Hi, Thanks for your answers.
I tried with ".local" at the end. It did not change anything for qping on execd. The order for name resolution is hosts and then, bind, according to /etc/host.conf In nsswitch.conf, I have (for host resolution) : > hosts: files dns I have checked my named log file and SGE log files... No errors. I have another master node, and if I try the qping command to an exec node, it works !!!!? My /etc/hosts files are the same, and /etc/resolv.conf is the same too... I compared my bind configuration on both DNS masters without success. The bind versions are differents but configuration files seems to be the same. My first master is in the list of administrative host. How can I check more precisely what happens ? -Regards, Remy. 2012/5/3 Simon Matthews <[email protected]> > > > On Wed, May 2, 2012 at 9:06 AM, Rémy Dernat <[email protected]> wrote: > >> Hi, >> >> I have a very strange hostname/IP resolve problem : >> >> >> From qmaster, if I qping a compute node : >> >> qping compute-0-4 537 execd 1 >> >> >> Answer of compute-0-4 : >> >> endpoint compute-0-4.local/execd/1 at port 537: can't find connection >> access denied: client IP resolved to host name "". This is not identical >> to clients host name "" >> endpoint compute-0-4.local/execd/1 at port 537: can't find connection >> >> I have the same problem for all my nodes but this problem does not occur >> with qsub or qstat... ! So SGE is working fine (except that little error). >> >> I know that I have changed some informations in my /var/named/... files >> and in my /etc/hosts , but I can not find my error. >> >> The qping in the other direction, execd to qmaster, is working fine. >> >> The result of >> hostname -s >> on qmaster give the content of $SGE_ROOT/$SGE_CELL/common/act_qmaster >> on all nodes and on qmaster. >> > > > It looks like the mapping of IP address -> hostname does not match the > mapping of hostname -> IP address. Depending on how your network resolves > the IP address, this could be in: > /etc/hosts > NIS "hosts" map > DNS > etc.. > > You need to check the order specified in /etc/nsswitch.conf and then > examine each mechanism (files, NIS, DNS) to see what name the IP address > of compute-0-4 resolves to in the first mechanism that resolves it. > > Simon >
_______________________________________________ users mailing list [email protected] https://gridengine.org/mailman/listinfo/users
