SymptomsOn a Parallels Operations Automation (POA) installation with Linux Shared Hosting NG that has a large amount of webspaces (upwards of 1,000), the following errors may be logged in the file /var/log/h2e-nss.log or /var/log/httpd/error_log on the web server in a cluster:
given time has been expired while waiting for response
CauseThis error means that the Parallels Configuration Service (shstg) is overloaded and is unable to issue a timely reply to Apache, SSH, or ProFTPd. The probable cause of the overload is that PostgreSQL 8 is used for the Configuration Database.
ResolutionWith POA version 5.3 and later versions, PostgreSQL 9 is supported as a configuration database for Linux Shared Hosting NG. Use the instructions below to update the configuration database from PostgreSQL 8 to PostgreSQL 9:
- Register the new hardware host or create the new shared VPS in POA.
- Install the package webcluster_cfg_db on the new host.
- Go to Top > Deployment Director > Server Manager > Web Clusters > click on web cluster > Configuration Databases tab.
- Click the Assign button and choose the newly added server.
- Wait for the synchronization process to finish.
- Remove the old configuration database.
Additional InfoThe task Deploy database for <webcluster name> to host <id> scheduled during new Configuration Database deployment may fail with an error message like this one:
createuser: could not connect to database postgres: fe_sendauth: no password supplied
Perform the following actions to resolve the problem:
- Log in to the host with Linux Shared Hosting NG Configuration Database by SSH as root.
- Change the line in the file /var/lib/pgsql/data/pg_hba.conf
host all all <MN_IP> 255.255.255.255 password
host all all <MN_IP> 255.255.255.255 trust
where <MN_IP> is the IP address of the POA Management Node.
- Add following line to /var/lib/pgsql/9.0/data/postgresql.conf
bytea_output = 'escape'
- Issue this command:
# service postgresql-9.0 reload
- Restart the failed task.