Pivotal Knowledge Base

Follow

GPText Upgrade 1.2.0.0 to GPText 1.2.2.0 Failed [CRITICAL]:-Installation Failed during Java Version Validation

Environment

Pivotal GPText

Symptom

Attempt to upgrade GPText 1.2.0.0 to GPText 1.2.2.0 failed with the following error:

******************************************************************************

Install Greenplum Text Search into </usr/local/greenplum-text-1.2.2.0>? [yes|no]
********************************************************************************
yes
20140115:21:36:18:026111 install.py:default-[INFO]:-Validating Java versions...
20140115:21:36:22:026111 install.py:default-[CRITICAL]:-Installation failed: Java version validation failed on sdw2. Failed to find java command in PATH.
20140115:21:36:22:026111 install.py:default-[INFO]:-Cleaning up temp files on GPDB hosts...
20140115:21:36:23:026111 install.py:default-[INFO]:-Waiting for worker threads to complete...
20140115:21:36:28:026111 install.py:default-[CRITICAL]:-Installation failed.

Cause

In this case cluster nodes have only 8GB memory configured in all nodes. So during gpext upgrade java processes failed.

When run "Java -version" across cluster nodes, many returned errors "Could not reserve enough space for object heap, Could not create the Java virtual machine." 

gpssh -f hostfile "java -version"
[sdw1] Error occurred during initialization of VM
[sdw1] Could not reserve enough space for object heap
[sdw1] Could not create the Java virtual machine.
[sdw2] Error occurred during initialization of VM
[sdw2] Could not reserve enough space for object heap
[sdw2] Could not create the Java virtual machine.
[ mdw] java version "1.6.0_22"
[ mdw] OpenJDK Runtime Environment (IcedTea6 1.10.4) (rhel-1.41.1.10.4.el6-x86_64)
[ mdw] OpenJDK 64-Bit Server VM (build 20.0-b11, mixed mode)

Resolution

If this is production, please add more memory to the cluster, for more details please read GPDB and GPText guide.

Workaround

We prefer to have more memory, if this is DEV/SAND cluster you may try changing overcommit_memory from 2 to 1 helps to resolve this issue.

gpssh -f hostfile "cat /proc/sys/vm/overcommit_memory"
gpssh -f hostfile "echo 1 > /proc/sys/vm/overcommit_memory"

 

Comments

Powered by Zendesk