Lists the issues that were resolved in HPE Ezmeral Data Fabric
release 6.2.0.
The following issues, which were reported by customers, are resolved in release 6.2.0.
Control System Issues
- MAPRDB-2287
- The first letter of each field name was missing when viewing field permissions.
- MON-1231
- Pagination functionality was not working as expected when changes to the number of
rows are made in pages other than page 1.
- MON-1968
- Unable to edit user email property in Volume User Disk Usage page.
- MON-2018
- The
Add Field option was disabled on the Add Change Log page.
- MON-2076
- Removing the last stream topic from the Remove Topic window list caused the Remove
Topic window to be closed, which was not the expected behaviour.
- MON-2321
- On the Service List page, the node pop-over did not display the list of nodes if
Elasticsearch had failed.
- MON-2374
- The Node Details page did not display logs for nodes that have two IPs.
- MON-2409
- The Control System did not display any feedback about page loading when navigating
from one cluster to another.
- MON-3404
- The Search field for a stream topic was not displayed properly on the Firefox browser
(Firefox version 58).
- MON-3438
- Page response was slow between clicks on the Alarms page.
- MON-3452
- Volumes Set Quota action cleared the system volumes but kept their checkboxes selected.
- MON-3476
- Users with
admin permission were not able to modify table ACLs.
- MON-3709
- The Service tab did not auto-refresh as per the refresh configuration settings.
- MON-3817
- The Topology view on the Nodes page did not refresh until one switched to another tab
and back.
- MON-3922
- Volumes prior to MapR 6.0 that lack volume ACEs were not displayed after being
upgraded to version 6.x.
- MON-4701
- Data was not returned for percentiles in the activity grid if the end time of the
query was in the future.
- MON-4776
- Oops error was displayed when mounting/unmounting multiple volumes that were selected
across volume grid pages.
- MON-4845
- Could not start the
apiserver as a user other than the
mapr user.
- MON-4862
- The Control System did not start after an upgrade to MapR 6.x if the installed license
did not have table support enabled.
- MON-4874
-
When the MAPR_EXTERNAL environment variable was set, the Control System was not
accessible and returned the following error: HTTP ERROR 503.
- MON-4892
-
The Snapshot tab on the Control System indicated that a license upgrade is needed
after upgrading from MapR 5.2.1 to MapR 6.1 even with an M5 enterprise license
(license permits all operations except tables/streams handling) already installed.
- MON-5011
- The Control System reported an incorrect value for CPU utilization when Drill queries
were executed on the cluster.
- MON-5057
- Table properties could not be modified from the Control System due to insufficient
permissions.
- MON-5075
- Was unable to modify existing volume quota using the Control System UI due to
insufficient permissions.
- MON-5116
- Fixed multiple Control System issues pointed out by a customer.
- MON-5139
-
Accessing the REST endpoints caused apiserver to freeze after a
certain number of requests.
- MON-5194
- The CPU utilization graph on the Control System had an incorrect scale on the
Y-axis.
- MON-5201
- The Control System failed to display Drill queries that were completed within the last
10 minutes.
- MON-5401
- The Control System crashed on one of the nodes.
- MON-5275
- The Control System failed to display MAC addresses on a multi-NIC environment.
- MON-5412
- The Control System failed to display the
resourcemanager service
under the MEP services list due to an unsupported shell command.
- MON-5774
-
The apiserver crashed when tmp directories were
mounted with the noexec flag.
- MON-5488
- An IO exception occurred when loading metadata of huge tables in the Control
System.
HPE Ezmeral Data Fabric Database Issues
- MAPRDB-1520
- OR conditions on array field path in correlation tracker were not handled
correctly.
- MAPRDB-1985
- Fixed a segmentation fault on the mapr dbshell
utility.
- MAPRDB-2062
- Failed to scan table on a remote secure cluster using the mapr dbshell utility because of a wrong ticket that was sent to
ZooKeeper.
- MAPRDB-2092
- Index addition fails if the write ACE of the table volume does not have the allow
access permission for MAPR_USER.
- MAPRDB-2096
- Enhanced the performance of HPE Ezmeral Data Fabric Database for
workloads with a large number of threads.
- MAPRDB-2113
- Fixed the database logic to select the index appropriate to the query being executed,
when there are multiple indexes over the same field in a table.
- MAPRDB-2125
- OJAI APIs failed to connect to ZooKeeper after applying the 6.1-EBF patch.
- MAPRDB-2156
- When running queries with set timeout, the number of threads on the HPE Ezmeral Data Fabric client increased exponentially and caused the
client to stop responding and become completely blocked.
- MAPRDB-2201
- Fixed a memory leak caused by a dangling reference of
MetaTable in
the BaseJson table.
- MAPRDB-2250
- Added a throttle for internal DB operations to prevent the cluster from becoming
unstable.
- MAPRDB-2303
- Added actual replication lag to replication/index lag alarm
- MAPRDB-2305
- The table edit API is missing the securitypolicy field.
- MAPRDB-2315
- Spark queries failed with out of memory error.
- MAPRDB-2322
- Added the
-showcol true option to column fields list. This helps the
Control System show all column fields for a given table.
- MAPRDB-2323
- The VOLUME_ALARM_TABLE_REPL_ERROR alarm was missing the details of the actual bucket
FID that produced the alarm, and the error code and description of the replication
error.
- MAPRDB-2328
- mapr copytable failed with ENOENT during create
of destination table, when source table has high number of tablets.
- MAPRDB-2345
- Tagging a column with security policies duplicated that column.
- MAPRDB-2350
- Fixed ACE permissions for MAPR_USER to allow CDC (Change Replica) to work.
Filesystem Issues
- COLD-184
collectd process constantly crashed when using the latest
collectd package.- CORE-387
- HPE Ezmeral Data Fabric user ticket got overwritten when Kerberos
authentication is enabled.
- CORE-472
- Fixed an issue where the
disklist.sh script was slow on nodes with a
high number of udev symlinks.
- CORE-476
- Fixed the empty MAPR_JMXAUTH value in the
env_override.sh script that
prevented NodeManager from starting on custom secure clusters.
- CORE-480
- When moving a node, the CLDB did not update the rackpath for volumes that have their
localpath set.
- CORE-566
- Fixed hoststats crash.
- CORE-579
rpm -U failed on HPE Ezmeral Data Fabric
patches.- CORE-611
- The maprcli disk list command reported incorrect
Storage Pool (SP) information.
- DFDEVOPS-1026
- Hadoop commands failed on 6.2EBF with the
mapr-patch-client package,
in client-only node.
- MFS-1984
- The maprcli dashboard info command
returned incorrect statistics about compressed and uncompressed data.
- MFS-2019
- The
apiserver crashed intermittently when there were multiple
networks in HPE Ezmeral Data Fabric subnets.
- MFS-2055
- CLDB crashed while processing alarms due to an error when opening one of the
clearedAlarm tables in kvstore.
- MFS-2079
- All update operations for the main volume Name container failed due to incorrect disk
permissions.
- MFS-2143
- Excluded audit data operations were not preserved on MFS restart.
- MFS-2209
- Fixed a Java exception in the
getVolName() function.
- MFS-2211
- Primary container instance was frozen during resync of orphan entries.
- MFS-2266
-
Remote wire-level secure Read RPC of unaligned compressed data failed.
- MFS-2294
- Fixed a CLDB exception that occurred when adding NFS version 4 nodes.
- MFS-2306
- Fixed a CLDB exception that occurred when nodes were removed.
- MFS-2307
- Added a cluster level flag to not offline SP when Read CRC error is encountered.
- MFS-2392
- gfsck failed on a secure cluster due to a missing
jackson-core jar file.
- MFS-2444
- FUSE failed to remove shared memory segments when running on a Kubernetes
cluster.
- MFS-2462
- Fixed DBworker thread crashes when looking up ACEs at the column level.
- MFS-2553
- Ecosystem jobs using ZooKeeper failed with the
java.lang.NoSuchMethod
error after applying the 6.1-EBF patch.
- MFS-2608
- Priorities of child processes were not in sync with that of the parent process when
the priority of the parent process is changed with the
renice Linux
command.
- MFS-2610
- Fixed hangs with Persistent Volume Mounts when the HPE Ezmeral Data Fabric ticket expired.
- MFS-2623
- Resolved a CLDB deadlock between two threads in the
ListSorterPurgeTask resource.
- MFS-2631
- Fixed CLDB shut down with
ConcurrentModificationException when
accessing Server.nfsIds.
- MFS-2638
- Store alarms sorted by alarm type in CLDB, to avoid sorting each time the maprcli alarm list -sortby alarmtype command is
used.
- MFS-2691
- Optimized HashTable walk for enhanced performance when fetching Muted and
RaisedAlarms.
- MFS-2695
- Cross cluster mirror failed with an Operation not Supported error after
enabling the snapshot lite feature.
- MFS-2700
- The FUSE kernel sent the wrong credentials to the HPE Ezmeral Data Fabric FUSE Process.
- MFS-2711
- Freed up CLDB CPU from background activity. Now expired snapshots are removed only
after the cluster is stable.
- MFS-2720
- Added a maprcli dump command to print
in-progress container replication.
- MFS-2732
- RPC connections between MFS and CLDB were failing.
- MFS-2757
- HPE Ezmeral Data Fabric services displayed incorrect services
when
systemd was updated.
- MFS-2767
- FileClient retried the same CLDB node on failure.
- MFS-2968
- Inactive alarms returned an incorrect alarm value.
- MFS-3291
- The number of flusher threads for HPE Ezmeral Data Fabric Streams
was not configurable. The default value of 64 threads was insufficient.
- MFS-4480
- Fixed intermittent NFS version 4 server crashes.
- MFS-4531
- Mirror source snapshots were not deleted after mirroring completed.
- MFS-4532
- Jobs failed with IO error as MFS processes could not establish connections with each
other, resulting in data remaining under-replicated for long periods of time until
manual intervention was done.
- MFS-4597
- Dependent services failed to start on a Warden restart.
- MFS-4562
- The
stat command on FUSE mounted files reported an incorrect IO block
size.
- MFS-4576
- Added log rotation support for the hoststats.err file.
- MFS-4605
- CLDB crashed when ACLs larger than 2K bytes were set.
- MFS-4670
- The CLDB process consumed massive amounts of memory at periodic intervals and caused
CLDB to fail.
- MFS-4752
- Fixed assert in
RpcBinding::GetCredentials.
- MFS-4776
- FUSE RPCs failed when a node's hostid is changed.
- MFS-4805
- Implemented a memory tracker for the NFS version 3 server.
- MFS-5226
- Tiered volume metadata path is trimmed down to 64 characters in the
mastgateway.log file.
- MFS-5356
- The
getAces() method returned a Null Pointer Exception when called on
a non-existent object.
- MFS-5390
- Added the filefilter feature to support blocking of
certain file extensions.
- MFS-5422
- The
create() method was creating files with the wrong
permissions.
- MFS-5430
- NFS server could not parse export files with line lengths greater than 8192
characters.
- MFS-5482
- Fixed memory leak in CLDB.
- MFS-5488
- Fixed intermittent failing of FileClient.
- MFS-5502
- PathWalk error occurred when a client node is upgraded but CLDB is of an older
version.
- MFS-5710
- Added log rotation for the
mfs.err and the mfs.out
log files to avoid filling local OS partition.
- MFS-5724
-
The MFS configuration parameter mfs.max.restore.count was not being
honored causing mirror resync operations to be delayed due to the lack of sufficient
restore slots.
- MFS-5732
- Change the severity of log messages so as to not unduly alarm customers.
- MFS-6666
- Implement a throttle to prevent the NFS server from overwhelming CLDB with too many
RPCs.
- MFS-6717
- Added lazy unmount option to allow NFS mounts to be unmounted when they are not
busy.
- MFS-6748
- Automatic Offload failed for EC Volumes.
- MFS-6785
- The mrconfig info containers rw command
was too slow to respond on a cluster with a large number of volumes.
- MFS-6873
- File operations across clusters failed when using FUSE.
- MFS-7126
- The maprcli disk remove command timed out on a
cluster with a large number of volumes.
- MFS-8454
-
Volume creation failed with the error: Could not get status of mount path.
- MFS-8459
-
Fixed volume access problems for volumes that reused the volume ID of deleted
volumes.
- MFS-8475
- The
createsystemvolumes.sh script took hours to complete when adding
a new node to a cluster with a large number of volumes.
- MFS-8482
- Tiering tasks are left orphaned when the cluster is in a semi-upgraded state.
- MFS-10328
- The maprlogin renew command failed to
refresh updated group membership details.
- MFS-10360
- Multiple MFS instances are being reset to 1 when CLDB is restarted.
- MFS-10813
- MASTGateway core crashes when gfsck is triggered on a
tiered EC volume with more than one parity nodes down.
- MFS-10845
- The maprcli volume create command does not
honor the impersonated user while creating the parent directory.
- MFS-11109
- Core dumps were generated frequently due to a memory leak issue.
- MFS-11221
- Fixed Drill query crashes.
- MFS-11295
- Offloading fails as MASTGateway is stuck in the compaction phase.
- MFS-11322
- MASTGateway truncates gateway IDs from the hostid file and passes the truncated
gateway IDs to CLDB.
- MFS-11431
- NFS v4 server crashes when running the connectathon lock test.
- MFS-11460
- The VOLUME_ALARM_COMPACTION_FAILURE alarm is repeatedly raised.
- MFS-11558
- The system should throw an error when a volume topology is changed to an empty
topology.
- MFS-11609
- With an EBF patch, Hadoop distcp jobs failing when using CLDB hostname and port.
- MFS-11618
- HPE Ezmeral Data Fabric does not return the correct file status
if the file path is relative.
- MFS-11619
- Global policy slave should periodically refresh policies since they may come as part
of mirroring.
- MFS-11621
- RDMA should be disabled when the MAPR_RDMA_SUPPORT is set to
false.
- MFS-11647
- SlowOpsTrace (Tracing Slow Operations) does not function.
- MFS-11649
- Add MAPR_EXTERNAL support for Loopbacknfs.
- MFS-11674
- Enable gfsck to check CRC without blocking operations on
the frontend volume.
- MFS-11682
- CLDB volume dump fails due to an unknown session key.
- MFS-11760
- Fixed auto offload of EC volumes being restarted repeatedly.
- MFS-11729
- mrconfig info threads crashes MFS when
hardlinks feature is not enabled.
- MFS-11731
- Avoid NFS process has Incorrect BuildVersion alarms if they are redundant.
- MFS-11761
gfsck -D is unable to report fid
corruption.- MFS-11765
- Fixed MASTGateway memory crashes.
- MFS-11774
- Offloads fail after a gfsck run.
- MFS-11779
- Fixed a MFS crash due to stack overflow.
- MFS-11808
- Added a command to check the MASTGateway balancer status.
- MFS-11809
- Added a configuration parameter to set the number of volumes to balance at a time,
rather than a default of 5 volumes.
- MFS-11823
- The maprlogin renew command fails to honor
ticket duration value.
- MFS-11835
- Fixed hangs in POSIX client when connections timeout after being inactive.
- MFS-11838
- Jobs fail with the Too many open files error.
- MFS-11855
- MASTGateway balancing does not start as the delete operation on volumes hangs.
- MFS-11858
- HPE Ezmeral Data Fabric needs to support n+n EC
schemes.
- MFS-11861
- Volume jobs are not automatically restarted when MASTGateway balancing aborts.
- MFS-11886
- NFSv4 fails to mount when
cldb.reject.root is enabled.
- MFS-11888
- On SLES 15, the NODE_ALARM_CORE_PRESENT alarm is raised even when
/opt/cores is empty.
- MFS-11901
- The Container Storage Interface (CSI) caches connections to CLDB and does not refresh
them.
- MFS-11902
- Fixed the
configure.sh script to work on SLES 15.
- MFS-11942
- Fixed an error in hoststats service due to AST timezone that is not resolvable
to a name using standard date functions.
- MFS-11947
- Rename operation fails when the directory being renamed is a symlink and the operation
is the first after the filesystem has been initialized.
- MFS-11959
- Fixed CLDB crash when a large number of virtusl IPs (VIPs) are added to NFS server
nodes.
- MFS-11968
- CLDB does not audit the tier lookup operation.
HPE Ezmeral Data Fabric Hadoop Issues
- MAPRHADOOP-102
- Fixed an error that occurred when copying ACES from the MapR-FS to the local FS.
Installer Issues
- MFS-11670
- Installer hangs when finalizing the cluster.
Security
- MON-5041
- Added a security mechanism to prevent Cross Site Request Forgery (CSRF) attacks on the
Control System.
- COMECURE-152
- Updated the Apache Commons Beanutils package to fix a security vulnerability
(CVE-2019-10086).
- COMSECURE-196
- Updated Jetty packages with security fixes.
- COMSECURE-325
- Remove misleading information message "No CLDB key found on host" for KMIP
configuration.
- COMSECURE-331
- Fixed security vulnerabilities in the /opt/mapr/lib/c3p0-0.9.1.2.jar file
(CVE-2018-20433).
- COMSECURE-334
- Fixed a dom4j vulnerability (CVE-2018-1000632).
- CORE-290
- Fixed permissions on files and directories in the
/opt/mapr
directory.
- CORE-293
- The
mapr-zookeeper and mapr-warden processes were
not properly started with systemd when the OS security updates were
installed.
- CORE-555
- Updated the XercesImpl dependency to fix a security vulnerability
(CVE-2012-0881).
- MFS-2551
- Fixed a Local Privilege Escalation vulnerability in
maprexecute.
- MFS-2554
- Fixed a vulnerability where the
cldbautdit.log file stored the
ldap.bindpassword in cleartext.
- MFS-2708
- Disk failure related log files were world writable causing a potential security
vulnerability.
- MFS-3310
- Added alert to warn customers about expiring SSL certificates.
- MFS-5236
- Fixed a Remote Code Execution vulnerability in the Warden Java JMX server.
- MFS-5235
- Fixed a Remote Code Execution vulnerability in the Gateway Java JMX server.
- MFS-5234
- Fixed a Remote Code Execution vulnerability in the CLDB Java JMX server.
- MFS-5229
- Fixed a Remote Code Execution vulnerability in the MAST Gateway Java JMX server.
HPE Ezmeral Data Fabric Streams Issues
- MS-811
- Segmentation fault occurred when deleting topics from monitoring streams.
Upgrade Issues
- ES-58
- Upgrading to MEP 6.0 on MapR 6.1.0 from MapR 5.2.1 with MEP 3.0.3 broke Elasticsearch
and FluentD.
- MS-925
- Upgrading to MEP 6.2 (Spark 2.4.0) broke consumption of Kafka / HPE Ezmeral Data Fabric Streams.
- MFS-2469
- CLDB secondary node was stuck in BECOMING_SLAVE mode during rolling upgrade from MapR
5.2.2 to MapR 6.1.0.
- MFS-2561
- Increased the value of the
cldb.datacontainer.block.updates.max.sec
parameter to avoid overwhelming the cluster with the VOLUME_ALARM_DATA_UNDER_REPLICATED
alarm after upgrade from MapR version 3.0.1.
HPE Ezmeral Data Fabric YARN Issues
- MAPRYARN-161
- The History server deletion thread stopped once it found an invalid application
directory.