Start a Test Queue
At a given site, a single queue is recommended, such as a single Accelerator setup, by default called "vnc". Such a setup is called a cluster by other systems. The scheduler in Accelerator does not have queues in the sense used by other batch systems.
In this tutorial, you will start a temporary Accelerator queue for our testing. This allows you to experiment with most of the administration tasks without disturbing your production Accelerator queue.
Find the Server Working Directory
$VOVDIR/../../vnc/vnc.swd
This directory contains the server configuration files and server/tasker logs, etc.
Start a Queue
The command to start a queue is ncmgr start. By default, this command will start the default queue in the directory of $VOVDIR/../../vnc.
-
Get usage of this command:
vncmgr: Usage Message USAGE: % ncmgr start [options] OPTIONS: -h This help. -force Do not ask confirmation. -block Do not return to the shell or command prompt after starting. This is only useful, and required, when starting Accelerator as a Windows service. -port <port|mode> Specify port number, port number list (colon separated) or port mode. Modes are: automatic - hash queue name into port number, do not start if port is unavailable. The default queue name 'vnc' hashes to port 6271. any - hash queue name into port number, try additional ports in increments of 1 until an available one is found. The default queue name 'vnc' hashes to beginning port 11437. Default: any -eventport <port[:port]> Colon separated update port list wherein each element may be of the form: N - An integer specifying the exact port number to use. automatic - Calculates hash of project name to derive a single port number. any - Calculates hash of project name to derive starting port number. Ports are checked sequentially up to 30000. Event port will be non-functional if there is a conflict with specified ports (default 5559). -webport <port|mode> Specify a dedicated web interface port for HTTP and HTTPS protocols. This port must be configured to enable REST API v3 interface, to enable the dashboard web UI page, and to enable SSL. A value of 0 directs VovServer not to open a web interface port. Specify port number, port number list (colon separated) or port mode. Modes are: automatic - hash queue name into port number, do not start if port is unavailable. The default queue name 'vnc' hashes to web port 6271. any - hash queue name into port number, try additional ports in increments of 1 until an available one is found. The default queue name 'vnc' hashes to beginning web port 9695. Default: Any -webprovider <provider> Specify the provider for HTTP and HTTPS protocols. This must be either "internal" or "nginx". Default: "internal" -roport <port|mode> Specify read-only guest access web interface port. A value of 0 disables this interface, requiring all web interface users to log in. Default: 0 -pubkey <key> Specify the public VOV security key to be added to a new project. If -pubkey is not specified, the public key listed in ~/.vov/userkey or %PROFILE%/.vov/userkey is added. -q, -queue <name> Name for queue (default is $NC_QUEUE if set, and otherwise vnc). -dir <dir> Directory of the server (default $VOVDIR/../../vnc). -dbhost <host> Host for database. -dbroot <path> Path on database host for database files. -dbport <port> Port of the database to listen for connections. -v Increase verbosity. EXAMPLES: % ncmgr start -port 6271 % ncmgr start -port 6271:6272:6273:any -force % ncmgr start -q bigqueue -dir /remote/queues
- Start the test queue in the home directory.
-
Name your queue vncdexin (make sure you pick a name that
does not conflict with any existing queue). The first three letters of the queue
name should be formed by prefixing your login with
vnc
For example, if your login is 'danny', use the name 'vncdanny' for your queue.
+ start a shell on the machine where your NC vovserver should run % cd # Go to your home directory % mkdir ncadmin # Create a directory for our testing queue % ncmgr start -dir ncadmin -queue vncdexin message: Checking the license... message: ... the license is good. message: Starting NC message: with name vncdexin message: on host alpaca message: in directory /home/dexin/ncadmin message: as user dexin Do you want to proceed? (yes/[no]) > yes message: Updating config file '/remote/release/VOV/2013.09/linux64/local/vncConfig/vncdexin.tcl' message: Waiting for server to be ready ... message: Sanity check... message: NC vncdexin@alpaca is ready.
Note: Make sure the directory $VOVDIR/local/vncConfig is writable to you, because a config file needs to be written in that directory to start a new queue. The Altair Accelerator installer should have set this.
Use a Specific Queue
Now that you have started a new queue, you have at least two Accelerator queues in your system. There are two ways to use a particular queue.
-
Use -q or -queue option:
# Submit a job to queue "vncdexin" % nc -q vncdexin run sleep 60 # List my jobs in queue "vncdexin" % nc -q vncdexin list # Get info about queue "vncdexin" % ncmgr info -queue vncdexin # Reset all taskers for queue "vncdexin" % ncmgr reset -queue vncdexin -taskers
-
Alternatively, you can set the environment variable NC_QUEUE:
You can also set an environment variable NC_QUEUE to the name of queue you want. Then you can execute all your nc and ncmgr commands in the context of that queue without having to use -q or -queue options.
% setenv NC_QUEUE vncdexin # Now submit a job to queue "vncdexin" % nc run sleep 61 # etc.
This way, you can easily switch between all the queues you have.