[torqueusers] initial pbs_server start hangs
knielson at adaptivecomputing.com
Mon May 16 10:12:39 MDT 2011
On 05/12/2011 03:17 PM, David Beer wrote:
> ----- Original Message -----
>> We are testing Torque v3.0.1, in preparation for an upgrade we have
>> During the initial install (on an empty system), the
>> "/etc/init.d/pbs_server start" command hangs, which is kind of
>> since we are doing a fully automated install and this issue requires a
>> manual intervention.
>> What seems to be happening is that during the initial startup process,
>> the pbs_server init script notices that the server_priv/serverdb isn't
>> there yet, so it tries to create it using "pbs_server -t create".
>> It's starts the creation in the background, waits until the serverdb
>> file has been created, and then stops the pbs_server again using the
>> killproc function.
>> However, it seems that the "pbs_server -t create" isn't creating the
>> serverdb file straight away, but buffering it. This causes the
>> pbs_server init script to wait for it indefinitely, and thus makes the
>> whole process hang.
>> If the pbs_server -t create process is killed with SIGTERM, the
>> cleans up after itself, flushes the files it's supposed to write and
>> then terminates. This forces the creation of the server_priv/serverdb
>> file, and thus makes the initial startup of pbs_server finish as
>> Is this a known issue? It seems like forcing pbs_server to flush the
>> serverdb file should fix this...
> Can you create a bugzilla for this issue? I haven't seen this issue in the past but I will check it out.
We may need a little more information about your automated install.
Part of installing torque is the initialization of the serverdb. Once
serverdb has been created then the use of /etc/init.d/pbs_server start
is ready to go. However, pbs_server start is not intended to be part of
the installation process. Can you make sure serverdb is initialized
before you begin to use pbs_server start?
More information about the torqueusers