Probable Cause: An invalid DIM argument was passed to an HPF library or intrinsic routine.
Probable Cause: The error message issued does not accurately reflect the problem that caused the program to abort.
Probable Cause: An invalid rank for BASE was passed to an HPF library or intrinsic routine.
Probable Cause: An invalid rank for INDEX was passed to an HPF library or intrinsic routine.
Probable Cause: An invalid rank for MASK was passed to an HPF library or intrinsic routine.
Probable Cause: An invalid rank for return value was passed to an HPF library or intrinsic routine.
Probable Cause: An invalid rank for SEGMENT was passed to an HPF library or intrinsic routine.
Probable Cause: An invalid rank BASE was passed to an HPF library or intrinsic routine.
Probable Cause: An invalid shape for INDEX was passed to an HPF library or intrinsic routine.
Probable Cause: An invalid shape for INPUT1 was passed to an HPF library or intrinsic routine.
Probable Cause: An invalid shape for INPUT2 was passed to an HPF library or intrinsic routine.
Probable Cause: An invalid shape for MASK was passed to an HPF library or intrinsic routine.
Probable Cause: An invalid shape for return value was passed to an HPF library or intrinsic routine.
Probable Cause: An invalid shape for SEGMENT was passed to an HPF library or intrinsic routine.
Probable Cause: An incorrect type parameter for an ARRAY was passed to an HPF library or intrinsic routine.
Probable Cause: An invalid type parameter for BASE was passed to an HPF library or intrinsic routine.
Probable Cause: An invalid type parameter for INDEX was passed to an HPF library or intrinsic routine.
Probable Cause: An invalid type parameter for MASK was passed to an HPF library or intrinsic routine.
Probable Cause: An invalid type parameter for return value was passed to an HPF library or intrinsic routine.
Probable Cause: An invalid type parameter for SEGMENT was passed to an HPF library or intrinsic routine.
Probable Cause: There is not enough processor memory to continue the execution of the program.
PSE startup errors are all flagged with the previous error message. One or more of the following errors may be appended to this error message.
Probable Cause: The directory containing the user executable is not uniformly exported to the client machines, or the file system is mounted on the remote client using a non-uniform path. If the file had just been created, such as part of a script, NFS on the remote node may not have been updated.
User Action: Check with your system administrator. This problem can occur when the file of interest has a non-unique pathname, such as multiple fully qualified pathnames, paths which start from root / to the same file. Typical examples of this are automounted file systems which users see as /usr/local/something; for example, automount introduces a symbolic link and mounts the actual filesystem as /tmp/_mnt/usr/local/something. In cases such as this, the multiplicity of pathnames should be preserved across all hosts.
Another common example of non-unique pathnames are when symbolic links are used. Symbolic links while causing problems on the one hand (as in automount), are a powerful tool by which system administrators can resolve the multiple-pathname problem.
Probable Cause: There are no available job slots on some of the selected machines; farm daemons have not been started; or user environment variables (see PSE_PARTITION, PSE_EXCLUDE, PSE_ON, and PSE_USE) have been set incorrectly.
User Action: In PSE physical mode (the default),
each application decomposes to at most one process peer for each
farm member in the partition. If the user program has been compiled
for n processes and the partition is composed only of
M < n processors, they must either use the virtual
mode, (see PSE_MACHINE=[physical|literal] ) or choose a
partition with a greater number of hosts.
Probable Cause: The current working directory of the user is not uniformly exported to the client machines or the file system is mounted on the remote client using a non-uniform path.
User Action: Check with your system administrator. This problem can occur when the file of interest has a non-unique pathname, such as multiple fully qualified pathnames, paths which start from root / to the same file. Typical examples of this are automounted file systems which users see as /usr/local /something; for example, automount introduces a symbolic link and mounts the actual filesystem as /tmp/_mnt/usr/local /something. In these cases, the multiplicity of pathnames should be preserved across all hosts.
Another common example of non-unique pathnames are when symbolic links are used. Symbolic links, while causing problems on the one hand (as in automount), are a powerful tool by which system administrators can resolve the multiple-pathname problem.
Probable Cause: User specified a nonexistent farm or the specified farm database is not accessible from the local machine.
User Action: By default, PSE assumes that the requested farm database is available using DNS/bind. System administrators must make sure that the named(8) daemon responsible for distributing the desired farm database is accessible to all farm nodes. The easiest means of accomplishing this task is to add a nameserver definition line to all /etc/resolv.conf files. Another convenient tool used to distribute DNS/bind information is to have the local network administrator add the local NS (nameserver) records to the master bind database. This is the most general solution as all DNS/bind servers forward domain requests up or down the domain hierarchy until it reaches a responsible named(8) daemon for resolution.
Probable Cause: The environment variable PSE_PARTITION is set incorrectly or there is a mistyped command line.
User Action: If the lspart command shows a similar error, unset the PSE_PARTITION environment variable or set it to one of the PSE partitions identified by:
% lspart -farm (name) -all
Probable Cause: There are no available job slots on the selected machines; farm daemons have not been started; or environment variables (see PSE_EXCLUDE, PSE_ON, and PSE_USE) have been set incorrectly.
User Action: Use the lspart command to determine the state of all machines in the requested partition. If the command identifies hosts as having No farm daemon running, there may be a network outage or the farm daemon may actually not be installed correctly. In either case, the local system administrator should be consulted.
If lspart seems to indicate that no problems exist with the requested farm nodes, then a review of the individual PSE environment variables may show unexpected results.
Finally, when a user's execution environment has been reviewed and is acceptable, the PSE application can be restarted. Some installations may decide to set individual host, Jobslot values to one (1) or to a minimal number to maximize single application performance by limiting the interference amount competing farm applications. Users in this situation should consider funneling applications to the farm using a simple batch queue, reducing the competition for a critical resource.
Probable Cause: The PSE_LOADSERVER environment variable identifies hosts which are not members of the specified farm, the requested farm is not currently active.
User Action: Undefine the PSE_LOADSERVER environment variable and issue an lspart(1) command. If the resulting output shows active farm daemons, reset the PSE_LOADSERVER environment variable to the loadserver set shown. If however, no active farm members are identified, consult with the system administrator and/or use the lspart -search command to find an active farm in their local domain.
Probable Cause: The PSE_FARM environment variable is not defined nor is a valid farm name specified using the -farm name command line option.
User Action: PSE attempts to find a default service named farm if no other information is given, displaying this error message. Every installation may differ in use of names to identify local farms. To find the set of valid farm names installed at the local site, issue the command:
% lspart -search
If problems persist after having specified a valid farm name (as shown by lspart), system administrators must assign the PSE Farm names to Internet port numbers in the /etc/services file. The service port number identified in /etc/services must match the PSE_SERVICEPORT definition in the farm database.
Probable Cause: The resource limits inherited from the controlling process cannot be used, a maximum limit on the specified host is substituted. Note that this message is displayed ONLY during verbose mode.
User Action: If user applications generate core dumps for no apparent reason, they should be re-run specifying -verbose to determine whether any relevant configuration differences exist between farm nodes. If the maximum limit value (stacklimit size,for example) is too small for the desired application, consult with the system administrators to reconfigure the system kernels with acceptable limits, or -exclude the offending host(s) from the user partition.
Probable Cause: The specified PVM file either does not exist, is empty, or is read access protected.
These messages appear when a command line argument is missing.
Probable Cause: PSE command parsing has detected an error. Either the argument to the command line option is missing, or the user entered a nondecimal value.
User Action: A list of accepted command line options is printed along with this error message.
Probable Cause: PSE command parsing has detected an error. This is the result of a missing argument to either the current option (as indicated) or to a previous command line option.
User Action: A list of accepted command line options is printed along with this error message.
Probable Cause: PSE host definitions must either be aliases to existing host names or reflect the actual hostnames as defined by the /etc/hosts file.
User Action: When specifying multiple hosts as a list, valid PSE separators are tabs, spaces, or commas. Any other separator causes the PSE parsing to fail with this error message. Note too, that lists separated by tab or space characters must be quoted to be passed by the user shell as a single command line argument to the PSE command parser.
