question Re: For CDH 5.7.0 what is the full path to an HDFS datanode's PID file? in Archives of Support Questions (Read Only)

For CDH 5.7.0 what is the full path to an HDFS datanode's PID file?

JeffKubina — Fri, 30 Jun 2017 21:05:31 GMT

For CDH 5.7.0 what is the full path to an HDFS datanode's PID file?

Re: For CDH 5.7.0 what is the full path to an HDFS datanode's PID file?

csguna — Sat, 01 Jul 2017 02:48:40 GMT

Could you check your /tmp .

Re: For CDH 5.7.0 what is the full path to an HDFS datanode's PID file?

JeffKubina — Wed, 05 Jul 2017 13:32:35 GMT

I did, it's not in /tmp.

Re: For CDH 5.7.0 what is the full path to an HDFS datanode's PID file?

csguna — Wed, 05 Jul 2017 14:58:57 GMT

check the status of datanode dameon in that host

meantime the pfile should be in

/var/run/hadoop-hdfs/hadoop-hdfs-datanode.pid

Re: For CDH 5.7.0 what is the full path to an HDFS datanode's PID file?

JeffKubina — Wed, 05 Jul 2017 15:43:35 GMT

I can get the PID via a "pgrep -fa datanode" but would prefer to find where
it writes its PID to a file.

Re: For CDH 5.7.0 what is the full path to an HDFS datanode's PID file?

csguna — Wed, 05 Jul 2017 16:28:53 GMT

Check the below location for datanode pid file

/var/run/hadoop-hdfs/

Re: For CDH 5.7.0 what is the full path to an HDFS datanode's PID file?

bgooley — Wed, 05 Jul 2017 16:41:56 GMT

@JeffKubina,

Since the supervisor starts CM managed processes, there is no PID file for the processes it starts.

Short answer: there is no PID file.

-Ben

Re: For CDH 5.7.0 what is the full path to an HDFS datanode's PID file?

JeffKubina — Wed, 05 Jul 2017 17:00:35 GMT

So what is the recommended way of getting the datanodes PID?

Re: For CDH 5.7.0 what is the full path to an HDFS datanode's PID file?

bgooley — Wed, 05 Jul 2017 17:10:48 GMT

@JeffKubina,

While there are a number of ways to get the PID (like ps aux |grep ....), this is one I just thought up:

/usr/lib64/cmf/agent/build/env/bin/supervisorctl -c /var/run/cloudera-scm-agent/supervisor/supervisord.conf pid `ls -lrt /var/run/cloudera-scm-agent/process/ | awk '{print $9}' |grep DATANODE| tail -1`

What it does is run supervisor's client using the Cloudera Manager sueprvisord.conf file to retrieve a pid file. The backticks execute a command to get the most recent process directory to pass into the "pid" command of supervisorctl. You can see the process names and statuses using the following:

/usr/lib64/cmf/agent/build/env/bin/supervisorctl -c /var/run/cloudera-scm-agent/supervisor/supervisord.conf status

-Ben

Re: For CDH 5.7.0 what is the full path to an HDFS datanode's PID file?

bgooley — Wed, 05 Jul 2017 17:13:13 GMT

@JeffKubina

Oops, that was overkill... forgot that supervisorctl status output has the PID column already... So, you can run:

/usr/lib64/cmf/agent/build/env/bin/supervisorctl -c /var/run/cloudera-scm-agent/supervisor/supervisord.conf status

The PID will be the 4th column.

-Ben

Re: For CDH 5.7.0 what is the full path to an HDFS datanode's PID file?

csguna — Thu, 06 Jul 2017 01:26:30 GMT

@bgooley

if the cluster is not managed by CM then the path for PID is the below

Correct me if I am wrong ,

/var/run/hadoop-hdfs/

Re: For CDH 5.7.0 what is the full path to an HDFS datanode's PID file?

JeffKubina — Thu, 06 Jul 2017 14:31:35 GMT

That is correct.