Applies To:Show Versions
Introduction to health and performance monitoring
BIG-IP Local Traffic Manager can monitor the health or performance of either pool members or nodes. Local Traffic Manager supports these methods of monitoring:
- Simple monitoring
- Simple monitoring merely determines whether the status of a node is up or down. Simple monitors do not monitor pool members (and therefore, individual protocols, services, or applications on a node), but only the node itself. The system contains two types of simple monitors, ICMP and TCP_ECHO.
- Active monitoring
Active monitoring checks the status of a pool member or node on an ongoing
basis, at a set interval. If a pool member or node being checked does not respond within a
specified timeout period, or the status of a node indicates that performance is degraded, Local
Traffic Manager can redirect the traffic to another pool member or node. There are many types
of active monitors. Each type of active monitor checks the status of a particular protocol,
service, or application. For example, one type of monitor is HTTP. An HTTP type of monitor
allows you to monitor the availability of the HTTP service on a pool, pool member, or node. A
WMI type of monitor allows you to monitor the performance of a node that is running the Windows
Management Instrumentation (WMI) software. Active monitors fall into two categories: Extended
Content Verification (ECV) monitors, and Extended Application Verification (EAV) monitors.
Note: If you configure a performance monitor, such as the SNMP DCA or WMI monitor type, you should also configure a health monitor. Configuring a health monitor ensures that Local Traffic Manager reports accurate node availability status.
- Passive monitoring
- Passive monitoring occurs as part of a client request. This kind of monitoring checks the health of a pool member based on a specified number of connection attempts or data request attempts that occur within a specified time period. If, after the specified number of attempts within the defined interval, the system cannot either connect to the server or receive a response, or if the system receives a bad response, the system marks the pool member as down. There is only one type of passive monitor, called an Inband monitor.
Comparison of monitoring methods
In the short description, briefly describe the purpose and intent of the information contained in this topic. This element is an F5 requirement.
About monitor settings
Every monitor consists of settings with values. The settings and their values differ depending on the type of monitor. In some cases, Local Traffic Manager assigns default values. This example shows that an ICMP-type monitor has these settings and default values.
The settings specify that an ICMP type of monitor is configured to check the status of an IP address every five seconds, and to time out every 16 seconds. The destination IP address that the monitor checks is specified by the Alias Address setting, with the value * All Addresses. Thus, in the example, all IP addresses with which the monitor is associated are checked.Name my_icmp Type ICMP Interval 5 Timeout 16 Transparent No Alias Address * All Addresses
Overview of monitor implementation
You implement monitors using either the BIG-IP Configuration utility or a command line utility. The task of implementing a monitor varies depending on whether you are using a pre-configured monitor or creating a custom monitor. A pre-configured monitor is an existing monitor that Local Traffic Manager provides for you, with its settings already configured. A custom monitor is a monitor that you create based on one of the allowed monitor types.
If you want to implement a pre-configured monitor, you need only associate the monitor with a pool, pool member, or node, and then configure the virtual server to reference the relevant pool. If you want to implement a custom monitor, you must first create the custom monitor. Then you can associate the custom monitor with a pool, pool member, or node, and configure the virtual server to reference the pool.
For a subset of monitor types, Local Traffic Manager includes a set of pre-configured monitors. You cannot modify pre-configured monitor settings, as they are intended to be used as is. The purpose of a pre-configured monitor is to eliminate the need for you to explicitly create a monitor. You use a pre-configured monitor when the values of the settings meet your needs as is.
The names of the pre-configured monitors that Local Traffic Manager includes are:
An example of a pre-configured monitor is the icmp monitor. The example shows the icmp monitor, with values configured for its Interval, Timeout, and Alias Address settings. Note that the Interval value is 5, the Timeout value is 16, the Transparent value is No, and the Alias Address value is * All Addresses.
If the Interval, Timeout, Transparent, and Alias Address values meet your needs, you simply assign the icmp pre-configured monitor directly to a pool, pool member, or node, using the Pools or Nodes screens within the BIG-IP Configuration utility. In this case, you do not need to use the Monitors screens, unless you simply want to view the values of the pre-configured monitor settings.Name icmp Type ICMP Interval 5 Timeout 16 Transparent No Alias Address * All Addresses
You create a custom monitor when the values defined in a pre-configured monitor do not meet your needs, or no pre-configured monitor exists for the type of monitor you are creating.
When you create a custom monitor, you use the BIG-IP Configuration utility or a command line utility to: give the monitor a unique name, specify a monitor type, and, if a monitor of that type already exists, import settings and their values from the existing monitor. You can then change the values of any imported settings.
You must base each custom monitor on a monitor type. When you create a monitor, the BIG-IP Configuration utility displays a list of monitor types. To specify a monitor type, simply choose the one that corresponds to the service you want to check. For example, if you want to want to create a monitor that checks the health of the HTTP service on a pool, you choose HTTP as the monitor type.
If you want to check more than one service on a pool or pool member (for example HTTP and HTTPS), you can associate more than one monitor on that pool or pool member.
Checking services is not the only reason for implementing a monitor. If you want to verify only that the destination IP address is alive, or that the path to it through a transparent node is alive, use one of the simple monitors, icmp or tcp_echo. Or, if you want to verify TCP only, use the monitor tcp.
Importing settings from a pre-configured monitor
If a pre-configured monitor exists that corresponds to the type of custom monitor you are creating, you can import the settings and values of that pre-configured monitor into the custom monitor. You are then free to change those setting values to suit your needs. For example, if you create a custom monitor called my_icmp, the monitor can inherit the settings and values of the pre-configured monitor icmp. This ability to import existing setting values is useful when you want to retain some setting values for your new monitor but modify others.
The example shows a custom ICMP-type monitor called my_icmp, which is based on the pre-configured monitor icmp. Note that the Interval value is changed to 10, and the Timeout value is 20. The other settings retain the values defined in the pre-configured monitor.Name my_icmp Type ICMP Interval 10 Timeout 20 Transparent No Alias Address * All Addresses
Importing settings from a custom monitor
You can import settings from another custom monitor instead of from a pre-configured monitor. This is useful when you would rather use the setting values defined in another custom monitor, or when no pre-configured monitor exists for the type of monitor you are creating. For example, if you create a custom monitor called my_oracle_server2, you can import settings from an existing Oracle-type monitor such as my_oracle_server1. In this case, because Local Traffic Manager does not provide a pre-configured Oracle-type monitor, a custom monitor is the only kind of monitor from which you can import setting values.
Selecting a monitor is straightforward. Like icmp, each of the monitors has a Type setting based on the type of service it checks, for example, http, https, ftp, pop3, and takes that type as its name. (Exceptions are port-specific monitors, like the external monitor, which calls a user-supplied program.)
By default, the value for the Alias Address setting in the monitors is set to the wildcard * Addresses, and the Alias Service Port setting is set to the wildcard * Ports. This value causes the monitor instance created for a pool, pool member, or node to take that node’s address or address and port as its destination. You can, however, replace either or both wildcard symbols with an explicit destination value, by creating a custom monitor. An explicit value for the Alias Address and/or Alias Service Port setting is used to force the instance destination to a specific address and/or port which might not be that of the pool, pool member, or node.
The ECV monitor types HTTP, HTTPS, and TCP include the settings Send String and Receive String for the send string and receive expression, respectively.
The most common Send String value is GET /, which retrieves a default HTML page for a web site. To retrieve a specific page from a web site, you can enter a Send String value that is a fully qualified path name:"GET /www/support/customer_info_form.html"
The Receive String value is the text string that the monitor looks for in the returned resource. The most common Receive String values contain a text string that is included in a particular HTML page on your site. The text string can be regular text, HTML tags, or image names.
The sample Receive String value below searches for a standard HTML tag:"<HEAD>"
You can also use the default null Receive String value [""]. In this case, any content retrieved is considered a match. If both the Send String and Receive String fields are left empty, only a simple connection check is performed.
For HTTP and FTP monitor types, you can use the special values GET or hurl in place of Send String and Receive String values. For FTP monitors specifically, the GET value should specify the full path to the file to retrieve.
Transparent and Reverse modes
The normal and default behavior for a monitor is to ping the destination pool, pool member, or node by an unspecified route, and to mark the node up if the test is successful. However, with certain monitor types, you can specify a route through which the monitor pings the destination server. You configure this by specifying the Transparent or Reverse setting within a custom monitor.
Sometimes it is necessary to ping the aliased destination through a transparent pool, pool member, or node. When you create a custom monitor and set the Transparent setting to Yes, Local Traffic Manager forces the monitor to ping through the pool, pool member, or node with which it is associated (usually a firewall) to the pool, pool member, or node. (That is, if there are two firewalls in a load balancing pool, the destination pool, pool member, or node is always pinged through the pool, pool member, or node specified; not through the pool, pool member, or node selected by the load balancing method.) In this way, the transparent pool, pool member, or node is tested: if there is no response, the transparent pool, pool member, or node is marked as down.
Common examples are checking a router, or checking a mail or FTP server through a firewall. For example, you might want to check the router address 10.10.10.53:80 through a transparent firewall 10.10.10.101:80. To do this, you create a monitor called http_trans in which you specify 10.10.10.53:80 as the monitor destination address, and set the Transparent setting to Yes. Then you associate the monitor http_trans with the transparent pool, pool member, or node.
This causes the monitor to check the address 10.10.10 53:80 through 10.10.10.101:80. (In other words, the BIG-IP system routes the check of 10.10.10.53:80 through 10.10.10.101:80.) If the correct response is not received from 10.10.10.53:80, then 10.10.10.101:80 is marked down.
With the Reverse setting set to Yes, the monitor marks the pool, pool member, or node down when the test is successful. For example, if the content on your web site home page is dynamic and changes frequently, you may want to set up a reverse ECV service check that looks for the string "Error". A match for this string means that the web server was down.
Monitors that contain the Transparent or Reverse settings
This table shows the monitors that contain either the Transparent setting or both the Reverse and Transparent settings.
|TCP||Transparent and Reverse|
|HTTP||Transparent and Reverse|
|HTTPS||Transparent and Reverse|
|TCP Half Open||Transparent|
The Manual Resume feature
By default, when a monitor detects that a resource (that is, a node or a pool member) is unavailable, the BIG-IP system marks the resource as down and routes traffic to the next appropriate resource as dictated by the active load balancing method. When the monitor next determines that the resource is available again, the BIG-IP system marks the resource as up and immediately considers the resource to be available for load balancing connection requests. While this process is appropriate for most resources, there are situations where you want to manually designate a resource as available, rather than allow the BIG-IP system to do that automatically. You can manually designate a resource as available by configuring the Manual Resume setting of the monitor.
For example, consider a monitor that you assigned to a resource to track the availability of an HTML file, index.html, for a web site. During the course of a business day, you decide that you need to restart the system that hosts the web site. The monitor detects the restart action and informs the BIG-IP system that the resource is now unavailable. When the system restarts, the monitor detects that the index.html file is available, and begins sending connection requests to the web site. However, the rest of the web site might not be ready to receive connection requests. Consequently, the BIG-IP system sends connection requests to the web site before the site can respond effectively.
To prevent this problem, you can configure the Manual Resume setting of the monitor. When you set the Manual Resume setting to Yes, you ensure that the BIG-IP system considers the resource to be unavailable until you manually enable that resource.
Resumption of connections
If you have a resource (such as a pool member or node) that a monitor marked as down, and the resource has subsequently become available again, you must manually re-enable that resource if the monitor’s Manual Resume setting is set to Yes. Manually re-enabling the resource allows the BIG-IP system to resume sending connections to that resource.
The procedure for manually re-enabling a resource varies depending on whether the resource is a pool, a pool member, or a node.
The Time Until Up feature
By default, the BIG-IP system marks a pool member or node as up immediately upon receipt of the first correct response to a ping command.
The Time Until Up feature provides a way to adjust the default behavior. This feature allows the system to delay the marking of a pool member or node as up for some number of seconds after receipt of the first correct response. The purpose of this feature is to ensure that the monitor marks the pool member or node as up only after the pool member or node has consistently responded correctly to the BIG-IP system during the defined time period. With this feature, you ensure that a pool member or node that is available only momentarily, after sending one correct response, is not marked as up.
A Time Until Up value of 0 causes the default behavior. When the Time Until Up value is a non-0 value, the BIG-IP system marks a pool member or node as up only when all pool member or node responses during the Time Until Up period are correct.
Dynamic ratio load balancing
You can configure Dynamic Ratio load balancing for pools that consist of RealNetworks RealServer servers, Microsoft Windows servers equipped with Windows Management Instrumentation (WMI), or any server equipped with an SNMP agent such as the UC Davis SNMP agent or Windows 2000 Server SNMP agent.
To implement Dynamic Ratio load balancing for these types of servers, BIG-IP Local Traffic Manager provides a special monitor plug-in file and a performance monitor for each type of server. The exception is a server equipped with an SNMP agent. In this case, Local Traffic Manager provides the monitor only; no special plug-in file is required for a server running an SNMP agent.
You must install the monitor plug-in on each server to be monitored, and you must create a performance monitor that resides on the BIG-IP system. Once you have created a monitor, the monitor communicates directly with the server plug-in.
Monitor plug-ins and corresponding monitor templates
For each server type, this table shows the required monitor plug-in and the corresponding performance monitor types.
|Server Type||Monitor plug-in||Monitor Type|
|RealServer Windows server||F5RealMon.dll||Real Server|
|RealServer UNIX server||f5realmon.so||Real Server|
|Windows server with WMI||f5isapi.dll or F5Isapi64.dll or F5.IsHandler.dll||WMI|
|Windows 2000 Server server||SNMP agent||SNMP DCA and SNMP DCA Base|
|UNIX server||UC Davis SNMP agent||SNMP DCA and SNMP DCA Base|
Monitor association with pools and nodes
You must associate a monitor with the server or servers to be monitored. The server or servers can be either a pool, a pool member, or a node, depending on the monitor type. You can associate a monitor with a server in any of these ways:
- Monitor-to-pool association
- This type of association associates a monitor with an entire load balancing pool. In this case, the monitor checks all members of the pool. For example, you can create an instance of the monitor http for every member of the pool my_pool, thus ensuring that all members of that pool are checked.
- Monitor-to-pool member association
- This type of association associates a monitor with an individual pool member, that is, an IP address and service. In this case, the monitor checks only that pool member and not any other members of the pool. For example, you can create an instance of the monitor http for pool member 10.10.10.10:80 of my_pool.
- Monitor-to-node association
- This type of association associates a monitor with a specific node. In this case, the monitor checks only the node itself, and not any services running on that node. For example, you can create an instance of the monitor icmp for node 10.10.10.10. In this case, the monitor checks the specific node only, and not any services running on that node. You can designate a monitor as the default monitor that you want Local Traffic Manager to associate with one or more nodes. In this case, any node to which you have not specifically assigned a monitor inherits the default monitor.
Some monitor types are designed for association with nodes only, and not pools or pool members. Other monitor types are intended for association with pools and pool members only, and not nodes.
Node-only monitors specify a destination address in the format of an IP address with no service port (for example, 10.10.10.2). Conversely, monitors that you can associate with nodes, pools, and pool members specify a destination address in the format of an IP address and service port (for example, 10.10.10.2:80). Therefore, when you use the BIG-IP Configuration utility to associate a monitor with a pool, pool member, or node, the utility displays only those pre-configured monitors that are designed for association with that server.
For example, you cannot associate the monitor icmp with a pool or its members, since the icmp monitor is designed to check the status of a node itself and not any service running on that node.
When you associate a monitor with a server, Local Traffic Manager automatically creates an instance of that monitor for that server. A monitor association thus creates an instance of a monitor for each server that you specify. This means that you can have multiple instances of the same monitor running on your servers.
Because instances of monitors are not partitioned objects, a user can enable or disable an instance of a monitor without having permission to manage the associated pool or pool member.
For example, a user with the Manager role, who can access partition AppA only, can enable or disable monitor instances for a pool that resides in partition Common. However, that user cannot perform operations on the pool or pool members that are associated with the monitor. Although this is correct functionality, the user might not expect this behavior. You can prevent this unexpected behavior by ensuring that all pools and pool members associated with monitor instances reside in the same partition.