Mechanism of Hostname Resolution in Ansys Gateway
The Ansys Gateway does not typically act as a reverse proxy for the final result transfer; instead, it brokers the initial handshake and job submission. When a client identifies itself via a short-name, that identifier is passed to the compute cluster. The compute nodes then attempt to establish a direct connection back to the client to transfer result files. If the compute node cannot resolve that short-name to an IP address using its own local DNS configuration, the connection will time out.
Likely Explanation vs. Confirmed Behavior
It is a confirmed architectural requirement that the compute nodes must be able to resolve the client's identity. However, the mechanism for handling short-names is not a specialized Gateway feature, but rather a dependency on the underlying OS network stack of the compute nodes. If a client is on a different subnet, the compute node's DNS search suffix list will likely fail to append the correct domain to the short-name, leading to the timeout.
Resolution Steps
To ensure the compute node can resolve the client without modifying global DNS search suffixes, use one of the following scoped methods:
- Client-Side FQDN: Configure the client to identify itself using its Fully Qualified Domain Name (FQDN) rather than a short-name. This removes the need for the compute node to guess the DNS suffix.
- Local Hosts Mapping: For static environments, add the client's IP and hostname to the
/etc/hosts (Linux) or C:\Windows\System32\drivers\etc\hosts (Windows) file on the compute nodes.
- Shared Staging Area: Configure the Remote Solve settings to use a shared network file system (NAS/SAN) for result staging. This changes the workflow from a push (node to client) to a pull (client from share), bypassing the need for the compute node to resolve the client's hostname entirely.
Verification
To verify if DNS resolution is the root cause, execute the following command from a compute node during a failed retrieval attempt:
nslookup [client-short-name]
If this returns Non-existent domain or times out, the issue is confirmed as a resolution failure.
Missing Diagnostic: Are you utilizing a shared filesystem for result staging, or is the system configured for direct transfer from the compute node to the local client machine?
ansys