Torque the wrong line.

One: The management node discovers that a node status is down , there are two reasons for this situation

The first reason:pbs_mom configuration File Error

1. Analysis : After the service Pbs_mom is started, the default is to first check if there is a config in the/var/spool/torque/mom_priv directory file, if any , to contact the management node according to the hostname of the pbsserver indicated under the file (whichever is the file), if the file is not available, according to/var/spool/torque/server_name File is indicated under the hostname of the Pbsserver to contact the management node, then if the two files indicated within the pbsserver IP address of the compute node can not ping , Then the compute node pbs_mom logs will report the following error:

10/16/2015 20:49:46;0001; pbs_mom.4538; Svr;pbs_mom; Log_error::inappropriate IOCTL for device (+) in tcp_connect_sockaddr, cannot connect to port 9 in Socket_connect_addr-er Rno:101 Network is unreachable

10/16/2015 20:49:46;0001; pbs_mom.4538; Svr;pbs_mom; Log_error::mom_server_all_update_stat, Couldnot Contact any of the servers-to-send an update

of course the management node will also show that the node status is down

2. Workaround :

Method ⑴ : If a config file exists under the/var/spool/torque/mom_priv directory , modify the file $pbsserver The value is the correct Management node hostname (the compute node can ping the host name)

Method ⑵ : when the config file does not exist in the/var/spool/torque/mom_priv directory , make sure that the/var/spool/torque/server_name The correct management node host name is the value of the

The second reason:/var/spool/torque/server_priv/nodes The host name is wrong within the file

1. Analysis : First in the compute node to view the Pbs_mom log, no errors found, and then view the Management node's pbs_server log, The following error message is found:

10/16/201521:22:26;0001; pbs_server.11915; SVR; Pbs_server; Log_error::get_node_from_str,node node1.wjcyf.com is reporting on node node1, which pbs_server doesn ' t know about

service Pbs_server found a host named node1.wjcyf.com , but according to Get_node_from_str know that the hostname should be Node1 , all the compute nodes cannot be contacted

2. workaround :

Change the host name of the compute node involved in the/var/spool/torque/server_priv/nodes file to the "uname-n" hostname, that is, if the compute node node1 on the Run command uname-n get the host name Node1, then the file within the Node1 hostname is written as Node1, if it is Node1.wjcyf.com, it's written as node1.wjcyf.com, etc.

This article from "Never Stop" blog, reproduced please contact the author!

Torque the wrong line.