Oninit
Talk to an expert →

Informix Error -55276: connect( ) to host 'host-name' using service 'service-name' failed.

Cause and resolution

connect( ) to host 'host-name' using service 'service-name' failed.

The possible causes of this error are as follows:

* The onprobe server is not running on the specified host.

* The onprobe server on the specified host uses a service other than specified.

Service-name resolution follows this precedence:

1. NIS services map when NIS (Network Information Service) is running.

2. File /etc/services on the local computer. If onprobe and oncockpit run on different computers, check whether the port number assigned to the specified service in /etc/services files on these computers matches.

* The specified TCP/IP address is not available on this computer. Check whether 'ping' and 'rlogin' commands can access the specified host. The 'ping' command checks the ICMP network layer. The 'rlogin' command checks the TCP network layer and specified host (but not specified service) availability.

* Connection establishment timed out without establishing a connection. The connection requests queue of the onprobe server might be full. The maximum queue length is 5.

* Unavailable system buffer space.

Oninit® Troubleshooting Guidance

Reasons / Common Causes

-55276 fires when connect() fails after host and service names both resolved — per the official guidance, several possible causes are named, this being the final step in the onprobe/oncockpit connection chain.

  1. The onprobe server not running on the specified host, per the official guidance.
  2. The onprobe server on that host using a different service than specified, per the official guidance — check NIS (ypcat services) or local /etc/services on both ends for a matching port number.
  3. The specified TCP/IP address not reachable from this computer, per the official guidance — checked with ping (ICMP layer) and rlogin (TCP layer plus host/service availability).
  4. The connection timing out because the onprobe server's connection-requests queue is full, per the official guidance — the documented maximum queue length is 5.
  5. System buffer space being unavailable, per the official guidance.
Solutions / Resolution
  1. Confirm the onprobe server is actually running on the target host.
  2. Compare the service/port used on both computers via NIS or /etc/services, per the official guidance, if onprobe and oncockpit run on different hosts.
  3. Use ping and rlogin to separately check ICMP reachability and TCP-layer/host-service availability, per the official guidance.
  4. If the server's connection queue is full (max 5), retry once load clears, per the official guidance.
Examples
Separating ICMP reachability from TCP/service availability
$ ping <host-name>
$ rlogin <host-name>

Per the official guidance, ping checks the ICMP network layer only; rlogin checks the TCP layer and the specified host's (though not the specified service's) availability.

Diagnostic Checks
  1. Confirm onprobe is running on the target host.
  2. Compare service/port definitions between the two computers.
  3. Run ping then rlogin to isolate ICMP vs. TCP-layer reachability.
  4. Check whether the server's 5-slot connection queue is full.
Related Errors / Related Topics
  • -55250 — "socket() failed." The first step in this same connection chain.
  • -55251 — "Could not translate service name... getservbyname() failed." An earlier step, resolving the service name this call uses.
  • -55252 — "bind() using service... failed." Another step in the same chain.
  • -55274 — "Could not translate host name... gethostbyname() failed." An earlier step, resolving the host name this call uses.

The final connection step failed after host/service names both resolved — confirm onprobe is actually running on the target host, then check reachability (ping/rlogin) and the server's connection queue.