Question about $Webnode.Webnodes.GetItem()

Hi,

We've been tracking a bug where a thread on one of our PRD instances will all of a sudden start stack tracing. The section of code is always random, but invariably, it is always on a webnode object which for whatever reason no longer has the core method associated with it, i.e. Webnode.BrowseableVolumes(). Almost always, the stack trace will keep on repeating on a particular thread (meaning that the stack trace is not causing the thread to simply die) but only that thread.

In this particular case, I traced the failure to $NGD.PresentationUtils.GetRootNode(), and the highlighted line:

In the stack trace, the prgCtx is defined as is obj. From the stack trace:

Local: obj (6) = #7a00c9f0
Local: s (7) = ''
Local: status (8) = ?
Local: contents (9) = ?
Local: i (10) = 0
Local: viewName (11) = 'DTree'
Args: prgCtx (0) = #423ce07

My theory is that something is corrupting $Webnode.Webnodes.fRegistry and one of the objects is getting messed up (although I cannot think how - we're running only OpenText, i.e. no 3rd party code whatsoever - this is a Shared Services Canada instance). My question is this: Does each thread have its own cache of Oscript objects like $Webnode.Webnodes? If so, that would explain why this is always occurring on just one thread and not the others (although I've seen up to 3 threads exhibiting this behaviour at any time). Our recourse is always the same. We take the server out of the load balancer and restart it and the problem goes away for a while.

We've opened tickets on this but haven't so far got any resolution, so I'm reaching out to the development community to see if others have encountered this sort of behaviour and what you've done about it.

-Hugh Ferguson

Find more posts tagged with

Comments

Hans Stoop

Hi Hugh,

I’ve seen this behavior and we also restarted the service. This was on a CS 10.0 system.

At least with the later 10.5 systems you can just restart the threads instead of the whole service. I would beinterest to know if somebody has this problem on CS 10.5.

Hans

From: eLink Entry: Content Server Development Forum [mailto:development@elinkkc.opentext.com]
Sent: Dienstag, 3. Mai 2016 15:03
To: eLink Recipient
Subject: Question about $Webnode.Webnodes.GetItem()

Question about $Webnode.Webnodes.GetItem()

Posted by Ferguson, Hugh On 05/03/2016 08:57 AM

Hi,

In this particular case, I traced the failure to $NGD.PresentationUtils.GetRootNode(), and the highlighted line:

In the stack trace, the prgCtx is defined as is obj. From the stack trace:

Local: obj (6) = #7a00c9f0
Local: s (7) = ''
Local: status (8) = ?
Local: contents (9) = ?
Local: i (10) = 0
Local: viewName (11) = 'DTree'
Args: prgCtx (0) = #423ce07

-Hugh Ferguson

[To post a comment, use the normal reply function]
Forum:	Content Server Development Forum
Content Server:	Knowledge Center

HFCDev

Hi Hans,

Restarting threads would be handy but I'd really like to know what's causing this. And yes, this is CS 10.0. The fact that it's happening on one thread only each time suggests that objects are cached at the thread level. if it wasn't for the fact that this is a Shared Services Canada instance (meaning *nothing* can be deployed unless it comes from OpenText or a "gold" partner), we can't even write our own patch that wour give us some specific dbug info like whether the object in the WebNodes for loop is properly defined, and if not, which one is being clobbered. In today's incident, our PRD copy has a near identical configuration to the point that even the length of the webnodes registry and all the object addresses (i.e. #abcdef00) actually were the same as what I saw in the trace file. In that case, it was the FolderContents object from the GUI Enhancements that was the object it was trying to access. I'm sure tomorrow, it will be another one entirely.

-Hugh

Chris Meyer

Hi Hugh: Open a script window in Builder and run this:

Object obj = #7a00c9f0scheduler.debugbreak()

Then inspect obj and see if you can figure out anything about the object. Is it a valid WebNodes object? Something else?

HFCDev

In our PRD copy that returns a perfectly valid WebNode object. I believe our problem has to do with a run-time corruption of the list of webnode objects but it only ever occurs for a particular thread, and not the other threads. The thread which exhibits the behaviour is random but once it starts, it's always that thread. Restarting the service fixes the problem in the interim.

-Hugh

Chris Meyer

Hi Hugh: I doubt it's the WebNodes registry per se, but rather the WebNodes object with reference #7a00c9f0. What's the subtype of this WebNode?

HFCDev

In this case it was the FolderContents webnode in the GUIEnhancements ospace.

-Hugh

Chris Meyer

Hi Hugh: Can you double check that? I just had a look at guicomponents and didn't see a "FolderContents" web node (there are other "FolderContents" nodes in there, but none a subclass of WebNodes).

HFCDev

It's here:

ossie.moore@capspire.com

As a side note.. thank you for this little gem Chris. Very helpful snippet when debugging a trace file.

Ossie

On May 3, 2016, at 12:22 PM, eLink Entry: Content Server Development Forum <development@elinkkc.opentext.com> wrote:

eLink : Question about $Webnode.Webnodes.GetItem()
Question about $Webnode.Webnodes.GetItem() Posted byMeyer, ChrisOn 05/03/2016 01:19 PM

Hi Hugh: Open a script window in Builder and run this:Object obj = #7a00c9f0scheduler.debugbreak()Then inspect obj and see if you can figure out anything about the object. Is it a valid WebNodes object? Something else?
[To post a comment, use the normal reply function]
Topic: Question about $Webnode.Webnodes.GetItem()
Forum: Content Server Development Forum
Content Server: Knowledge Center