Resolved Issues (MapR 6.0.0)
The following MapR issues, which were reported by customers, are resolved in Version 6.0.
| Component | Number | Description | Resolution |
| CLDB | 14105 | Log a proper error message when nodes attempt to the register with duplicate IP addresses | |
| CLDB | 22044 | cldb.log is not rotated properly | |
| CLDB | 24413 | cldb servers keep crashing with [java.lang.ArrayIndexOutOfBoundsException: 4 at com.mapr.fs.cldb.topology.Topology.addServerBackToRRList(Topology.java:5857) ] | |
| CLDB | 24846 | topology gets lost after cldbfaileover | |
| CLDB | 25590 | Wrong fileserver port information served by slave cldb | |
| CLDB | 25800 | Unable to restore mapr.cldb.internal from dump | |
| CLDB | 28633 | Immediately after a master CLDB failover, users were able to remove a node which had not yet registered with the new master CLDB. | With this fix, node cannot be removed for up to 15 minutes after a master CLDB failover. |
| CLDB | 29588 | CLDB deleted wrong CID 1 causing CLDB volume to be wiped out | |
| FileClient | 22571 | To use the MapR Client on Windows, it was necessary to set up the spoof user
username, uid, and gid in the core-site.xml file because these
properties were needed for translating cluster uid back to username (for job
submission). |
With this fix, you need not set up the spoof user username, uid, and gid in the
core-site.xml file as username is available via the
ticket. |
| FileClient | 23880 | When multiple applications attempt to load the same MapR native library using the root class loader, an exception was returned because MapR native library can be loaded only once and an application had already successfully loaded the MapR native library. | With this fix, an exception is not returned when applications try to load an already loaded MapR native library. |
| FileClient | 24140 | On Windows, MapRClient generates core for Python APIs | |
| FileClient | 24247 | FSDataInputStream the number of bytes read, possibly zero, or -1 if reach end-of-stream not met | |
| FileClient | 24249 | ERROR Client fs/client/fileclient/cc/client.cc:435 Thread: 7321 MapClientInitParams: Unable to get setSeperateAuditPathfieldID from Java | |
| FileClient | 24965 | Bug in FileClient initialization | |
| FileClient | 25198 | Insert query with union all failed by InvalidInputException on hive-1.0 and Tez | |
| FileClient | 25471 | Failed to fetch attributes for volume | |
| FileClient | 26790 | FUSE: CP command fails with ESTALE error | |
| FileClient | 27757 | YARN Distributed shell application fails after RM failover | |
| FileClient | 28440 | InputStream.read() API returns invalid data | |
| FileClient | 28528 | FsAction.READ_EXECUTE on directory result in AccessControlException | |
| FileClient | 29121 | InputStream.read() API returns invalid data | |
| FileServer | 12856 | Improve behavior of "hadoop fs -rmr", do not build entire file list in memory, unlink files as their meta info is returned by readdir | |
| FileServer | 23530 | SPs went offline on 2 nodes at same time and 2 containers stuck in BM state | |
| FileServer | 24232 | Unable to access or remove directory with stale files | |
| FileServer | 24779 | Assert in mapr::fs::Replication::FreeReplica | |
| FileServer | 24805 | Summary: Reduce Lock contention in SetAttr by using Fid lock in shared mode when possible. | |
| FileServer | 24884 | Stale container cache in mfs can potentially corrupt data | |
| FileServer | 25848 | Error In GetXattr for Fid x.x.x error=2 while accessing a file | |
| FileServer | 26025 | Security::Decrypt() returning EINVAL, causes MFS to assert in ReplicateOps::OpRespond() | |
| FileServer | 26093 | mfs assert in unlink after promoting destination mirror to RW | |
| FileServer | 26288 | MFS went to hung state looping in mapr::fs::CidCache::LookupServerBinding(unsigned long) () | |
| FileServer | 26351 | SSD Trimming is not supported by all vendors | |
| FileServer | 26570 | A successful gethostbyname() call returns h_addr_list as NULL causing MFS to crash | |
| FileServer | 28794 | punch hole is slow as it is doing log flush un-necessarily | |
| FS:Audit | 24585 | Extra logging in cldbaudit.log.json | |
| FS:Audit | 24915 | expandaudit command creates huge error files | |
| FS:Cache | 28775 | mfs write take long time due to uncache | |
| FS:Container | 23876 | Container was in resync for 12 hours | |
| FUSE-based POSIX Client | 25290 | Fuse process crashes frequently | |
| FUSE-based POSIX Client | 28907 | The MapR FUSE-based POSIX client required user/group to exist before allowing creation of files in MapR-FS with those UIDs/GIDs. | With this fix, the MapR FUSE-based POSIX allows file creation in MapR-FS with UIDs/GIDs that may not exist on the cluster or the client host. |
| FUSE-based POSIX Client | 28910 | When enabling the setgid bit on a file, if the the primary group set in the executing process did not match the actual group of the file, then POSIX-compliant file systems proceeded without throwing an error. | With this fix, the file system now gives EPERM error when setting the setgid bit on file, if the primary group of the executing process does not match the actual group of the file. |
| FUSE-based POSIX Client | 28914 | Unlink operation sometimes left the file in inconsistent state when accessed using the FUSE-based POSIX client. | With this fix, the unlink operation will succeed and the file cannot be accessed. |
| FS:Hardlinks | 24238 | Multi-mfs crashes at mapr::fs::Container::IAllocFromAg | |
| FS:Mirror2RW | 20644 | Mirror schedule stop with inuse flag 0 | |
| FS:Mirror2RW | 25308 | mfs asserts in GetChildAttrInodeGotten() | |
| FS:Replication | 25147 | Mfs detected deadlock condition during container replication and result to server restart | |
| FS:Replication | 25443 | Container in resync mode forever; corresponding resync on upstream is missing | |
| FS:Replication | 26054 | Container stuck in resync state without any progress with single valid copy | |
| hoststats | 11349 | hoststats fails to start on one node | |
| JM:UI | 24700 | JobTracker UI fails with NullPointerException for hive parameter | |
| DB | 24054 | MapR-DB performance was significantly degraded after it became 4 nodes cluster | |
| DB | 25241 | MapR-DB: FuzzRowFilter is not returning correct result | |
| DB | 25485 | Scanning MapR-DB table fails with Input/Output error due to corrupt column family data | |
| DB | 26416 | Assert during free in mapr::fs::LDBBlock::FetchBlockCheckCrcAndCleanup | |
| DB | 28023 | Segmentation fault in SegmentScanner::Init | |
| DB | 26338 | QueryCondition using in conditional does not include last value in provided list | |
| MapReduce | 24135 | Limits class does not construct Configuration correctly, enforces wrong counter limit, causes TaskTracker failures | |
| MapReduce | 24235 | deadlock at com.mapr.log4j.CentralTaskLogAppender | |
| MapReduce | 25373 | Jobs failing after upgrade to 5.2 due to NativeIO.POSIX.open not supported on Windows in standalone mode | |
| MapReduce | 25599 | Warden might start JobTracker twice | |
| NFS | 23698 | Change of NFS Export Path | |
| NFS | 24315 | dd command reads more data than the actual data when iflag=direct is used | |
| NFS | 24434 | rsync is not working properly with hardlinks | |
| NFS | 24630 | full time stamp shows negative numbers after applying patch 39544 | |
| NFS | 28016 | NFS keeps trying to access dead file server | |
| NFS:LoopBack | 23652 | MapR client (POSIX, FUSE, native) does not automatically refresh service ticket | |
| Security | 27152 | MapR-FS crashed because the number of bytes to encrypt exceeded 256MB. | This number of bytes to encrypt can now be greater than 256MB. |
| Tez | 20965 | Insert query with union all failed by ArrayIndexOutOfBoundsException on hive-1.0 and Tez | |
| UI:Browser | 23257 | VIP created using "Select the desired network interfaces" does not appear in NFS Setup in MCS | |
| UI:CLI | 24280 | dashboard info failed with "TimeoutException during RM Status thread execution" | |
| Warden | 24119 | Warden calculation of MFS,nodemanager memory consumption is incorrect | |
| Warden | 24562 | Warden is not being nicer to CLDB in latest releases | |
| YARN | 23571 | Spark Web UI failed with ERROR 500 when Zero configuration RM HA enabled | |
| YARN | 24392 | mapr 5.2 RM not coming up | |
| YARN | 24477 | LocalVolumeAuxService instance continues to work if init failed | |
| YARN | 24812 | Hive couldn't find finished job status | |
| YARN | 25001 | FairScheduler: potential livelock due to maxAMShare limitation and container reservation | |
| YARN | 25175 | Resource Manager fails to load state with InvalidProtocolBufferException | Property "yarn.resourcemanager.fs.state-store.remove-corrupted-data" works as
expected:
|
| YARN | 25359 | RM crashes due to inability of maprfs to remove application's directory from RM state store | |
| YARN | 25387 | NullPointerException in Error in handling event type NODE_ADDED to the scheduler | |
| YARN | 26425 | ResourceManager goes loops infinitely in the FairScheduler.preemptResources method while holding FairScheduler lock, causing all RPC threads blocked by lock contention | |
| YARN | 27919 | Memory Leak in org.apache.hadoop.fs.FileSystem$Cache | |
| YARN | 28366 | FairScheduler.preemptMapping leak |