Invalid dimension, even straight out of table header

[nu2QlikBM]

Hi , I am new to the platform and forum too. Although I have already  learnt a lot from postings on the forum. I have a quirky issue that has recently sprung up on my app. It has been showing connection error sporadically. I checked by opening a data free version , and it seems to  indicate some dimensions are  INVALID. Initially I thought it could be because some of them were calculated dimensions and added to Master dimension sections. To mitigate, I have  even duplicated and repopulated versions free- from -Master dimensions. And now, even regular (as imported from table ) headers/fields are showing up as  Invalid dimensions on the app- a week after creation and validation! This is leading to similar , frustrating connection  error. 

I am on Enterprise module, the apps  are in adhoc space, and have not been published.  I have emptied chrome cache multiple times.

Please advise.

[stevedark]

Hi @nu2QlikBM 

I am not sure where you are seeing this connection error. Could you possibly post a screen grab showing the error?

Invalid dimensions happen when the data model changes in an app and a field that was there when the object was built is no longer present.

Does the load script specify all fields in the load script or do you have SELECT * in the script. If you do, is it possible that the structure of the underlying table has changed?

Steve

[nu2QlikBM]

Hi Steve,  Please find the screenshots attached. The connection error pops up the moment I try to open any one of the  two sheets on my app. 

As for the data structure being altered. I am checking that with out database management  team. I must mention the app does not read directly from database (a different story!) , but I "attach" the excel download manually and refresh the tool. First time the error showed up was after one such data upload. However, I do not think that a structural change  is likely  because I have an excel Power query  tool built around it, which  has been reading the data error free.  Also as you will see in one of the SS, the script identifies the header (marked in red) , but shows up as invalid dimension in the app.

Thanks

Mithun

[stevedark · Accepted solution]

Hi @nu2QlikBM 

Let's try and get the connection error sorted first, as that could make the other issue go away.

Unfortunately, that error can occur for a whole number of reasons.

The most likely is some kind of connection disruption between the client and the server. Are you using a VPN to access the server at the moment? Are you confident that connectivity between you and the server is solid? One way of removing this kind of issue is to remote desktop onto the Sense server itself and try working from there.

I have seen issues when the same user account is accessing the server more than once concurrently. Are you sharing your login with anyone? It can also occur if the same user account is used for the services as is being used as a client - do you have a separate service account used for reloads or is that a user account?

It looks like you have used the data manager to bring data in. I have seen issues with this. You could switch to using the data load editor, which gives you better control of your load. Unlock the script and remove anything extraneous.

The size of the app can also have a bearing. Can you try adding a WHERE statement to your load to only pull through a small subset of the data and see if it happens still. If you have a date in the source data you could try WHERE Date >= today()-7 or you could get a subset based on the first letter of a field by going WHERE left(Name, 1) = 'v'.

Check the data model in the Data Model Viewer, ensure that all tables associate (have a line between them) and the keys used are sensible.

I see that you are on a very old version of Sense. Upgrading to a later version may solve the problem.

Hope that something there helps. If not you could try speaking to your Qlik partner, or using support@qlik.com

Please post back here and let me (and others) know how you get on.

Cheers,

Steve

[nu2QlikBM]

Hi @stevedark  

Thanks for the detailed responses. Let me work through each of those and get back to the forum to mark the post , if it were to solve the issue.

I have tried both with and without VPN connections. I have checked for WebSockets as well. I will pass on your suggestions to our IT and Qlik management team prior with regard to the remote access. 

Thanks, Mithun

[nu2QlikBM]

Hi @stevedark ,

You were right. The load through  data manager is buggy, and recently  it had brought in an extraneous column! But it had been fine for months earlier. The loading from Data load editor  seems not to have that problem (at least for now). However, it leaves the script for the prior LOAD, and thus offsets the data, unless I manually deelte  each prior LOAD script. 

Is that the only remedy? Again, pardon my naivete , but  I had been earlier advised (internally) to not load through editor nor to UNLOCK the editor to be able to easily navigate without significant scripting (not my background) requirement. 

[stevedark]

Hi @nu2QlikBM 

Personally I never use the Data Manager, but then I am a developer by trade and prefer looking at lines of source code rather than floating bubbles.

I always recommend people use the Data Load Editor, but then I am biased.

My reasoning for this recommendation is partly that the Data Manager can take you a certain amount of the way with your load, but the moment you need to do something complex you have nowhere to turn. You just as well be using another tool that doesn't have a powerful scripting language!

The other reason is that I have seen instances of the Data Manager going wrong. The one that caused a client of mine lots of grief was that the Data Manager was loading from a cached data set, even when the underlying data had changed. This may have been a bug in a previous version of Sense (it was a while back), but it put me right off. I have seen a number of instances of issues caused by the Data Manager on Community, and I am sure I would see more if I actually used it, or encouraged my customers to do so.

When I run training courses I always tell people not to click the Data Manager, and often when someone's hand goes up as they have a problem it is because they have clicked the button I told them not to! In fairness though, this is in part that it is a mish mash of Data Load Editor and Data Manager.

The Data Manager demos well and may make Qlik look good when presented from a podium. The whole drag and drop thing gives the impression that handling data is easy, and it really isn't sometimes. Better, in my opinion, to accept that and work with a toolset that gives you a fighting chance of dealing with it (with some help from the Community if required).

If you unlock the script and only edit the script using the Data Load Editor from here on in I believe you will have an easier ride.

Please feel free to direct your colleague to me if they wish to have a discussion about it on Community! 😃

Cheers,

Steve

[nu2QlikBM]

Hi @stevedark ,

Thanks for those valuable insights. I think the "sell" of drag and drop tie ups trickles down deeper into organizations than  intended, especially to those newly exploring DA. I will pass the information to our colleagues, some of them  being developers (unlike me) may already be aware of the pitfalls. The data manager suggestion may have been for reducing the learning  curve for new users like me.

Thanks,

Mithun