Am 18.06.2014 um 15:20 schrieb Esztermann, Ansgar: > OK, I've found the culprit, but I do not really understand the reason. > >>> It was only a test whether by default call of the function will output the >>> short name or the FQDN. I still wonder, why this happens. > > I've looked at the source code (from MPICH) so see how this works. > HYDU_sock_is_local() turns out to be pretty intelligent. It looks at > hostnames as a first step only, so if these do not match (e.g. one is FQ > while the other is not), it will look up IP addresses and compare these. If > necessary, it does this for all local interfaces. > > However, there is an overriding environment variable > (HYDRA_LAUNCHER_AUTOFORK). If this is set to 0, no, false (or similar), qrsh > is used even for local processes. Now, according to the source, this variable > defaults to true. However, this is not what I observe. > Explicitly setting HYDRA_LAUNCHER_AUTOFORK fixes the problem.
Well, maybe Intel changed the default behavior. Thx for sharing your findings. -- Reuti > A. > > -- > Ansgar Esztermann > DV-Systemadministration > Max-Planck-Institut für biophysikalische Chemie, Abteilung 105 > > _______________________________________________ > users mailing list > [email protected] > https://gridengine.org/mailman/listinfo/users _______________________________________________ users mailing list [email protected] https://gridengine.org/mailman/listinfo/users
