Infrastructure Map Server Forum
Welcome to Autodesk’s Infrastructure Map Server Forums. Share your knowledge, ask questions, and explore popular Infrastructure Map Server topics.
cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 

Warning MG6 and Hyper threading

6 REPLIES 6
Reply
Message 1 of 7
dhaipola
474 Views, 6 Replies

Warning MG6 and Hyper threading

We were experiencing a consistent 524 error on our MG6 site making it almost unusable.
The problem only occurred on our production server a Dell 2650. After 3 weeks of trouble shooting and ZERO help from AutoDesk, the problem was diagnosed as MG6 having a problem with the virtual processors created by the Hyper-threading. Disabling Hyper-threading in the BIOS solved the problem instantly. Autodesk is not aware that this a problem.
6 REPLIES 6
Message 2 of 7
Anonymous
in reply to: dhaipola

We cannot reproduce this problem. Some
investigation has been done by the Autodesk MapGuide Software
Development team.

Below is pasted the findings.

--
Chris L. Amos
Autodesk MapGuide Quality
Engineering

 


size=2>==================================================

face=Arial size=2>
"Hey guys,

I've been doing some R6 testing with the hyperthreading
test box. The map I am testing contains 3 SDF, 3 SHP, 1 DWG, and 1 raster layer.
The test has run for 2 1/2 hours and processed >20,000 requests and
>100,000 layers. I will let the machine run overnight but I am not expecting
any unusual behaviour. Here's the setup:

Dell Precision 450, 1 GB DDR, 2 Xeon 2GHz cpus,
hyperthreading enabled

Windows XP Professional SP1 with all current critical
updates as of March 31, 2003

MapGuide Server R6, 6.0.4.6

MapGuide Provider for SHP R6, 6.0.3.6

When I set up the machine, I specifically chose Windows
XP. Windows 2000 Professional is not a recommended operating system for
hyperthreaded processors. Windows 2000 Server versions should be
ok.

 

[Reference]



 

 

I also found this article in MSDN. It suggests any
program using QueryPerformanceCounter may fail on hyperthreaded machines. I
think we are ok with MapGuide but it may be something we can look
into.


I am also checking out the Dell support site. There
seems to be lots of support traffic related to
hyperthreading
."

 


size=2>==================================================

 

 

 



style="PADDING-RIGHT: 0px; PADDING-LEFT: 5px; MARGIN-LEFT: 5px; BORDER-LEFT: #000000 2px solid; MARGIN-RIGHT: 0px">We
were experiencing a consistent 524 error on our MG6 site making it almost
unusable.
The problem only occurred on our production server a Dell 2650.
After 3 weeks of trouble shooting and ZERO help from AutoDesk, the problem was
diagnosed as MG6 having a problem with the virtual processors created by the
Hyper-threading. Disabling Hyper-threading in the BIOS solved the problem
instantly. Autodesk is not aware that this a problem.
Message 3 of 7
dhaipola
in reply to: dhaipola

Chris,
We are running a dell 2650 with dual 2.8 xeon
Running windows 2000 advanced server
4GB DDR 200 MHz ram
The problem is easily reproducible on our machine
Go into the bios enable hyperthreading and the 524 errors
come back.
Turn it off and the errors go away.
I read the article from MSDN

RESOLUTION
To resolve this issue, contact the manufacturer of the program that does not run correctly to determine whether there is an update available to address this issue.

The text above is from the article in question, does autodesk have a fix?


I appreciate that AutoDesk went to some length to test for this issue but I am surprised that you could not reproduce it.
There are two differences between your test and our installation
1. The problem occurred almost 100% while panning I’m not sure how the 20000 requests were generated but if it was not 20000 pans then the problem would never show up.
2. Our site also contained some 30 shapefiles, quite a bit larger then the 3 file listed in the test.
I am very interested in the outcome of this testing.
I am also interested in how to solicit autodesk attention earlier on in the troubleshooting process.
Danny
Message 4 of 7
Anonymous
in reply to: dhaipola

Danny,

Does your hardware use identical memory modules
throughout? If not, this could be the problem.

 

See below for the details of the further
testing done by the Autodesk MapGuide Software Development team.

--
Chris L. Amos
Autodesk MapGuide Quality
Engineering

 

========================

"I have successfully run a
simulation of 16 concurrent users using
randomized sequence of
200 pan requests under Windows 2000 Advanced Server for 17 hours.  Over
80,000 requests and 600,000 layers have been successfully returned from 6
different SHP files.  The CPU utilization has been pegged at 100% for all
"4" CPU's.

The test
machine is a Dell Precision 450 with dual 2.0 GHz Xeons and 1GB of ECC
DDR.  The motherboard uses 
class=453203919-08042003>an 
Intel E7505 chipset.  The memory
configuration is two 512MB DIMMS. 


There
have been issues posted on the Dell community web site related to mixing
memory types and the Intel E7505 chipset:



href="http://forums.us.dell.com/supportforums/board/message?board.id=pw_bios&message.id=564&highlig...



For the
E7501 chipset used in the PowerEdge 2600, Intel states:



color=#0000ff>Intel does not test, recommend, or support mixing of memory
types within the same server system.
Functionality issues may occur if mixed
memory types are installed in the same server system. Intel recommends that
memory modules of identical size, type, banking and stacking technology, and
vendor are installed in each server system. Intel will not provide support for
issues encountered when mixed memory configurations are in use.



href="http://support.intel.com/support/motherboards/server/se7501hg2/tested_mem.htm">http://support...



class=453203919-08042003>The PowerEdge 2650 uses a ServerWorks GC-LE
chipset.  I have not found any references to hyperthreading instability for
this chipset.  However, based on the confirmed issues with the Intel
chipsets, it would be wise to verify the memory configuration for the 2650 to
rule out hardware related instability.  This means the 2650 should be
configured with 4 identical 1GB DIMMs.



class=453203919-08042003>
face=Arial color=#0000ff>If the memory configuration is ruled out, we may have
to proceed with a more rigorous
analysis."


========================

 

 

 


style="PADDING-RIGHT: 0px; PADDING-LEFT: 5px; MARGIN-LEFT: 5px; BORDER-LEFT: #000000 2px solid; MARGIN-RIGHT: 0px">
Chris,

We are running a dell 2650 with dual 2.8 xeon
Running windows 2000
advanced server
4GB DDR 200 MHz ram
The problem is easily reproducible
on our machine
Go into the bios enable hyperthreading and the 524 errors

come back.
Turn it off and the errors go away.
I read the article
from MSDN

RESOLUTION
To resolve this issue, contact the manufacturer of the
program that does not run correctly to determine whether there is an update
available to address this issue.

The text above is from the article in question, does autodesk have a fix?


I appreciate that AutoDesk went to some length to test for this issue but I
am surprised that you could not reproduce it.
There are two differences
between your test and our installation
1. The problem occurred almost 100%
while panning I’m not sure how the 20000 requests were generated but if it was
not 20000 pans then the problem would never show up.
2. Our site also
contained some 30 shapefiles, quite a bit larger then the 3 file listed in the
test.
I am very interested in the outcome of this testing.
I am also
interested in how to solicit autodesk attention earlier on in the
troubleshooting process.
Danny

Message 5 of 7
dhaipola
in reply to: dhaipola

Chris,
We have 4GB DDR,200MHz,4x1GB DIMMS Factory installed by dell.

I would like to send you a scan of the packing list
so you can forward it to technical support.
Can you send me your email address.
Thanks for your help.

Danny
Message 6 of 7
andreva
in reply to: dhaipola

We experienced also a problem concerning MG 6.3 an HyperThreading (HT). On a Windows 2000 Server machine with Xeon 2.8Ghz and hyperthreading on we encountered problems concerning the Oracle Spatial Data Provider (OSDP), which was not able to make a connection to OSDP datasources, which resulted in a CPU hang-up (50% or 100% CPU usage on the MapServer instance). This occurred while working in the MG Author. HT is not supported for Win2000 (see Intel site), and Setting HT OFF in the BIOS resulted in a stable machine.

However, the same problem occurred on a simple REAL dual processor Compaq server (2x Xeon 500Mhz, 1,5 Gb RAM), also in combination with ODSP in MG63. After physically removing the second processor the machine became stable since then.

So, it seems that MG62SP2 and MG63 in combination with ODSP using Oracle Spatial (Oracle 8.1.7 EE, even with Patch for bug 1575474, see below) may be instable in allocating treads using multiple processors. Concerning Autodesk this is not an known issue.

Andre van Atten
Isis Benelux BV
www.isis.nl
---------------------

Bugs Fixed in Patch 1575474:

Intermittant Transform problem (no bug number)
Memory leak fixed for R-tree (Bug: 1396837)
R-tree tablespace parameters bug (no bug number)
R-tree error in queries (Bug: 1683199))
Message 7 of 7
Anonymous
in reply to: dhaipola

Have found the same problem on a compaq dual
processor.  MG 6.3, Oracle 901, with OSDP

 


--
Johan Nel
Open Spatial
Technologies

 

 


style="PADDING-RIGHT: 0px; PADDING-LEFT: 5px; MARGIN-LEFT: 5px; BORDER-LEFT: #000000 2px solid; MARGIN-RIGHT: 0px">
We
experienced also a problem concerning MG 6.3 an HyperThreading (HT). On a
Windows 2000 Server machine with Xeon 2.8Ghz and hyperthreading on we
encountered problems concerning the Oracle Spatial Data Provider (OSDP), which
was not able to make a connection to OSDP datasources, which resulted in a CPU
hang-up (50% or 100% CPU usage on the MapServer instance). This occurred while
working in the MG Author. HT is not supported for Win2000 (see Intel site),
and Setting HT OFF in the BIOS resulted in a stable machine.

However, the same problem occurred on a simple REAL dual processor Compaq
server (2x Xeon 500Mhz, 1,5 Gb RAM), also in combination with ODSP in MG63.
After physically removing the second processor the machine became stable since
then.

So, it seems that MG62SP2 and MG63 in combination with ODSP using Oracle
Spatial (Oracle 8.1.7 EE, even with Patch for bug 1575474, see below) may be
instable in allocating treads using multiple processors. Concerning Autodesk
this is not an known issue.

Andre van Atten
Isis Benelux BV
www.isis.nl

---------------------

Bugs Fixed in Patch 1575474:

Intermittant Transform problem (no bug number)

      Memory leak fixed for R-tree (Bug:
1396837)
      R-tree tablespace parameters
bug (no bug number)
      R-tree error in
queries (Bug: 1683199))

Can't find what you're looking for? Ask the community or share your knowledge.

Post to forums  

Autodesk Design & Make Report