formating HDFS /grid/hadoop/hdfs/nn is not enforced/tested in charm

Bug #1425456 reported by Samuel Cozannet on 2015-02-25
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
hdp-hadoop (Juju Charms Collection)
Undecided
Juju Big Data Development

Bug Description

This bug comes when joining a hdp-pig-4 charm to an existing HDP cluster made at least of a YARN Master (hdp-hadoop-7) and compute nodes (same versions)

By default, when creating the cluster, the charms will format a HDFS partition in /grid/hadoop/hdfs/nn.

When adding the PIG node, it will try to do it again. Except that if it has already been done, it won't work because it's not enforced or tested.
It then goes into an infinite loop of:

unit-yarn-master-0[28041]: message repeated 3809 times: [ 2015-02-25 09:33:59 INFO unit.yarn-master/0.namenode-relation-joined logger.go:40 Re-format filesystem in Storage Directory /grid/hadoop/hdfs/nn ? (Y or N) Invalid input: ]
unit-yarn-master-0[28041]: 2015-02-25 09:34:00 INFO unit.yarn-master/0.namenode-relation-joined logger.go:40 Re-format filesystem in Storage Directory /grid/hadoop/hdfs/nn ? (Y or N) Invalid input:

The correct behavior should be
* to test if the partition is already formated.
* Pass if yes (we don't know if data has been posted there, don't want to erase it)
* Create/Format otherwise

Changed in hdp-hadoop (Juju Charms Collection):
assignee: nobody → Juju Big Data Development (bigdata-dev)
To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Other bug subscribers