Slow Windchill performance for some users (PDMLink 10 M030)

[egifford]

We’re experiencing speed issues with Windchill PDMLink 10 M030 that seem to get progressively worse over time. The issue appears to be directly related to the user, it is not specific to a client PC, number or size of workspaces, product context, network issues or other factors.

Here are the symptoms & setup:

• Windchill login screen when connected to Pro/E (we’re on Wildfire 4 M220) is slow, it takes 20+ seconds to come up


• Very slow repaint when switching workspace being viewed. This happens whether it’s in the browser in Pro/E or an external browser.The user has to watch the “spinning wheel of death” for 30 seconds before a 45 object workspace displays. A user not experiencing the issue can switch the workspace being viewed and it takes under 3 seconds, even if the workspace has hundreds of objects.


• The issue follows the user. If user “A” with the speed issue connects to Windchill from user “B’s” PC where “B” has no speed issue, user “A” still experiences the speed issue (even in an external browser ).


• We had tried clearing the browser cache, java cache, clearing the windchill cache & reconnecting to the server and going back to a java 6.13 on the client PC from 7, but it makes zero difference. Plus, as mentioned, the issue follows the user from PC to PC.


• Not related to the number of workspaces or data size in the workspaces. A user with normal performace might have 10 workspaces with hundreds of object each. A user experiencing slow performance might have 5 workspaces each with no more than 50 objects.


• All users are authenticating through a connection to Active Directory.


• Level of authorization to the data and group assignment does not appear to be related (under the product team):
  
  
  
  • User B is individually assigned to a Product Team as product manager, no speed issues.
  
  
  
  
  • User A is a in a group assigned as a member for the same Product. She has speed issues.
  
  
  
  
  • Pull A from the group, assign them individually as a product manager of the product, speed issue persists.
  
  
  
  
  • (so it’s not something related to the extra layer of “group” in the authentication / rights process)
  
  


• All workstations are Windows 7 64 bit, mix of Dell T3500 & Dell T3600 PCs. Browser is 32 bit IE 9.


• Windchill is running as two separate servers (app & database) on Win Server 2008 R2.Speed issue existed on previous Oracle DB and after recently moving to Microsoft SQL server.

I may try rebooting the entire Windchill system overnight tonight. But, even if that clears it up I’m sure it will creep back in. Anyone have any suggestions on where the root cause might be? Has anyone else resorted to regularly rebooting Windchill for performace issues.

Thanks,

Erik

[SteveVinyard]

Any extra "status columns" in the View in the workspace that is so slow?

[cid:image001.gif@01CE4CAE.E63B0090]

Steve Vinyard
Senior Solution Architect

[egifford]

Steve,

We do have some status columns that are known to effect speed of workspace display, but the display is set the same for all users. A user experiencing the slow speed might only have 5 objects in a workspace 20 in the next while a user with no issues might have hundreds of objects in each workspace. If the system has to check for out of date status, renames etc, wouldn't a larger workspace display slower than a small one? The workspace display speed is also always paired with the long wait for the Windchill login prompt to appear when launching a connected session of Pro.

Thanks for the prompt reply & for all the posts you write to the group. You've provided a lot of good information that has helped me in the past.

Regards,


Erik C. Gifford
IT Dept., CAD Systems Administrator
G.W. Lisk Co., Inc.
Clifton Springs, NY

[SteveVinyard]

OK, well when it doubt go with the scientific method.


1- Remove all configs

a. config.pro, config.sup, win etc etc...anything and EVERYTHING not OOTB

2- Start with entirely clean caches

3- Install Java 6.30ish (1.6.30). 6.13 is too old

a. Ensure all of your existing java stuff gets removed properly beforehand so you start fresh.

Also do the following:

1- Try restarting Windchill and clearing the caches on it as a test (we are trying to find the culprit first, not a long term fix)

2- Look at the logs closely while testing the slow issue

3- Try thing later in the evening when no one is on the network and see if there is a big difference.

4- See if you can narrow it down to particular users. Perhaps you have an LDAP query issue.

If you can get it working consistently fast then we will start by adding things back in one by one and figure out which is the issue.



[cid:image001.gif@01CE4CB6.61066150]

Steve Vinyard
Senior Solution Architect

[ddemay]

Do you have all the known indexes for the database tables configured? PTC
has been updating a master list here for 10.0-10.1 all datecodes:

[ddemay]

Should also add, when you have multiple method servers running. One may be
slower than the other. The tomcat process may actually be the slower
bottleneck. Clear your tomcat cache.



When you state it is user based and this user is testing from a different
PC, how long are you waiting in between trying the tests. I'm fairly
certain the system will keep sessions alive for at least 30 minutes usually
even when the IP address of the requesting client changes.



This would translate to if you switch machines or clients too quickly,
you'll still perceive same performance issue.

[gwalker]

TheFor example, ifa user with slowresponsetransferred from another site, they may still be connecting to the remote site file server.

[lreddy-2]

Please check for any browser add-ons that might be affecting the performance and disable them.