Friday, March 5, 2010

Reducing the Type manager import time

Flex PLM  provides the flexibility to export the complete type manager or one node at a time. If you have around 30000 attributes defined in the system , then the best practice is to import the type manager type by type and not try to load it completely at a single stretch . Since , the type manager loader does not load in incremental commits. If the loader fails at line 29000 , it rolls back everything , and you have to start fresh.

   Maybe it would have been better if , after every type is loaded , a commit is been done and a restart point is created, so that the second time , when the type manager is imported , it does not start from the begining.

    Also , if it take 2 hours to load data upto line 20000 , it stops and does not tell is there is an issue at line 20001 as well .

  The second  methodology can be that , when the file is loaded , before loading , the type manage loader program should list out all the errors in advance. So that , the user can go and rectify all the errors and then load the type manager. At least he will save a lot of time.

Messed up Cluster Setup

Day 0:Thats it , development is freezing from today and from tommorrow , IT testing is going to start and on the actual clustered flexPLM system. Everyone, get ready from tommorrow , as we need to do IT testing. The test lead will distribute the test scripts to you guys, All the best.

Day 1:
Tester:  the test server is not up and running since the system admin did some mistakes to install FlexPLM system on a clustered environment.  we have the test scripts, but do not have a system.

Manager: We are talking to the sys admin and the issue will be resolved , he has a back up of all the servers.

Day 2: Tester: Is the system up and running today ?

 Manager: No still, we tried to restore the server on a single server mode , but its not working may be some changes in the property file.

 Day 3 : Still the test system is not up and running.

The point is , is it difficult to switch from server running 1 master and 3 slaves to just one master. I think in a clustered environments , slaves can be turned off at any time  . Slaves are to support the master. All the 4 systems use the same database and can be run standalone, but why is it difficult to restore a cluster.

 First of all , is there a procedure to back up a clustered FlexPLM system ? How do you rehost a clustered FlexPLM system ?  

Thursday, March 4, 2010

FlexPLM Data Loaders

An developer who has worked with XML will definitely loath working on the way the object loaders work in the FlexPLM system. I mean, i had to create a 200 attribute long mapping list for all my attributes first. What the heck. and everytime i need to check the position . I am waiting , if the future releases , a flexObject will be in a form of an XML document rather than the pain old hashmap .

Also , it would be nice if we could use an XML mapping document rather than a tilde (~ )delimited flat file .

XML is so developer friendly.

FlexPLM - Managing the alternate keys

Today , its been almost an year of a healthy up and running FlexPLM system. There is a new project, to integrate it with a legacy system. Its an Inbound integration where around 700 new attributes (Single list and Multilist) will be interfaced. But the values in flexPLM were totally different that the values we configured for the FlexPLM system during our design sessions an year ago, we never assumed that this need for a new integration willl arise in future.

There were two options to interface , either do the translation on the middleware or create alternate Keys in the Flex System.

   Creating alternate keys for all 700 attributes is a small project on its own. I think we should write a program to create the alternate keys rather than doing it manually. We can use the FlexAttribute API's to load the moa List and the single list values. What do you guys suggest. can this be a generic program , using which we can automate the process of creating alternate keys ?