Task #8018
When the hazlecast ports of cns are blocked to each other, the orchid log-in doesn't work
100%
Description
In our CN upgrade process, we will ran a script which will block the hazelcast ports among the three CNs. So the communication among them will be disconnected. In this situation, the orchid log-in on the search interface doesn't work.
The error message shows the connection to hazelcast server was timeout in the SessionHelper class on d1_portal_servlet.
The class reads the hazelcast server address configuration from the file /etc/dataone/portal/hazelcast.xml. The file is configured:
128.111.54.78
64.106.40.9
160.36.13.153
So the client will connect to cn-dev-orc-1 from cn-dev-ucsb-1. However, the connection is blocked and the timeout happened. So we have to manually comment out the address 64.106.40.9 and 160.36.13.153 and force the client to connect hazelcast server on cn-dev-ucsb-1. After the installation is done and regular operation is restored, we have to uncomment the configuration. This make the installation complicated.
I propose to use another hazelcast-client.xml file which will be used in this class. In the hazelcast-client.xml file, the hazelcast server address will be localhost. So the server still is distribute on three nodes, but the client only connects to the local host and we don't change the configuration anymore . This change introduce a new configuration file but automate the installation process.
History
#1 Updated by Jing Tao about 7 years ago
- Assignee changed from Ben Leinfelder to Jing Tao
It was fixed on the 2.3 branch and trunk
#2 Updated by Jing Tao about 7 years ago
- Status changed from New to Closed
- % Done changed from 0 to 100
