Server Hangs when running process

[mce]

Hi,

When I execute a process that transfers data from an external file to a TM1 cube, it hangs and does not do anything and stops anybody login or use TM1. Then I have to kill the TM1 service and restart.
I could not find anything in the tm1smsg.log and tm1s.log files about this problem and it does not create an error log file for the process as well. I simplified the process by removing text data items, leaving only numeric data items, but it did not help.

What might be happening? How can I diagnose the problem?

Regards,

[jim wood]

Firstly I would try upgrading up to 9.1 sp4 at the earliest to see if you still have the problem,

Jim.

[mce]

I am afraid, upgrade is not possible for this problem.

[mce]

Assuming there is something wrong with eiter the data source or with the process settings,
- why is just a data upload process for one cube capable of crashing the whole TM1 server?
- why do all users have to stop using TM1 just because I am uploading a small set of data for just one cube?
- am I missing anything here or is this the way TM1 is delt with?

[rkaif]

The SQL statement which you are using within the process may need optimization. Have you tried running the same SQL statement within the DB itself (any other IDE than TI Process like SQL plus if you are using Oracle).

What are you your Hardware specs? number of Processors? Total RAM? and the RAM still available?
Which back-end database you are using?
Is this problem with a specific TI Process or with every other process?

[Alan Kirk]

Firstly I would try upgrading up to 9.1 sp4 at the earliest to see if you still have the problem,

I wouldn't. The OP is on 32 bit. 9.1 is so memory hungry compared to earlier versions that I wouldn't think of going to 9.1 or higher unless you have either (a) a very small model or (b) have done extensive testing to ensure that you're not going to be bouncing your head against the memory ceiling on a daily basis. There's a reason that I upgraded to 9.0 and not higher.

[Alan Kirk]

The SQL statement which you are using within the process may need optimization. Have you tried running the same SQL statement within the DB itself (any other IDE than TI Process like SQL plus if you are using Oracle).

The OP said that the data source was a file, not an ODBC connection.

When I execute a process that transfers data from an external file to a TM1 cube,

[Alan Kirk]

Assuming there is something wrong with eiter the data source or with the process settings,
- why is just a data upload process for one cube capable of crashing the whole TM1 server?

It doesn't appear to be crashing it but rather hanging. Given that the data source is a file connection shouldn't be an issue; the process will either find the file or it won't and if it doesn't it would just abort. That makes it unlikely that the data source is the actual problem.

- why do all users have to stop using TM1 just because I am uploading a small set of data for just one cube?

Because TM1 can't know in advance that you're updating only one cube. It can't even know in advance which cube you are updating; for example for all TM1 knows you may have code in there that retrieves the cube name from the data source or from a lookup of a control cube. Accordingly it (in versions like 9.0 anyway) locks the server so that the update can be done without anyone getting in the way of that or them getting out of date data.

When this sort of thing happens it's usually because there's an endless loop in the TI code; a While statement which doesn't have an increment, for example. It may be an idea to post the process file so that we can check it for that; easier than speculating.

[lotsaram]

Hi,

When I execute a process that transfers data from an external file to a TM1 cube, it hangs and does not do anything and stops anybody login or use TM1. Then I have to kill the TM1 service and restart.
I could not find anything in the tm1smsg.log and tm1s.log files about this problem and it does not create an error log file for the process as well. I simplified the process by removing text data items, leaving only numeric data items, but it did not help.

What might be happening? How can I diagnose the problem?

Regards,

What everyone seems to have missed except Alan in the last post is that mce is running version 9.0. I'll make it more explicit for mce's benefit - in 9.0 there is no object locking model only server wide locks. Therefore it is perfectly normal (and expected) for all logged on users to experience a "hang" for the duration of a TI process running and for no new users to be able to log in while a process runs.

Quite likely there is no "problem" at all other than mce is executing a long running process without realising that hanging users and preventing new connections is an expected outcome from doing this!

Does the process really "hang" (that is fail to finish) or would it actually complete successfully if allowed to run? How large is the source text file and how many records does it contain? If the file is very large then quite a long duration for the process would be expected.

If the process does complete but takes some time then the solution is clear - only run the process after hours so users don't get locked out. On the other hand if the process genuinely does fail to complete in an infinite time then chances are there is a problem with some while loop logic somewhere in the process so analysing any while loops would be a good place to start.

[mce]

Thanks for the responses.
- My data source contains numeric and text data items and has less than 1000 records.
- I initially used an ODBC connection to the Exel file to load data, but after having this issue I converted it into a text file and tried the same, and had the same issue.
- When I ignored the text formatted data items from process, it was running successfully and quickly.
- I found out that when I try typing data to one of those text formatted cells in the cube, the server hangs again. This was explaining that the issue was not to do with the process or data source.
- I checked the rules for that cube, and removed all rule statements that was using the text formatted elements (although they were all right) that was causing the hang. Then tried typing again and it did not help.
- I am now trying to investigate why the server hangs when I type data in a text formatted cell, although it was working all fine previously and no change has been done to that cube.
- Do you have any ideas that might help me in this?

Regards,

[mce]

More notes,
- There are multiple text formatted data items in the measure dimension of this cube. Most of them working fine when typing data. But only few of them causes the server hang when typing data.
- I have the same cube in test server having the same rules and same cube structure. It does not have such an issue.
- Do you think it should fix the problem if I delete and re-create the cube?

[mce]

The issue is disappeared when I delete the following feeder that was related to the string element:

Code: Select all

['Live','Base','Name']=>	['Live','Name'];

The rule for that feeder was the one below:

Code: Select all

['Live','Name']=S:IF(ATTRS('Oth_Periods',!Oth_Periods,'ActPlan')@='Plan', DB('MyCube',!Oth_Scenarios,'Base',!Dim_Divisions,!Dim_Positions,'Name'),STET);

I am re-working the feeder statement.
I did not know that such a feeder would cause the server hang and it did not until I try changing the values in string source cells.
Is there a rule in TM1 that we should not have feeders targeting input string cells haivng the same string cells as the source?

[lotsaram]

Your feeder statement has a circular reference and is recursively feeding itself!

Once a numeric feeder has fired it is switched off however string feeders are different they will continue to fire on every source string change. I would not have thought that a circular reference in a string server would cause a server to hang (the rule equivalent of a never ending while loop maybe?...) but perhaps this is your problem.

Edit: on a second look the area statement of the feeder might be muddled. Are "Live" and "Base" both members of the same dimension? If so you definitely don't want Live to feed itself. The correct (simplified) feeder would be ['Base'] => ['Live']

[mce]

Thanks for your comments. I will keep in mind not to let strings feed themselves to avoid circular issues.
Base and Live are in different dimensions.