Object value,
          int tries) {
    OperationFuture<Boolean> result = null;
    OperationStatus status;
    int backoffexp = 0;

    try {
      do {
        if (backoffexp > tries) {
          throw new RuntimeException("Could not perform a set after "
                  + tries + " tries.");
        }
        result = cbc.set(key, exp, value);
        status = result.getStatus(); // blocking call, improve if needed
        if (status.isSuccess()) {
          break;
        }
        if (backoffexp > 0) {
          double backoffMillis = Math.pow(2, backoffexp);
          backoffMillis = Math.min(1000, backoffMillis); // 1 sec max
          Thread.sleep((int) backoffMillis);
          System.err.println("Backing off, tries so far: " + backoffexp);
        }
        backoffexp++;

        if (!status.isSuccess()) {
          System.err.println("Failed with status: " + status.getMessage());
        }

      } while (status.getMessage().equals("Temporary failure"));
    } catch (InterruptedException ex) {
      System.err.println("Interrupted while trying to set.  Exception:"
              + ex.getMessage());
    }

    if (result == null) {
      throw new RuntimeException("Could not carry out operation."); // rare
    }

    // note that other failure cases fall through.  status.isSuccess() can be
    // checked for success or failure or the message can be retrieved.
    return result;
  }
}

There is also a setting you can provide at the connection-level for Couchbase Java SDK that will also help you avoid too many asynchronous requests:

List<URI> baselist = new ArrayList<URI>();
        baselist.add(new URI("http://localhost:8091/pools"));

        CouchbaseConnectionFactoryBuilder cfb = new CouchbaseConnectionFactoryBuilder();
        cfb.setOpTimeout(10000);  // wait up to 10 seconds for an operation to succeed
        cfb.setOpQueueMaxBlockTime(5000); // wait up to 5 seconds when trying to enqueue an operation

        CouchbaseClient myclient = new CouchbaseClient(cfb.buildCouchbaseConnection(baselist, "default", "default", ""));