[Brian Huneycutt has contributed today’s article]
Configuration Manager 2007 can operate successfully in a wide variety of network environments. However, there are still some factors to consider in regards to planning the location of site systems. This is of particular importance for those that Site Component Manager watches, such as a management point or state migration point.
Understanding Configuration Manager Sites is a core document that contains a section on site system connectivity.
“Sites are typically configured so that the clients and site systems have fast connectivity with each other, usually local area network (LAN) speed.”
In the same document is a section on intra-site communications that covers the following.
“When Configuration Manager 2007 components that are within the site boundaries communicate with each other, they use either server message block (SMB), HTTP, or HTTPS, depending on various site configuration choices you make. Because all of these communications are unmanaged—that is, they happen at any time with no consideration for bandwidth consumption—it is beneficial to make sure these site elements have fast communication channels.”
One thing that we do not specifically mention is network latency. Our guidance operates under the assumption that part of a fast communication channel is not only high speed, but low latency as well. We recently had a customer report some issues in a high latency environment so we wanted to share how the symptoms can appear.
Attempts to install, alter, or even uninstall managed site systems (management points and PXE service points for example) in a high latency environment may fail. The key to diagnosing issues related to latency lies in reviewing the time stamps in the log files. If polling operations take several seconds to complete and later timeout then latency is likely too high. Here is an example taken from the sitecomp.log file of a site server where there was an attempt to uninstall a state migration point:
//Starting bootstrap operations... 3/27/2011 12:25:50 PMInstalled service SMS_SERVER_BOOTSTRAP_SITESERVER. 3/27/2011 12:26:18 PMStarting service SMS_SERVER_BOOTSTRAP_SITESERVER with command-line arguments "CP3 D:\SMS /deletefile \\REMOTESERVER\D$\SMS\bin\i386\firstname.lastname@example.org"... 3/27/2011 12:26:29 PMCould not start service SMS_SERVER_BOOTSTRAP_SITESERVER. The operating system reported error 997: Overlapped I/O operation is in progress. 3/27/2011 12:59:21 PM //
Notice the delays between operations. All of these are typically a second or two apart. Instead, we have 28 seconds, 11 seconds, and an eventual timeout a little over 30 minutes later. A network trace taken at the same time should illustrate lengthy delays between calls to check the status of the service. Ultimately, this is too much latency for successful management of a remote site system.
If improving network conditions is not possible in these scenarios, the next recourse is to put a new dedicated site on the same end of the high latency connection. This way intra-site communications can happen properly; the local segment should be low latency, and the remote connection uses the LAN Sender component between sites, which is much more tolerant of latency.
This posting is provided "AS IS" with no warranties, and confers no rights.