No ability to manage NG cluster if webserver adding failedMost of cluster management operations are performed using asynchronous POA tasks (one or many). To prevent errors that may occur during simultaneous execution of several operations cluster is locked for the time of execution of one operation. Operation is considered finished when last task related to it was successfully completed or deleted from Task Manager. So, if operation fails (some task related to it fails) cluster is still locked and management controls in POA UI are disabled. To enable management controls POA administrator has to finish current operation: check why tasks failed and restart them or cancel.
There is one exception from this rule: director switching operation locks cluster only when running (there are running tasks related to it).
No ability to manage cluster if broken webserver was set as DirectorSwitching director locks cluster only while running (there are tasks in running state) or there is another operation, which was started after failed switching.
If LB is deployed as Parallels Server Bare Metal (PSBM) Virtual Machine, all IP addresses are removed after its restartConfiguration will be restored either after any new web space provisioning or by the following command on POA Management Node:
# /usr/local/pem/etc/web_cluster/lvsctl.py update --cluster-id=<cluster id>
IP address is not set up on loopback interface on web serverIP address can be configured manually by the following command on web server:
# service shstg reconfigure
Error 'Host # with similar properties already exists' appearsDuring NG cluster creation web servers are being registered as normal POA hosts and later are being converted to cluster part by the task 'Registering Web Server' . So if the latter task fails and you want to recreate the cluster you have to manually delete half-registered web server in POA Provider Control Panel at Top > Deployment Director > Server Manager > Hardware Nodes.
Configuration Database synchronization failure cause stopping PostgreSQL service on corresponding serverIf webspace or cluster replication failed Configuration Database on the CDB server becomes invalid. POA WebCluster Service Controller issues request to stop database service on the failed CDB server to force web servers reconnect to other CDB server. If customers databases are hosted on same failed server they became inaccessible.
See the main Knowledgebase article #114326 Linux Shared Hosting NG: General Information, Best Practices and Troubleshooting for more information about NG Hosting in Parallels Automation.