If you would like to read the first part in this article series please go to Back Pressure in Exchange 2010 (Part 1).
Introduction
In part one of this two-part article on the back pressure feature in Exchange Server 2010, we covered the calculations that Exchange uses to determine the high, medium and normal back pressure levels for disk utilization. Here in part two we’ll look at how a breach of these levels will look as far as the Event Viewer is concerned as well as covering how things look when trying to telnet to a transport server to manually submit a new SMTP message. We will then finish by looking at the back pressure settings for memory utilization.
Event Viewer
Let’s start part two by having a look at what we see in the Event Viewer when a back pressure event has occurred that has been caused by a lack of disk space. In Figure 2-1, we can see event ID 15004 from a source of MSExchangeTransport that informs us that the resource pressure has increased from normal to high.
Figure 2-1: Event ID 15004 – High Back Pressure Event
You can see from Figure 2-1 that two separate resources are detailed, namely the disk space utilization associated with the queue database and queue database transaction logs. Specifically, you can see that the disk space utilization associated with the queue database is shown as 96%, as is the disk space utilization associated with the queue database transaction logs. This is because these two resources are located on the same disk.
We can also see from Figure 2-1 that the 96% utilization for the queue database disk is rated as medium usage. After this, we see that the normal, medium and high utilization percentages are stated as 93%, 95% and 97% respectively, as covered in part one of this article. Remember, these values are based on an example disk partition size of 20GB. Immediately below this information, we can see that the 96% utilization for the queue database transaction log disk is rated as high usage; therefore this event has been logged because the queue database transaction log disk has moved from normal to high usage not because the queue database disk has moved from normal to medium usage. After this information, we see that the normal, medium and high utilization percentages for the queue database transaction log disk are stated as 88%, 90% and 92% respectively, as covered in part one of this article. Again, these values are based on an example disk partition size of 20GB as the queue database and queue database transaction logs are located on the same disk.
Scrolling further down the information text within event ID 15004, we see that we are advised that the Hub Transport server is no longer accepting messages from other Hub Transport servers, the Internet, the pickup and replay folders, mailbox servers or the queuing database. Additionally, the Hub Transport server will not be delivering messages to remote domains. This is because the level of resource utilization has reached a value of high, meaning that full back pressure is applied.
Figure 2-2: Services Affected After Back Pressure Event
With full back pressure applied, connecting systems will not be able to send inbound messages to the Hub Transport server. For example, Figure 2-3 shows you what will be seen if you try to telnet to the Hub Transport server on port 25 and attempt to manually send an SMTP message. After entering the MAIL FROM: SMTP verb, we are advised that the service is not available via the 4.3.2 Service not available status message.
Figure 2-3: Service Not Available Status Message
You will see a similar thing in the SMTP protocol logs if you have them enabled. An example is shown in Figure 2-4 where you can see the 4.3.2 status message as well as the tarpit being used.
Figure 2-4: Protocol Log Example
While event ID 15004 refers to a resource moving from one back pressure level to another, you will see event ID 15006 logged if the resource issue is associated with a lack of disk space. This event ID is clear in that it advises you that “Microsoft Exchange Transport is rejecting message submissions because the available disk space has dropped below the configured threshold”. This is shown in Figure 2-5.
Figure 2-5: Event ID 15006
Finally, you will also see event ID 15005 logged when the resource back pressure has moved back to a normal state. You can see this event shown in Figure 2-6.
Figure 2-6: Event ID 15005 Advising Normal Service Resumption
Back Pressure and Memory Utilization
There are two different areas regarding memory utilization and back pressure in an Exchange Server 2010 environment that I want to cover. They are the memory used by the EdgeTransport.exe process in particular, and the memory used by all other processes on the server.
The EdgeTransport.exe file is used by the Microsoft Exchange Transport service. To calculate the high, medium and normal back pressure threshold levels of memory utilization for the EdgeTransport.exe service, Exchange Server 2010 first determines the high threshold utilization as a percentage of overall physical memory. Once this has been determined, the medium and normal thresholds are then determined.
The high threshold for the EdgeTransport.exe process is set at 75% of the overall physical memory, unless the process reaches 1TB of memory usage in which case the high threshold is also triggered. For example, consider a transport server with 8GB of memory. The high threshold will be met if the EdgeTransport.exe process reaches 6GB of memory utilization.
It is stated that the medium back pressure level for EdgeTransport.exe memory utilization is calculated from the lesser of two values. Those values are 73% of the physical memory, or the previously calculated high level minus 2%. For example, in our transport server with 8GB memory we can calculate that the medium back pressure level will be 73% of 8GB, which is 5.84GB. We previously calculated the high back pressure level to be 6GB, so 6GB minus 2% is 5.88GB. Therefore, the lesser of these two values is 5.84GB which will be set as the medium back pressure level. The normal back pressure level is also the lesser of two values, namely 71% of the physical memory, or the previously calculated high level minus 4%. In our example, these two figures are 5.68GB and 5.76GB respectively which means that a figure of 5.68GB will be used. Once back pressure is invoked for memory utilization of the EdgeTransport.exe process, unused memory is reclaimed which is termed garbage collection.
For the memory used by all other processes on the server, Exchange Server 2010 sets the high back pressure setting to be 94% of the total physical memory in the server. Once back pressure is invoked for memory utilization of all other processes, a feature known as message dehydration occurs which essentially looks at forcing the message to be re-read from the queue database rather than memory.
Other Information
If you are keen to understand the various settings that are used for back pressure, they can be found in the EdgeTransport.exe.config file that is, by default, located in the \Program Files\Microsoft\Exchange Server\V14\Bin folder on the drive where you installed Exchange Server 2010. However, it is very important to note that Microsoft strongly discourages making any modifications to these settings so please be aware of this.
Finally, if you’re interested in knowing exactly what actions are taken by Exchange when the various resources are under pressure, these are documented in the table titled Back pressure actions taken by Hub Transport and Edge Transport servers when responding to resource pressure in the topic titled Understanding Back Pressure. Rather than repeating those actions here, please take the time to read this table as it will help you understand just how the Exchange server is affected by the various actions.
Summary
That completes our brief look at the back pressure feature found in Exchange Server 2010. If you have ever run out of disk space on an Exchange server, or come close, it’s quite possible that you will have come across the back pressure feature before. If it is new to you, hopefully this article has explained the key concepts that will allow you to better understand it.
If you would like to read the first part in this article series please go to Back Pressure in Exchange 2010 (Part 1).
