Does anyone cycle their engine daily?

We don’t and have neverf seen a need to as long as everything is configured properly. No leaked handles for example.

We used to recycle the cloverleaf every month, when the cloverleaf was on the windows box. After we move the cloverleaf to the Unix box, we never recycle it.

I once went too long without doing the quarterly maintenance downtime for Cloverleaf on AIX.  One thing to be aware of is that AIX has a max file size and once that file size is reached it will not allow any more writes to that file.  We have an interface process that hadn’t been cycled for over 1/2 year so the log file reached that max file size.  One of the threads evidently was writing data to the log file for each transaction.  Every transaction on that thread errored out because it could not write to the log file.  At that point it became evident that just because things are running smoothly doesn’t mean that I should ignore the quarterly maintenance.

And it is not abnormal for our interface processes to run without intervention once they are put into production.  We have around 80 threads on 3 sites and get paged maybe once every 2 months and most of the time those pages have to do with the sending and receiving systems, not Cloverleaf.

We cycle the process logs and smat files daily and move them to an archive, which helps us keep file sizes beneath the limit (2GB per file for our system).  We will maybe do 2-3 downtimes a year to shrink databases and clear shared memory, but in retrospect we also were switching HIS, RIS, and a bunch of other things during the last year, so that may be an inflated value.

If you’re not on Windows and keep your files below the max size you should be able to run mostly indefinitely.  Daily or weekly cycling seems excessive.

We used to, but once we added an emergency room tracking system, the enterprise couldn’t tolerate even the few minutes it took.

So we now cycle and archive SMAT & logs daily, and maybe do a clean up quarterly.

We are CIS 5.7 on AIX.  Two sites, each about the same as yours.

I cycle save the live threads weekly.  If I have to go back and find a message, it’s a little easier than looking through daily files.

If you mean reboot the system, then that’s more of a quarterly time period.  It takes at least 20 minutes to accomplish everything, so I don’t like to do it too often.

We cycle and archive the logs and smat files daily.  We’ve run on HP-UX and more recently RedHat. The only time we bring it down is to make hardware changes. If you’re on Windows, that’s a whole different story.

Weekly is fine for us, with a prefab script which is scheduled to run on windows every sunday. Works fine.

We cycle SMAT files daily, message tracking files daily, log files (we log message translation details) every four hours.

Unless there is a server (AIX) issue or change, we don’t touch the engines.  If there is a thread disconnection, Ops (24×7) instructions are to bounce the thread as a simple start point.

We have around 75 production sites.

I’m on RedHat, and I cycle my SMAT every week or two.  I have my logfiles set to autocycle at 500mb.  Configured this way, my processes run all the time unless a cycle is needed for config changes.

We are on Windows Server 2008 with version 5.7 sp2. We cycle all our prod threads and archive the files twice a day.

We have recently moved to Server 2008 and 5.7 sp2. Once we moved to the new version and server the Cycle_save script provided by Quadramed no longer works and they don’t know why. Can you help with a script?

That would be much appreciated.

Thanks,

Ron

Here is a routine we use to cycle logs, smat files, input/output files. We are running 5.6 on RedHat so your mileage may vary, but it will give you an idea of all the things we touch on. We run this under cron. It gets all of its relevant data from NetConfig so you don’t have to maintain lists of processes, threads, files, etc. It can be run as offen as appropriate for your environment.

Cheers.

Thanks very much for the response. I’ve updated the Daily_SMAT_Cycle2 file with my interface names and the CycleSMAT file with my test environment name. When I run the Daily I get the following messages:

Process command port not defined!

Process is probably not running.

I get these messages 3 times. Probably one pair for each Proc set. Any ideas of what I don’t have setup?

Regards,

Ron

The error message is telling you that a process is down when you are trying to perform an operation that requires it to be up.