Workload Automation Autosys Edition (AE) Office Hours Transcript (12/11)

[Chris_Stallone]

Jim Spath :

hello

 

 

Jim Spath :

I'm starting to use Java for utility programs, via a tip  Mike Woods.  Does anyone have best practices for coding against the API?

 

 

Chris Stallone :

Morning Jim

 

 

Jim Spath :

(I plan to post some of what I write in the community forum but am prohibited  wholesale code sharing due to corporate rules).

 

 

Michael Woods :

@Jim - I wouldn't recommned that you bombard the application server with constant reqeusts, just as you would't run autorep -j ALL constantly.

 

 

Michael Woods :

@Jim, I guess a good rule of thumb is us the same common sense that you would use for the CLI.

 

 

Jim Spath :

@Mike fair enough. What I am writing now will replace sendevent commands with java code, so these will run at the same frequency as the current jobs.

 

 

Michael Woods :

@Jim, OK.  I did dreceive your email, but haven't had a chance to review it yet.  Will get back to you later today or tomorrow the latest.

 

 

Francine: Good morning...a member of our scheduling team is trying to create a calendar to schedule a job to run bi-weekly, is there a keyword that can be used for example to schedule a job on the 11th of the month and run that job every 14 days thereafter?

 

 

Jim Spath :

One example I have that works on UNIX but not directly on Windows is: sendevent -E SET_GLOBAL -G VDATE=`date +20%y%m%d%H%M%S`

 

 

Jim Spath :

@Francine for bi-weekkly jobs, you could create a calendar. Since there is not a "start on an even or odd week" rule, extended calendars are not helpful.

 

 

Jim Spath :

@Francine another trick is to add a sendevent that re-schedules the next run as part of the box. But that's a pain.

 

 

Francine :

ok, thanks Jim

 

 

Jim Spath :

On another topic, we've started to leverage the "send email on failure" feature.  Any gotchas?

 

 

Jim Spath :

(such as returned undeliverable, spam rules, etc.)

 

 

Michael Woods :

@Francine,  the WCC create calendar has a function where you can generate every x number days or weeks.  I'm running a test when my image starts up

 

 

Jim Spath :

@Mike - that is a standard calendar (that would need to be maintained) right?

 

 

Francine :

@Jim - yes, that's my understanding...we're trying to aviod maintenance

 

 

Michael Woods :

@Francine, yes it is a standard calendar.

 

 

Jim Spath :

On a similar topic, do others keep calendars in sync between production and non-production (assuming both tiers exist)? If so, how?

 

 

Michael Woods :

@Jim, The only gotcha for email I have seen is regarding the  address.  If it wasn't a valid one some mail servers may block it  being sent.

 

 

Michael Woods :

@Jim, there is a setting in the config file if you need or want to change it.

 

 

Jim Spath:

@Mike - yes we plan to use a common service address, not mine or someone else on the team as the "", or "reply-to".  I'm wondering if others do this or use personal/corporate email and how they maintain these. In UNIX command line jobs, the /var/spool/mail/autosys box gets returns, and I rarely look at them.

 

 

Jim Spath :

(I should go on mute and let others speak :-) )

 

 

Lenn Thompson - CA :

@Jim No way! We really appreciate your questions and you sharing your experience and expertise!

 

 

Michael Woods :

@Jim,  I have mostly a service or group dlist being used.  The other is a doNotReply type if nobody will be checking

 

 

Francine :

Is there a configuration option to send an email only on the second job failure event  - perhaps within a certain time frame?

 

 

Jim Spath :

Here is another topic - do others export JIL on a routine basis?  We've done it daily, for years, and it has "saved our bacon" more than once.

 

 

Michael Woods :

@Francine, the current options are on failure or job termination.

 

 

Francine :

@Mike - ok, thanks

 

 

Jim Spath :

@Francine - you should post that as an "Idea" on the Community site...

 

 

Francine :

@Jim - Good idea,...will do

 

 

Lenn Thompson - CA :

It looks like a couple of you aren't members of the CA Workload Automation Community. I'd encourage you to join! https://communities.ca.com/community/ca-workload-automation

 

 

Jim Spath :

Here is another pain point - upgrading agents.  We have 3 or 4 systems running 4.5 agents, most run 11.3, while the core scheduler version is 11.3.6 (INC1).  I guess this is not a question, though.

 

 

Michael Woods:

@Jim,  I'm sure the others on here have seen different things, but yes.  A regular backup of jil and other objects like machines is not a bad idea and can be quicker that trying to get a DB restored for example.

 

 

Jim Spath :

@Mike I back up calendars, including cycles, weekly.  Machines, jobs, and global variable state get dumped daily.

 

 

Jim Spath :

For those in the " CA Workload Automation Community", how do you get notifications of posts/questions/events - (1) email, (2) browse the site, or (3) RSS?

 

 

Francine :

@Mike - is CA working any agent related tools for deployment upgrades etc?

 

 

Joe Neumann :

There are customers who run monthly jobs that extract the JIL data Mike notes. They have a lot of set job flows which change little. If you have a more dynamic instance, you should get copies more often.

 

 

Michael Woods :

@Francine,  There is a project for managing agents already underway.  I do not have a list of of all features or dates at the moment and would need to get Product Management  to get an definitive answer.

 

 

Francine :

@Mike - ok, thanks for the info

 

 

Jim Spath :

One more... how do others measure performance of their AutoSys instances?  We occasionally get the "it's slow" comment, and it gets trickier to find bottlenecks as the application evolves.

 

 

Michael Woods :

@Jim,  I guess it depends on what is 'slow'.  The traditional lag or latency in scheduling events or running a command or WCC.

 

 

Jim Spath :

Night operators complain about WCC. Few end users notice if jobs start seconds later than the schedule says.  One of my pet peeves is jobs set up to run once a minute, when this requirement should be handled by an application, not AutoSys.

 

 

Michael Woods :

@Jim,  There are some tweaks that can be made to WCC to improve performance.  The other part to consider is checking the responses  AE itself to narrow down the bottleneck.

 

 

Jim Spath :

@Mike we keep a lot of history (I have several tickets open about history not being purged) which can trigger this. Some SQL tuning helped.

 

 

Jim Spath :

Thanks everybody. See you on the intertubes!

 

 

Lenn Thompson - CA :

We have a couple minutes left in our session. If you still have questions, please visit the CA WA Community and ask them there. we will get the same experts in there to answer them.

 

 

Lenn Thompson - CA :

Also, we will be publishing a transcript of today's session to the community this afternoon.

 

 

Michael Woods :

@Jim, I don't remember if they updated the 11.3.5 tuning guide for WCC for 11.3.6, but it should be applicaable for both.  If you are running on Linux, the entropy method can come into play when working with EEM.

 

 

Jim Spath :

@Mike - almost all on Windows not, save the DB (Oracle on AIX).

 

 

Jim Spath :

/not/now/

[Lenn Thompson]

Thanks for posting, Chris!