Migrate TM1 DB from Windows 08R2 to Linux RedHat 6.5

[wilsonmax]

Recently, I've installed TM1 10.2 on Linux RedHat 6.5
I tried to migrate the TM1 Models from an existing TM1 10.1.0.1 in Windows Server 08R2 by copying the whole data folder, since it is file-based.
After that, setup and started the TM1 instance in Configuration in the Linux Server
It could be started and my TM1 Client could connect to it.
However, after login, I found that the system could not recognize all the files.
It was just showing some of the TIs instead of the complete system.

May I know if there is any extra steps/config need to be done for Linux?

Remarks:
I tried the same migration and connecting method for Windows-to-Windows, and it was successful.

[moby91]

However, after login, I found that the system could not recognize all the files.
It was just showing some of the TIs instead of the complete system.

The names of the object files the system did not recognize, did they start with a capital letter ?

If so, this is a known issue:
Regarding file names Windows is key insensitive, thus it does not matter if file names are using capital letters or not, eg you can access the same TI process file by "TIprocess.pro" or by "tiprocess.pro".
But Unix and Linux are key sensitive, thus it does matter if file names are using capital letters or not, eg the file names "TIprocess.pro" and "tiprocess.pro" label two different TI processes.


Find attached the TXT-file "utol.txt".

Copy it to your Linux box into your TM1 server's database directory, rename it into utol.exe or utol.sh and make it executable be setting the executable-bit.
If you run the utol script it will convert all filenames and directory names into lowercase.


PS
You might contact IBM Cognos Support and ask for the document "TM1 Platform Portability Technical Note" covering the differences between running TM1 on Windows vs running TM1 on Unix or Linux platforms.

[David Usherwood]

IBM provide the tm1xfer utility to convert server directories between Windows and the Unix family. I used it to migrate to Ubuntu and it worked well.

[jim wood]

David,

Is that available on the IBM support site for download or do you need to contact thm for it?

Jim.

[wilsonmax]

However, after login, I found that the system could not recognize all the files.
It was just showing some of the TIs instead of the complete system.

The names of the object files the system did not recognize, did they start with a capital letter ?

If so, this is a known issue:
Regarding file names Windows is key insensitive, thus it does not matter if file names are using capital letters or not, eg you can access the same TI process file by "TIprocess.pro" or by "tiprocess.pro".
But Unix and Linux are key sensitive, thus it does matter if file names are using capital letters or not, eg the file names "TIprocess.pro" and "tiprocess.pro" label two different TI processes.


Find attached the TXT-file "utol.txt".

Copy it to your Linux box into your TM1 server's database directory, rename it into utol.exe or utol.sh and make it executable be setting the executable-bit.
If you run the utol script it will convert all filenames and directory names into lowercase.


PS
You might contact IBM Cognos Support and ask for the document "TM1 Platform Portability Technical Note" covering the differences between running TM1 on Windows vs running TM1 on Unix or Linux platforms.

Thank you for your reply and the tool in the attachment!
I tried both method yesterday: i) convert the filename using utol.sh; ii) tm1xfer utility tool and they are OK

[wilsonmax]

IBM provide the tm1xfer utility to convert server directories between Windows and the Unix family. I used it to migrate to Ubuntu and it worked well.

Thank you, David!

It worked well~

[wilsonmax]

David,

Is that available on the IBM support site for download or do you need to contact thm for it?

Jim.

I found that the tm1xfer utility is under the directory of tm1, e.g.: C:\Program Files\ibm\cognos\tm1_64

Reference:
How to use tm1xfer
http://pic.dhe.ibm.com/infocenter/cx/v1 ... 1xfer.html

[David Usherwood]

@Jim - I took the Red Hat version and frigged it unmercifully until it came up. From memory the frigging was mainly around telling IBM's version checking routines that they were looking at the right versions of modules (libncurses and libz.so), plus using the then-undocumented cfg setting IPAddress=T. I don't expect Support to be accepting any calls

[jim wood]

Sounds like fun David. I might give it a try on a VM.