Frequently Asked Questions and Troubleshooting Tips

I'm doing an installation and configuration in a Windows environment - can I use PowerShell?

PowerShell is not supported; we strongly recommend not using PowerShell.

How do I contact Altair Engineering to get additional support, report a bug, or request a feature?

You can contact Altair Support at: https://community.altair.com/community.

Why can't I access Monitor's historical license usage through Accelerator?

Accelerator ships with a version of Monitor that is licensed to monitor current license activity only. This edition is called LMS (Monitor Small). To access Monitor's historical license usage information, you must have the full version of Monitor.

What do I do in the event of a server failover or crash?

You can find a checklist for system recovery on the System Recovery page. You can find this address with the command:

nc cmd vovbrowser -url /cgi/sysrecovery.cgi

Where is the policy.tcl file? What about the taskers.tcl file? The resources.tcl and security.tcl files?

All .tcl configuration files for Accelerator are located at $VOVDIR/../../vnc/vnc.swd

How do I enable the retrace of more than 400 jobs at a time?

The limit to how many jobs can be run/retraced at any given moment is defined by the maxNormalClients config variable. To change the variable, you can use the command:

vtk_server_config "maxNormalClients" maxnumberofjobs

How do I receive email notifications on job completion?

To receive automatic notification of major FlowTracer and Accelerator events, you should use the vovnotifyd daemon.

How do I track the memory usage of taskers?

VOV automatically keeps track of tasker memory usage. vovtasker keeps logs of 1 minute, 5 minute, and 10 minute load averages of the machines where taskers are running on. The tasker reports are available on the Tasker Load page. The Accelerator URL can be found with the command:

nc cmd vovbrowser -url /cgi/taskerload.cgi

Why are my jobs taking so long?

There are multiple reasons why FlowTracer jobs may be retracing slowly. Fortunately, Accelerator produces reports to help diagnose any problems. Read about available reports at Resource Plots.

How do I change to another version of Accelerator?

To upgrade Accelerator software, refer to Upgrade Accelerator.

How do I access information on license usage?

Accelerator does not have this functionality. This functionality belongs to Monitor. If you have Monitor installed and fully working, you can display license information at on the FTLM page.

Why are my licenses not fully utilized? I'm sure they're completely booked.

Your licenses are not fully utilized probably because they are not overbooked.

Essentially, the jobs being run do not use a license 100% of the time. Because there are jobs booked for licenses 100% of the time, there will be times where licenses are not utilized. This is because one or more jobs will still be running, but be done with the license that was booked. To rectify this, jobs are queued for more than 100% of the licenses, allowing another job to start and utilize the open license.

How do I share licenses between jobs in queue?

Read more on license at License Sharing Support.

My tasker is sick! What do I do?

Your tasker is sick because it has not sent out a heartbeat for at least 3 minutes. This may mean your tasker has crashed or disconnected. Once you have identified a sick tasker, you can proceed to troubleshoot it to fix the problem.

This list may be helpful:

Check to make sure the machine itself is healthy. Make sure it is running, connected to the network, and not jammed.
Check to see if vovtasker or vovtaskerroot is still running. If it isn't, then the tasker program itself has crashed. You should restart the tasker program with:
```
vovtaskermgr start
```
Check to see if vovtasker or vovtaskerroot is stuck. If it is, Linux commands such as strace and pstack should provide you with enough information to fix it.

My tasker is healthy, but all jobs sent to the tasker come out failed. What is going on?

Your tasker is what is called a black hole. It appears healthy, but is in fact unable to execute jobs. There is functionality to enable automatic detection of black holes in the Black Hole Detection page.

When a black hole is found, it would be prudent to send a simple job such as cp or sleep to the tasker to confirm its black hole state.

I want to give a different amounts of resources to different sites. How can I do that?

FairShare is a mechanism to allocate CPU cycles among groups and user according to a policy. This would be your best bet.

How do I limit a resource for a particular user?

Although it is not recommended, information on limiting users can be found on the Limit Users page.

My job was killed because it failed to start within 1m00s!

This can be caused by a bad NFS mount point, or an automounter that is so overloaded, that it fails to mount the run directory for the job in under a minute. Although this is a hardware problem, there is a workaround by changing the variable VOV_MAX_WAIT_NO_START to a value over 1 minute.

How do I setup prioritized licence usage?

For example, to use the licence FOO_BAR_A first, then the licence FOO_BAR_B second, use:

vtk_resourcemap_set FB-lic UNLIMITED "Licence:FOO_BAR_A  OR Licence:FOO_BAR_B"

In the jobclass. To set it in a resource map, use:

set VOV_JOB_DESC(resources) "Licence:FOO_BAR_A  OR Licence:FOO_BAR_B"

Why am I missing the plots when I look at a resource or license report?

Probably, what is causing the plots to be missing is a name resolution issue. To fix this, make sure VOV_HOST_HTTP_NAME is set correctly. If all else fails, set this to the host's IP address, not network name. To update a running server, you must use the command:

vtk_server_setenv VOV_HOST_HTTP_NAME XXX

vovresourced is not starting, says 'Failed to source'' too many resources'!

Most likely, you have exceeded the limit for resource maps in use. To raise this limit, change the maxResMap value in policy.tcl.

How do I ensure that a tool is preemptable robustly?

Sometimes a tool will crash when preempted. To test whether this is Altair, or the tool vendor, try and run the tool without Altair binaries (pure UNIX code) and see if the tool still crashes. The steps to do this are as follows:

Start the EDA tool(s) which you wish to test.
Use the UNIX command ps to find the PID of the EDA tool(s):
```
% ps | grep firefox
PID TTY          TIME CMD
349 ?        00:24:19 firefox
```
Send TSTP and CONT signals 10 seconds apart repeatedly. Try this in your shell:
```
% kill -TSTP 349 ; sleep 10 ; kill -CONT 349 ; sleep 10 ; kill -TSTP 349 ; (etc...)
```

Following these steps, if the tool crashes, then the problem is independent of Altair, as not a single line of Altair code was executed.

I set a configuration in the policy.tcl file, but it is not taking effect!

Most likely, the file has not been read yet. Try a:

% nc cmd vovproject sanity

I have a lot of log files, how can I remove the older files?

An easy way to remove files that are over 60 days old is using the vovcleanup command:

% nc cmd vovcleanup -proj

Note: When preemption is heavily used, log files tend to build up.

How do I test a policy change before releasing it to production?

To test policy changes you can use the soft release mechanism. Here's a summary:

Create a test queue.
Set up the test to use files from the repository of the master queue.
Test a hot file in a sandbox, identify and fix the errors before releasing it to the production domain.

I am upgrading the software - how do I suspend Accelerator from dispatching jobs?

Typically, to minimize the impact of upgrading the overhauling the system, the vovserver is stopped from dispatching new jobs, while jobs that are running are allowed to complete on the vovtasker. There is more than one way to do this: Cold Upgrade, Hot Upgrade and Rolling Hot Upgrade. For more information and instructions, refer to Upgrade Accelerator.