This page contains the following topics:
New In This Release
- Feature: Task Nodes (nodes that contribute only compute resources and no storage) are now supported gracefully. See Task Nodes.
- Feature: v2.1.2 on Amazon's Elastic MapReduce (EMR) cloud now supports Spot Pricing for adding and removing task nodes.
- Feature: Source code for Hadoop ecosystem components is now released to GitHub. See Source Code for MapR Software.
- Feature: Component dependencies are now defined and downloadable from a Maven repository. See Maven Repository and Artifacts for MapR.
- Feature: The platform-independent
maprfslibrary now allows developers to build and run applications without platform-specific dependencies. See Accessing MapR-FS in Java Applications.
- Feature: HiveServer2 is now included in the MapR distribution, enabling multiple concurrent client sessions with the Hive server. See Configuring Hive.
- Change: The contents of
maprfs-0.1.jarare separated into two parts,
maprfs-jni-2.1.2.jar, which enables loading on distributed class-loader environments that use the
maprfslibraries to access the MapR distribution for Hadoop from multiple contexts. For details on application impacts, see topic My application that includes maprfs-0.1.jar is now missing dependencies and fails to link.
- Change: The topology for new volumes and nodes now defaults to
/data, instead of
/default-rack, which improves usability for the future event when a node must be gracefully decommissioned. See Node Topology and Managing Data with Volumes.
- Change: Balancer status added to the MapR Control System dashboard. See Balancer Settings.
- Change: Several properties in the
mapred-site.xmlfile now enable imposing a maximum cap on the number of task slots allocated on a node.
Resolved issues are listed in the sections below.
- (Issue 6822) Whirr now runs correctly on EMR clusters.
- (Issue 8668) Storage pools no longer display with the same names.
- (Issue 8995) GIDs now correctly and consistently translate to group names.
- (Issue 9089) The
hoststatsservice no longer leaves defunct
- (Issue 9104) The
sp list -vwas added to list all disks in a storage pool.
- (Issue 9268) Disk failures are now correctly handled and generate logs.
- (Issue 9270) Disk errors updated with more detailed error messages.
- (Issue 9447) A static libMapRClient.so under the
/support/pythondirectory now provides further support for linking to this library on Linux.
- (Issue 9449) Changes to how input splits are determined for enhanced performance.
- (Issue 9623) Hadoop streaming jobs with binary mappers now work correctly.
- (Issue 9304) Nodes synchronizing in a star replication pattern no longer cause an assert in rare circumstances.
- (Issue 8853) Log files can be exempted from regular log file cleanup to allow longer retention.
- (Issue 8942) Central user log directories now clean up correctly.
- (Issue 9736) Centralized logging now disabled by default.
- (Issue 8191) Improvements to the algorithm used to balance virtual IPs to nodes in a VIP pool.
- (Issue 8990) JobTracker now fails over correctly on network disconnect.
- (Issue 9385) Added comments to warden.conf regarding usages of JT and TT memory settings
- (Issue 8385) Occasional hangs during TaskTracker restarts addressed.
- (Issue 9132) Saving Metrics configuration in the MCS no longer overwrites the existing MySQL user password.
- (Issue 8599) The
version.shscript now correctly reports the HBase version after installing a different version of HBase.
Known issues are listed in the sections below.
TaskTracker incorrectly kills task attempts and considers the attempts failed.
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.
MapReduce and Hadoop Patches Integrated Since Last Release
MapR 2.1.2 includes the following Apache MapReduce and Apache Hadoop patches that are not included in the MapR distribution for Hadoop version 2.1.1:
- [HADOOP-6835] Processing of concatenated gzip files now continues to the end of the concatenated stream, matching the behavior of the gzip(1) utility. This patch does not affect support for
- [MAPREDUCE-2340] Optimization of
JobInProgress.initTasks()for increased performance.
- [MAPREDUCE-4986] The
getAllowedLocalityLevelmethod uses preemption timeouts to disable locality delay scheduling for starved pools when
mapred.fairscheduler.preemptionis set to the default value of
- [MAPREDUCE-830] Bzip2 splitting now supports Text data.