Architecting, installing and maintaining your SAS environment

Is there a way to find out when LASR server crashes ?

Reply
Contributor Go
Contributor
Posts: 74

Is there a way to find out when LASR server crashes ?

Hi All,

 

You think I can schedule a simple unix script to send me an alert when LASR server crashes ??

Trusted Advisor
Posts: 1,756

Re: Is there a way to find out when LASR server crashes ?

[ Edited ]

Hello @Go,

 

excellent question. The quick answer is yes, of course, you can do almost anything you need!

The detailed answer will depend on the definition of "crash" for this particular case. With this I mean:

 

- If you want to monitor that the LASR server is providing service, you would need to monitor the TCP port where LASR is listening and/or the system process that started the sas process (LASR) with the command line as expected.

 

- If you want to monitor the LASR tables that are loaded, you can create a script of your own containing code as explained in this link http://support.sas.com/documentation/cdl/en/inmsref/67629/HTML/default/viewer.htm#p050lknh5xepngn1s2... and then you can customize it to monitor your requirements, and check your important tables if they are loaded and how.

 

If there is something else you will need to monitor, please let us know.

 

Hope it helps.

Kind regards,

Juan

SAS Employee
Posts: 518

Re: Is there a way to find out when LASR server crashes ?

@Go,

 

From my point of view, better to find out why your LASR server is crashing. The LASR server might crash in a few circumstances:

 

  • A human error, for an example someone killed the process
  • An internal exception within LASR servers
  • Third-party application killing the LASR server, such a Linux Out-Of-Memory killer

First off, this is distributed or non-distributed LASR? Depends on the answer to this question, I will provide you a different set of instructions.

 

@JuanS_OCS,

If you want to monitor that the LASR server is providing service, you would need to monitor the TCP

That's not enough because the TCP port might be in the listening state, but the LASR server unresponsive. The better way to verify if the LASR server is responding is to try to assign SASIOLA library.

Contributor Go
Contributor
Posts: 74

Re: Is there a way to find out when LASR server crashes ?

@JuanS_OCS and @alexal ..

 

Thanks for the detailed reply, ours is distributed environment...! we are still trying to find out the cause/solution ...

SAS Employee
Posts: 518

Re: Is there a way to find out when LASR server crashes ?

@Go,

 

Check that pstack is installed, by running 'pstack' at the command line.  If it is, create a file called pstack.sh, which contains:
 

#! /bin/sh
 
pstack $2 > /tmp/sastb.out

 
Make the script executable.
 
Set variables like the following in the /TKGrid/tkmpirsh.sh file on all TKGrid nodes:
 

export TKMPI_DEBUGGER=/opt/sas/pstack.sh
export TKMPI_DEBUGONEXCEPTION=1

 
Restart the LASR server. After the LASR crash, check /tmp on all the machines. One or more may have a traceback in /tmp/sastb.out (only if the LASR server throws an exception).  Send this file wherever you find it.

Contributor Go
Contributor
Posts: 74

Re: Is there a way to find out when LASR server crashes ?

SAS Employee
Posts: 518

Re: Is there a way to find out when LASR server crashes ?

@Go,

 

It appears that you have a network scanner that is running in aggressive mode. There are two ways to avoid the problem you are experiencing:

 

  • Do not run a network scanner on the LASR server
  • Start LASR server using specific TCP port range and exclude that range from scanning

P.S.: Why you didn't tell me that you have a track open? My co-worker and I just did a double amount of work :-)

Contributor Go
Contributor
Posts: 74

Re: Is there a way to find out when LASR server crashes ?

@alexal,

 

I am sorry  thats my colleague and im his brand new team mate... didnt realize you work for SAS directly, so wanted see if sas community here has other ideas... thanks for your support... we will go with sas track from here.... 

Ask a Question
Discussion stats
  • 7 replies
  • 308 views
  • 6 likes
  • 3 in conversation