I checked the failed job log and found that there was one such item, java.lang.NoSuchField Error: HBASE_CLIENT_PREFETCH.   Then I look at the classpath of the failed job, find different versions of the jar package of HBase in hive. aux. jars. path, delete the jar package here, restart hiveserver2, and everything returns to normal. ... View more

thanks first. I uninstall spark1, then install spark2.1，everything is OK ... View more

I have done what you said.    In addition to the DFS directory'/dnn/data1/dfs/dn/current', the data disk of the data node is the YARN container log file.  And the log file is only about 5G. I tried the command   'lsof | grep deleted'  ,and found no deleted files occupied.   It's no use restarting YARN's job history alone. Finally, I had to restart HDFS, and then something magical happened.   At the moment of restart completion, Non DFS Used dropped directly from 54T to 4T.   I'm curious to know how datanode calculates the remaining available capacity of a node.   I read the source code, but found no commands, such as du, df. ... View more

HI, thanks first for reply. for disk reserved block ,I have done it . tune2fs -m 1 /dev/sda1 ,(and sdb1,sdc1,sdd1) and the value of cluser Non DFS Used is growing slowly。just like From 44T to 54T in 7 days。 how to reduce the Non DFS Used ? ... View more

hu,I have done it. I would like to mention a few points. 1. The disk mount information of the host must be backed up before the host is offline (partition UUID mount information is very important), and the /etc/passwd and /etc/group files of the host must also be backed up. 2. Do not check the Unauthorization button when deleting all roles on the host from the cluster. Otherwise, you need to wait for replication and the action will become slow. 3. Shortly after DN is offline, HDFS will detect insufficient replicas and start replicating. Some configurations can be adjusted, but because I don't want to restart the cluster, there will be no modification, which will not have much impact on the operation of the cluster. 4. it is best to build local private CM, CDH and OS resources. This will greatly reduce the speed of node installation. 5. After the host reinstalls the operating system, it is necessary to mount the disk to its original location through the previously backed-up partition UUID to ensure that the data can be reused. 6. It is very important to manually add users and groups backed up before adding them on the host to ensure that the mapping relationship is the same as the original one, so that the privileges of these data can be correctly identified by HDFS again smoothly. 7. After everything goes smoothly, the HDFS DFS fsck / command can quickly check whether the replica is sufficient and stop the previous three replica recovery actions in time.   My English is not good, the expression may not be very accurate, there are problems that can continue to discuss with me. ... View more

Hi,      I'm cluster manager,and CDH version is 5.7.2.       The same trouble,If I can change some params in CM to solve this problem.  ... View more

Thank you so munch！ I change the group of '/tmp/logs' to hadoop , and  restart the JobHistoryServer role, everything being OK.        So happy ! ... View more

Hi , my cluster is CDH 5.7.2,CM5.7.0,  and I meet the same touble. we set   dfs.permissions.superusergroup=supergroup    ;  and we run the mapreduce application by 'hdfs' user, the hdfs file like this:       drwxrwx---   - hdfs       supergroup          0 2018-06-05 15:01 /tmp/logs/hdfs   and the linux mapping of user to group is :         hadoop:x:497:hdfs,mapred,yarn         supergroup:x:505:hdfs,yarn   what should I do to resolve this problem?  thanks you very much.   ... View more

大家好，我来自中国。刚接触hadoop两年时间。主要负责大数据集群的运维工作。欢迎多多交流。 ... View more

I think it's the copyright issue of cloudera. Modifications should not be allowed. Personal view only。。。 ... View more