TM1 Forum

I appreciate if any one could answer my following questions:
1)When I Use SaveDataAll TI function in a Process, where will be the data save? Is there any specific location? Is there any specific file-naming format for the saved data file?
2)Let suppose, if by chance I lost values under one element today? can it be restored from the previous saved data file (atleast values as of on last saved time)?
3)Can I simply copy the objects (cubes, rules, views) from one server to another?

Thanks in advance.

1)When I Use SaveDataAll TI function in a Process, where will be the data save? Is there any specific location? Is there any specific file-naming format for the saved data file?

The data will be written to the *.cub files (one for each cube) in your server's data directory. In addition, a new transaction log file (tm1s.log) will be created and the old one will be renamed with a timestamp.

2)Let suppose, if by chance I lost values under one element today? can it be restored from the previous saved data file (atleast values as of on last saved time)?

How exactly did you lose the data? The full cube can be restored from the *.cub file and individual transactions can be backed out using the transaction log (provided logging is switched on for that cube).

3)Can I simply copy the objects (cubes, rules, views) from one server to another?

You can copy the object files between data directories and they will be loaded (but only after server restart) and recognised correctly provided you also copy all the objects that these objects depend on (e.g. a cube file needs all the dimension files too, a view needs its cube etc). You do not need to update the }Cubes or }Dimensions dims, it will be done automatically.

ryan wrote:I appreciate if any one could answer my following questions:
1)When I Use SaveDataAll TI function in a Process, where will be the data save? Is there any specific location? Is there any specific file-naming format for the saved data file?

While people don't mind answering questions when you can't find the answers, that one's a little on the basic side for anyone who has made even a cursory examination of the documentation:
http://publib.boulder.ibm.com/infocente ... ry_N140458

You should have at the very minimum read the System Architecture chapter of the Operations guide (and a significant chunk of the User Guide) before trying to administer a TM1 system.

ryan wrote:2)Let suppose, if by chance I lost values under one element today? can it be restored from the previous saved data file (atleast values as of on last saved time)?

You'd need to copy the files in the data directory to another location, and load them into another session, whether a Local session or another remote session using a different name (say, your development server if you have one). You could then export the data from that session and reload it on your production server.

You then take a long, hard look at your processes to find out how something like that could happen "by chance", and make sure that it never does again.

ryan wrote:3)Can I simply copy the objects (cubes, rules, views) from one server to another?

Yes, but for safety both servers should be down at the time. Certainly you can't do a "live" promotion of the objects from (say) a development session to a production session unless you use a tool like Cubewise's Vizier.

2)Let suppose, if by chance I lost values under one element today? can it be restored from the previous saved data file (atleast values as of on last saved time)?

How exactly did you lose the data? The full cube can be restored from the *.cub file and individual transactions can be backed out using the transaction log (provided logging is switched on for that cube).

In a dimension there was an element 'Fixed Costs' (weight:1,Level 0) which was carrying its own values entered by business people. Today we enhanced the Level to 1 and added two children Fixed Cost(which is new element) and Misc(already existing element which is having values).

Once we populated Fixed Costs into the template, Fixed Costs element values are overridden by Misc (since Fixed Cost is new element and doesn't have values). Once we noticed the problem, we reverted the level of Fixed Costs from 1 to 0. But the worst happened as it lost all the business entered values.

So we are thinking atleast to restore yesterday saved data from that cube. We want to restore only that particulars elements older existing values.

What we have planned is:
-If we know the steps to restore or copy the cube to test environment using the production copy, once we are confident that values are available then we dump the values to .csv file from test environment and load it to production environment for that element.

Thanks in advance.

ryan wrote:

2)Let suppose, if by chance I lost values under one element today? can it be restored from the previous saved data file (atleast values as of on last saved time)?

How exactly did you lose the data? The full cube can be restored from the *.cub file and individual transactions can be backed out using the transaction log (provided logging is switched on for that cube).

In a dimension there was an element 'Fixed Costs' (weight:1,Level 0)

Level 0 maybe, weight 1, no. "Weights" only exist when an element is part of a consolidation, and in respect of that specific consolidation. An element which is part of multiple consolidations can have a different weight in each consolidation. An element which is not part of any consolidation, or is looked at outside of a consolidation, has no weight at all.

Again, it's useful to read through the manuals or invest in training to get the basic concepts before deploying a TM1 application operationally.

ryan wrote:which was carrying its own values entered by business people. Today we enhanced the Level to 1 and added two children Fixed Cost(which is new element) and Misc(already existing element which is having values).

Once we populated Fixed Costs into the template, Fixed Costs element values are overridden by Misc (since Fixed Cost is new element and doesn't have values). Once we noticed the problem, we reverted the level of Fixed Costs from 1 to 0. But the worst happened as it lost all the business entered values.

That's hardly "the worst" when you understand what a consolidation is. For numeric elements, it is always a calculated value. If you change an N level element to a consolidation, you lose any existing values, period. Which, I'll admit, should be shown as a boldfaced warning in the relevant section of the Developers' Guide but isn't; the author probably assumed that it was too obvious, though in reality it may not be to a completely new user.

ryan wrote:So we are thinking atleast to restore yesterday saved data from that cube. We want to restore only that particulars elements older existing values.

What we have planned is:
-If we know the steps to restore or copy the cube to test environment using the production copy, once we are confident that values are available then we dump the values to .csv file from test environment and load it to production environment for that element.

You're overlooking a potentially simpler option, depending on how long you've been running your system, how far back you have log files for, and whether you have logging turned on on the cube (which, in most cases, you should, at least for entries which aren't bulk-uploaded by TI such as the ones you're describing here). You could search the transaction logs for entries which had been made to that particular element (consult the manuals for "View Transaction Log"), copy them, save them to a data file then re-load them using TI.

Alan Kirk wrote:You're overlooking a potentially simpler option, depending on how long you've been running your system, how far back you have log files for, and whether you have logging turned on on the cube (which, in most cases, you should, at least for entries which aren't bulk-uploaded by TI such as the ones you're describing here). You could search the transaction logs for entries which had been made to that particular element (consult the manuals for "View Transaction Log"), copy them, save them to a data file then re-load them using TI.

That sounds a lot riskier/harder (due to all of the "ifs" required) than unzipping a backup, deleting rux and cho files (for speedier startup), starting it up on dev, exporting the lost data to a txt file, and using a trivial TI process to reimport. Let's just say that in my early days I wasn't a stranger to "making a quick change on prod" to poor effect....

Matt

mattgoff wrote:

Alan Kirk wrote:You're overlooking a potentially simpler option, depending on how long you've been running your system, how far back you have log files for, and whether you have logging turned on on the cube (which, in most cases, you should, at least for entries which aren't bulk-uploaded by TI such as the ones you're describing here). You could search the transaction logs for entries which had been made to that particular element (consult the manuals for "View Transaction Log"), copy them, save them to a data file then re-load them using TI.

That sounds a lot riskier/harder (due to all of the "ifs" required) than unzipping a backup, deleting rux and cho files (for speedier startup), starting it up on dev, exporting the lost data to a txt file, and using a trivial TI process to reimport. Let's just say that in my early days I wasn't a stranger to "making a quick change on prod" to poor effect....

The "ifs" were the point in this case. It sounds like this is a very new setup, which means that there probably aren't months or years of log files. Logging on data cubes is on by default and they would have had to know how / when to turn it off for the values not to be there especially given that the OP stated that the values had been input by end users rather than TI. In short, everything they need is probably in the logs. And given that the guy's first question was essentially "what is a data directory", I'd say that this side of taking a training course there would be a HUGELY greater chance of screwing something up along the way by following your approach than by sticking to what is much closer to an end user interface like the log files extract.