| Author |
Message
|
| mattfarney |
 Posted: Thu Dec 20, 2007 12:40 pm Post subject: Message truncation/garble issue |
 |
|
Disciple
Joined: 17 Jan 2006
Posts: 167
Location: Ohio
|
Running WBI v5 on Windows (running on virtual blade servers).
Every now and then, we are getting an XML message that is either truncated or garbled. The occurance rate is very low (on the order of 1 in 100k). The volume of message is going to prevent debugging in the real world, and we have been unable to replicate the problem in another environment.
Example of truncation:
| Code: |
| <?xml version=1.0><BOD><aBOD><Key>Hello</Key></a |
Example of garbled:
| Code: |
| <?xml version=1.0><BOD><aBOD><Key>Hello</Key><aBO<Key>Hello2</Key><aBOD><BOD> |
Has anyone seen anything like this before? It is as if the temporary memory space of the broker flow is either getting corrupted or not getting cleaned out.
-Matt |
|
| Back to top |
|
 |
| jefflowrey |
| Posted: Thu Dec 20, 2007 1:33 pm Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
Are you sure that the incoming physical XML message is not truncated, or does not contain (at the point of truncation) illegal XML characters?
_________________
I am *not* the model of the modern major general. |
|
| Back to top |
|
|
| mattfarney |
| Posted: Thu Dec 20, 2007 1:41 pm Post subject: |
|
|
Disciple
Joined: 17 Jan 2006
Posts: 167
Location: Ohio
|
Yes, as best as I can tell.
To help with support, we send a tagged copy of the message when it enters WBI and one when it leaves.
In these cases, those messages show a difference in size/content.
As much as I'd like to blame something other than MQ/WBI, at the moment, it is by far the most likely culprit.
-mf |
|
| Back to top |
|
|
| jefflowrey |
| Posted: Thu Dec 20, 2007 1:48 pm Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
Can you use one of those tagged messages to replicate the truncation?
Or does it just "work" the second time?
If you can replicate the problem, then you can run a trace.
If you can't replicate the problem, then you either aren't preserving the message contents fully - or the problem is really as intermittent as you think it is. And thus... support is going to have a very hard time replicating it too....
Would you be offended if I reminded you that v5 is going out of support in less than a year, and now's a really good time to start migrating to v6 or v6.1 - which may also cause this problem to go away, if it's actually WMB's issue?
_________________
I am *not* the model of the modern major general. |
|
| Back to top |
|
|
| mattfarney |
| Posted: Thu Dec 20, 2007 2:24 pm Post subject: |
|
|
Disciple
Joined: 17 Jan 2006
Posts: 167
Location: Ohio
|
Replicating from the tagged messages is what I am trying currently. (Not easily possible, since the copies of the messages are parsed by another program for key info to save filespace).
When the original messages are regenerated/retransmitted, they process with no problems.
I'm certainly not offended by the suggestion to upgrade... 
-mf |
|
| Back to top |
|
|
| jefflowrey |
| Posted: Thu Dec 20, 2007 2:38 pm Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
| mattfarney wrote: |
| Replicating from the tagged messages is what I am trying currently. (Not easily possible, since the copies of the messages are parsed by another program for key info to save filespace). |
Which means that these copies do not prove or disprove that the original input bitstream has only valid XML characters in it. 
_________________
I am *not* the model of the modern major general. |
|
| Back to top |
|
|
| mattfarney |
| Posted: Thu Dec 20, 2007 2:52 pm Post subject: |
|
|
Disciple
Joined: 17 Jan 2006
Posts: 167
Location: Ohio
|
Agreed.
Nothing short of a trace or manual message inspection will though right?
-mf |
|
| Back to top |
|
|
| jefflowrey |
| Posted: Thu Dec 20, 2007 3:26 pm Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
| mattfarney wrote: |
Agreed.
Nothing short of a trace or manual message inspection will though right?
-mf |
Yep... which again, hard to do if you can't replicate the problem...
You could consider putting in mirrorq, temporarily, if the Broker flow is fed from a queue. Or some such... And then do something to remove messages after a couple of days or etc., to allow for keeping the "backup" queue smaller than "way too big".
Then at least you have a backup that you can confirm message content against.
rather than leaving trace running over a span of 100K messages...
_________________
I am *not* the model of the modern major general. |
|
| Back to top |
|
|
|
|
