Troubleshooting

In this page you can find the answers to some errors that may occur while starting the Ophidia framework the first time. The errors refer mainly to what is shown by the Ophidia terminal or in the Ophidia server log (commonly server.log).

Issue

The Ophidia terminal shows the following error when it starts:

SOAP 1.1 fault SOAP-ENV:Server [no subcode]
"Connection refused"
Detail: connect failed in tcp_connect()

Solution

The terminal isn’t able to reach the Ophidia Server using the connection arguments specified. Possible sources of this error can be:

  • a firewall is blocking terminal-server communications;
  • connection arguments are incorrect;
  • the Ophidia Server is not running.

Issue

The Ophidia Server can’t startup showing the following message in the log:

SSL Server Context Error
Server shutdown
Error 30 fault is internal [no subcode]
"SSL/TLS error"
Detail: Can't read CA file

Solution

The server can’t start due to a problem with the SSL keys used for client-server encryption. Possible sources of this error can be:

  • SSL keys aren’t valid or haven’t been created properly. The Server configuration page shows a sample procedure to generate keys;
  • the SSL key password set in the server configuration file is incorrect. Check the CERT_PASSWORD argument in the server.conf file. It should matcht the one used to generate the keys;
  • the server can’t find the keys in the location specified. Check the paths specified in CA and CERT configuration parameters inside the server.conf file.

Issue

The terminal shows the following error when a command is submitted:

Error on serving request [4]: server IO error.

At the same time the server log reports:

R5: unable to connect to OphidiaDB. Check access parameters.

Solution

In this case the server can’t interact with the OphidiaDB. Possible sources of this error can be:

  • the MySQL server hosting the OphidiaDB isn’t reachable (behind a firewall) or it’s not running;
  • MySQL server access parameters specified in the server configuration file ophidiadb.conf are not correctly set (see server configuration);
  • even though the parameter access are correct, the user specified in the configuration file cannot access the database. In this case the MySQL server user’s grants should be fixed.

Issue

The terminal shows the following error when a command is submitted:

Error on serving request [4]: server IO error.

At the same time the server log reports:

Authentication by public key failed
Error during remote submission
R4: error in serving the request; reporting the notification ...

Solution

The Ophidia server is not able to connect remotely through SSH to the node where the Slurm master is running for job submission. Possible sources of this error can be:

  • the configuration values specified in the server.conf file aren’t correct. You should check IP_TARGET_HOST, SUBM_USER, SUBM_USER_PUBLK and SUBM_USER_PRIVK (see server configuration);
  • the server public key is not allowed to perform a password-less key-based connection to the SSH target node;
  • the target node doen’t allow SSH connection or key-based connections.

Issue

The terminal shows the following error when a command is submitted (via SSH channel):

Gtk-Message: Failed to load module "canberra-gtk-module"
Gtk-Message: Failed to load module "pk-gtk-module"

Solution

The problem is due to absence of some packages used to display graphical outputs. Some commands exploiting this feature are view and check.

Install the packages required to solve the issue. For example, the following command should be adequate for CentOS 7.

yum install libcanberra-gtk3 PackageKit-gtk3-module xdg-utils