What is the best way to deploy from dev to prod?

[Andy]

Dear all,

I have one physical server where two TM1 server instances are running. The 'developing' TM1 server instance and a 'production' TM1 server instance.

What is the best way to synchronize these two ones or
what is the best way to deploy everything or only changes from the 'Developing' to the 'Production' TM1 Server instance?

I tried TM1 replication, but there it is not possible to synchronize/replicate processes and I understand that it is not working without problems (http://www.tm1forum.com/viewtopic.php?f=3&t=3870)
I also tried 'TM1 Simple Replicate'. That is working fijne. But it takes long and I cannot see any changes (between the instances). So, I have to synchronize everything (but this button is missing).

How do you do the deployment from one TM1 server instance to the other? What is the best way to keep them synchronized.

PS:
It should be free of charge

Thanks in advanced.

[Ken Vuong]

Hi Andy,

I dont think there is any easier way and it's just the case of downing your Prod server, back it up, copy the delta files over from dev using the date stamp as quide.
I guess you will also need to know the extension of the files .PRO - process, .RUX - rules, .CUB - cubes, .SUB - subset...etc

I have to do this quite alot of time and usually during lunch time or out of hours

I wouldn't go anywhere near Replication/Synchronisation.
Hope this helps

Ken

[Andy]

Hi Ken,

thanks for your answer.

I did copy the datafiles (*.rux, *.pro, etc) from one Tm1 Server instance to the other.
And then I strated the instances again.

BUT:
The start from the instance takes more then one hour and
we are using plannings applications on the Application portal and these application are gone every time.
No physically, (there still in the mapD:\Program Files\ibm\cognos\tm1_64\webapps\pmpsvc\WEB-INF\applications) but there not anymore on the Application server.

Therefore I am looking for another way to deploy/synchronize the application.

[Steve Vincent]

It's a common question & one IBM still hasn't come up with a decent solution for, probably because it doesn't look sexy enough in a sales meeting

Replication is at best, flakey. It's main issue it the reliance upon logged changes. If you manipulate a lot of data then replication could take longer than it's taking to restart your services. As you've already pointed out it also only deals with data and objects, TIs, applications etc aren't part of that suite.

For data replication i ended up writing my own TIs. It's handling cubes & dims just fine, but i'm ignoring subsets & views because both require a service to be offline in order to copy across correctly. Likewise rules, TIs etc aren't covered in that either.

I even tried using SVN options, these seem to handle binary files OK (rules, TIs, Dims) but failed to handle cubes correctly. The main issue was on time stamping of the objects - TM1 will update the time on a lot of stuff even if it's not been changed, and SVN just didn't do what we needed it to without being a major admin task.

So, as it stands i'm still using the age old method i used to update websites with 20 years ago - manually copy & paste the changed items i've kept a record of to the production service. IBM, if you are listening, this is the type of stuff that *really* benefits a business...

[LanceTylor]

Hi Andy,

I am in a similar situation. I would like to find an easy way to promote TM1 Objects and applications without any errors. Below is the recommended approach from IBM.

1. This is the process that IBM outlined for 9.5.
https://www-304.ibm.com/support/docview ... wg21423395

2. This is the process that IBM outlined for 10.1.1 - Using performance modeller to transfer objects and applications (transferring data section)
http://public.dhe.ibm.com/software/data ... mdl_ug.pdf

Currently we are using TM1 10.1.1 and all the development is done within Architect and a couple applications are deployed through Performance Modeller. My first attempt at promoting changes from DEV to TEST was using the first process (9.5) which included transferring the whole data directory. This worked the first time but when I attempted it again (after user testing and development changes) it didn’t work and I was forced to revert back to the backup and manually update changes. I still don’t know exactly what the problem was but it would think it relates to application references and the cube views it creates?

I have yet to try the latest and greatest method which is through performance modeller but to be honest I am a bit skeptical.

Has anyone else tried this approaches and had a positive experience?

Lance

[George Regateiro]

We have started to use a git system to track changes to the rules and processes, but the actual moves are still the manual method. If something like a process needs to go in without a downtime we will manually create it and then I will copy the entire object over the weekend. This is not ideal, but it is what you live with.

So far I have not been impressed with the performance of the Performance Modeler migrations. It leaves alot to be desired.

[Harvey]

I have a blog article on this which you might find interesting. It covers WinMerge, SVN and various other issues. You can find it here.