New In This Release
- Performance improvements: MapR 2.1 has increased performance
- Starting in version 2.1, MapR is compatible with Hadoop 1.0.3.
- Improved CLDB/HA failover due to faster network failure discovery.
- Continuous client access throughout failover process.
- Faster CLDB failover.
- Faster NFS VIP failover.
- Moves across volume boundaries are now supported with the
hadoop fs -mvcommand.
- (Issue 5537) Fixed rare MapR application crashes due to Java JVM dumping core. This is a Java Runtime Environment issue.
- (Issue 8692) Initialization scripts now point to the correct binary on RedHat and CentOS.
- (Issue 7275) The
hadoop fs -mvcommand can now move data across volumes.
- (Issue 8237) Old FSID values are now removed correctly.
- (Issue 8170) Cache expiration no longer causes binding to return null and generate a core.
- (Issue 7909) Multihomed nodes now failover correctly with VIPs.
MCS and CLI
- (Issue 8266) The
maprcli volume removecommand properly handles filters passed with the
- (Issue 8642) The
maprcli disk removecommand now correctly handles flat files.
- (Issue 7893) A given CLDB node can now have multiple IP addresses in
- (Issue 8578) CLDB failover time decreased significantly.
- (Issue 8598) CLDB now properly sets device information when relinquishing an IP address.
- (Issue 8874) CLDB failover during rereplication no longer raises false under-replication alarms.
- (Issue 8922) CLDB no longer indefinitely and incorrectly marks some containers as spurious.
- (Issue 4981) Hive now supports lower replication settings for intermediate data.
- (Issue 8325) Diagnostics info files now generate only for failed task attempts.
- (Issue 8459) Distributed cache log messages now have information about job ID and file sizes.
In some cases, Map/Reduce preemption may not work with Map Task prefetch and Expresslane. MapR recommends using prefetch, preemption and expresslane mutually exclusively.
The bug https://issues.apache.org/jira/browse/HIVE-2907 is not yet fixed in MapR's release. It may cause the dropping of a table with a large number of partitions to fail due to an out-of-memory error.
There are no obvious indications when a MapR license expires, except the degradation of services to the non-licensed level. The workaround is to pay attention to License expiration alarms before the actual expiry.
The job tracker must be manually restarted after upgrading from any MapR version prior to 2.0 to MapR version 2.0 or later on a live cluster.
When a node's IP number is reassigned to another node, remote procedure calls are misdirected to the wrong node. This can result in widespread errors affecting any system that use the RPC binding layer, including container resync operations that never finish or the creation of several superfluous containers. Restart the cluster to clear this condition.
If you use the latest version of any browser to view the MCS, the CLDB, HBase, JobTracker, and job log pages do not load when you click on the respective links. As a workaround, use the following instructions to disable mixed-content-blocking in your browser:
Scroll to the bottom of the screen, and click Show all content. The page refreshes and displays mixed content.
- At the top of the page, to the left of the address bar, click (the shield icon).
- In the pop-up window that appears, click the down arrow next to "Keep Blocking", and select Disable Protection on This Page. The page refreshes and displays mixed content. The shield icon is replaced with a warning symbol to indicate that the page is displaying both secure and non-secure items.
- Click (the shield icon) on the right side of the address bar.
- In the icon dialog box, click Load anyway. The page refreshes and displays mixed content. The URL in the address bar shows https crossed out to indicate that the page is displaying both secure and non-secure items.
Map/Reduce and Hadoop Patches Integrated Since Last Release
Hadoop Common Patches
MapR 2.1 includes the following Apache Hadoop patches that are not included in the MapR distribution for Hadoop version 2.0:
[HADOOP-8329] HADOOP-8329. Build fails with Java 7.
[HADOOP-8430] Backport new FileSystem methods introduced by HADOOP-8014 to branch-1
[HADOOP-6546] BloomMapFile can return false negatives
[HADOOP-8151] Error handling in snappy decompressor throws invalid exceptions
[HADOOP-6642] Fix javac, javadoc, findbugs warnings
[HADOOP-7539] merge hadoop archive goodness from trunk to .20
[HADOOP-7602] wordcount, sort etc on har files fails with NPE
[HADOOP-7594] Support HTTP REST in HttpServer
[HADOOP-7661] FileSystem.getCanonicalServiceName throws NPE for any file system uri that doesn't have an authority.
[HADOOP-7649] TestMapredGroupMappingServiceRefresh and TestRefreshUserMappings fail after HADOOP-7625
[HADOOP-7215] RPC clients must connect over a network interface corresponding to the host name in the client's kerberos principal key
[HADOOP-7509] Improve message when Authentication is required
[HADOOP-8445] Token should not print the password in toString
[HADOOP-8587] HarFileSystem access of harMetaCache isn't threadsafe
[HADOOP-7836] TestSaslRPC#testDigestAuthMethodHostBasedToken fails with hostname localhost.localdomain
[HADOOP-6975] integer overflow in S3InputStream for blocks > 2GB
[HADOOP-8552] Conflict: Same security.log.file for multiple users.
[HADOOP-8612] Backport HADOOP-8599 to branch-1 (Non empty response when read beyond eof)
MapR 2.1 includes the following Apache MapReduce patches that are not included in the MapR distribution for Hadoop version 2.0:
[MAPREDUCE-336] The logging level of the tasks should be configurable by the job
[MAPREDUCE-4359] Potential deadlock in Counters
[MAPREDUCE-2452] Delegation token cancellation shouldn't hold global JobTracker lock
[MAPREDUCE-3993] Graceful handling of codec errors during decompression
[MAPREDUCE-4385] FairScheduler.maxTasksToAssign() should check for fairscheduler.assignmultiple.maps < TaskTracker.availableSlots
[MAPREDUCE-2779] JobSplitWriter.java can't handle large job.split file
[MAPREDUCE-4036] Streaming TestUlimit fails on CentOS 6
[MAPREDUCE-4355] Add RunningJob.getJobStatus()
[MAPREDUCE-4415] Backport the Job.getInstance methods from MAPREDUCE-1505 to branch-1
[MAPREDUCE-4154] streaming MR job succeeds even if the streaming command fails
[MAPREDUCE-4464] Hostnames with an underscore no longer cause reduce tasks to fail.