Troubleshooting Cisco Prime Access Registrar
This chapter provides information about techniques used when troubleshooting Cisco Prime Access Registrar (Cisco Prime AR) and highlights common problems.
This chapter contains the following sections:
•Gathering Basic Information
•Troubleshooting Quick Checks
•aregcmd and Cisco Prime Access Registrar Configuration
•RADIUS Request Processing
•Other Troubleshooting Techniques and Resources
•Checking AR Server Health Status
Gathering Basic Information
Table 28-1 lists UNIX commands that provide basic and essential information to help you understand the Cisco Prime AR installation environment.
Table 28-1 UNIX Commands to Gather Information
|
|
/usr/bin/uname -r |
Solaris release level |
/usr/bin/uname -i |
Machine hardware name |
/usr/bin/uname -v |
Solaris version |
/usr/bin/uname -a |
All system information including hostname, operating system type and release, machine model and type |
/usr/sbin/prtconf |
System configuration information including memory capacity, machine type, and peripheral equipment |
/usr/sbin/df -k |
File system disk space usage including partitions, capacity, and space used |
/usr/bin/ps -ef |
Currently running processes |
/usr/sbin/psinfo -v |
Information about processors |
/usr/bin/pkginfo -l CSCOar |
Software package information about Cisco Prime AR version number and installation directory |
Note More information about these commands and their options is available using the man command in a terminal window on the Sun workstation.
Troubleshooting Quick Checks
Many of the most common problems can be diagnosed by doing the following:
•Check disk space
•Check for resource conflicts
•Check the Cisco Prime AR log files
Disk Space
Running out of disk space can cause a number of problems including:
•Failure to process RADIUS requests
•Parts of the Cisco Prime AR configuration disappearing in aregcmd
•Failure to log into aregcmd
Check that the Cisco Prime AR installation partition ($INSTALL) and /tmp are not at capacity.
Resource Conflicts
Resource conflicts are a common reason for the Cisco Access Registrar 4.1 server failing to start. The most common resource conflicts are the following:
•Cisco Network Registrar is running on the Cisco Prime AR server
•Another application is also using ports 1645 and 1646
•A network management application is using the Sun SNMP Agent
No Co-Existence With Cisco Network Registrar
Cisco Network Registrar cannot coexist on a machine running Cisco Prime AR for this reason. You can determine if CNR is running by entering the following command line in a terminal window:
pkginfo | grep -i "network registrar"
Port Conflicts
The default ports used by the Cisco Prime AR server are ports 1645 and 1646. You should check to determine that no other applications are listening on the same ports as Cisco Prime AR.
You can check to see which TCP ports are in use by entering the following command line:
netstat -aP tcp
You can check to see which UDP ports are in use by entering the following command line:
netstat -aP udp
Note If you configure the Cisco Prime AR server to use ports other than the default, you will have to specifically add ports 1645 and 1646 if you want to also use those ports.
Server Running Sun SNMP Agent
If you plan to use the Cisco Prime AR server's SNMP agent, you cannot use the Sun Microsystems SNMP agent that comes with the Solaris operating system.
Cisco Prime Access Registrar Log Files
Examining the Cisco Prime AR log files can help you diagnose most Cisco Prime AR issues. By default, the Cisco Prime AR log files are located in /opt/CSCOar/logs. Table 28-2 lists the Cisco Prime AR log files and the information stored in each log.
Table 28-2 Cisco Prime AR Log Files
|
|
agent_server_1_log |
Log of the server agent process |
ar-status |
Log of Cisco Prime AR stop and start using the arserver utility |
aregcmd_log |
Log of commands executed in aregcmd (very useful for tracing the steps that took place before a problem occurred) |
config_mcd_1_log |
Log of the mcd internal database |
name_radius_1_log |
Log of the radius server process |
name_radius_1_trace |
Debugging output of RADIUS request processing (only generated when the trace level, set in aregcmd, is greater than zero) |
Modifying File Sizes for Agent Server and MCD Server Logs
The two parameters added to the car.conf file under $BASEDIR/conf affect the agent_server_logs and config_mcd_server_logs logs files:
•AGENT_SERVER_LOG_SIZE (10 MB by default)
•AGENT_SERVER_LOG_FILES (2 by default)
You will find these new parameters at the beginning of the car.conf file. When the log file size reaches the value set in AGENT_SERVER_LOG_SIZE, a rollover of the agent_server_log_file occurs. The value set in AGENT_SERVER_LOG_FILES specifies the number of log files to be created.
Using xtail to Monitor Log File Activity
A useful way of monitoring all of the log files is to run xtail, a utility provided with Cisco Prime AR. The xtail program monitors one or more files and displays all data written to a file since command invocation.
Run xtail in a dedicated terminal window. It is very useful for monitoring multiple logfiles simultaneously, such as with a command line like the following:
xtail $INSTALL/logs/*
Note Cisco AR 4.1.5 and later include the millisecond field in the logs' timestamp.
Modifying the Trace Level
By modifying the trace level, you can gather more detailed information in the log files about what is happening in the Cisco Prime AR server. There are five different trace levels. Each higher trace level also includes the information logged using lower trace levels. The different trace levels provide the following information:
•Level 0—No tracing occurs
•Level 1—Indicates when a packet is sent or received and when a status change occurs in a remote server (RADIUS Proxy and LDAP)
•Level 2—Information includes the following:
–Which services and session managers are used to process
–Which client and vendor objects are being used to process a packet
–More details about remote servers (RADIUS Proxy and LDAP), packet transmission, and timeouts
–Details about poorly-formed packets.
•Level 3—Information includes the following:
–Tracing of errors in Tcl scripts when referencing invalid RADIUS attributes
–Which scripts have been run
–Details about local userlist processing
•Level 4—Information includes the following:
–Advanced duplication detection processing
–Details about creating, updating, and deleting sessions
–Tracing of all APIs called during the running of a script
•Level 5—Provides information about policy engine operations
Installation and Server Process Start-up
The installation process installs the Cisco Prime AR software to the specified installation directory and then starts the server processes. This process rarely fails but the following checks should always be performed:
•Ensure that there is an installation success message at the end of the pkgadd dialog, otherwise check the dialog for the problem
•Follow the installation instructions carefully especially when performing an upgrade. For example, when upgrading to 1.6R1, 1.6R2, or 1.6R3, a post-installation upgrade script needs to be run
•Pay attention to the information included in README files
At the end of a successful installation, arstatus should show the following four server processes:
> $INSTALL/usrbin/arstatus
AR RADIUS server running (pid: 6285)
AR MCD lock manager running (pid: 6284)
AR MCD server running (pid: 6283)
AR Server Agent running (pid: 6277)
If any of the above processes are not displayed, check the log file of the failed process to determine the reason. The MCD processes might fail to start if Cisco Network Registrar is installed on the same machine.
The manual method of starting and stopping the Cisco Prime AR processes is using the arserver utility.
To start Cisco Prime AR processes: arserver start
To stop Cisco Prime AR processes: arserver stop
To restart Cisco Prime AR processes: arserver restart
aregcmd and Cisco Prime Access Registrar Configuration
While troubleshooting, you should always use the aregcmd command trace to turn on tracing. With tracing active, Cisco Prime AR generates debugging output to the log file name_radius_1_trace.The syntax is:
trace [<server>] [<level>]
When you do not specify a server, Cisco Prime AR sets the trace level for all servers in the current cluster. When you do not specify a trace level, the currently set level is used. The default trace level is 0.
Running and Stopped States
Cisco Prime AR can be in two states, running or stopped. In either state, all four Cisco Prime AR processes remain running. The state of Cisco Prime AR will be displayed when logging into aregcmd or by using the aregcmd status command:
status
Server 'Radius' is Running, its health is 10 out of 10\
The start and stop commands allow Cisco Prime AR to move between states. Reload is equivalent to a stop followed by a start if Cisco Prime AR is already running, and just a start if it is already stopped.
stop
Stopping Server 'Radius'...
Server 'Radius' is Stopped
start
Starting Server 'Radius'...
Server 'Radius' is Running, its health is 10 out of 10
reload
Reloading Server 'Radius'...
Server 'Radius' is Running, its health is 10 out of 10
During the transition from running to stopped, Cisco Prime AR stops processing new RADIUS requests and releases resources such memory, network and database connections and open files.
During the transition from stopped to running, Cisco Prime AR reverses this process by opening a connection with its internal database, reading configuration data, claiming memory, establishing network connections, opening files, and initializing scripts. During this transition, problems can occur. Cisco Prime AR might fail to start and display the following:
reload
Reloading Server 'Radius'...
Cisco Prime AR failed to move from stopped state to running:
status
Server 'Radius' is Stopped
This might occur for a number of reasons including the following:
•An invalid configuration
•Insufficient memory
•Listening ports already in use by another application
•Unable to open files
•Unable to initialize scripts
Check the name_radius_1_log file for the one of these indications.
RADIUS Request Processing
The main technique for troubleshooting RADIUS request processing in Cisco Prime AR is to examine the name_radius_1_trace log file with the trace level set to 5. Most issues are fairly self-explanatory. Some issues that can arise are as follows:
•Cisco Prime AR has marked a remote server as down
•A resource manager has run out of resources (for example, user or group session limit has been reached or no more IP addresses are available)
•A configuration error (such as an accounting service not being set)
•A run time error in a script
Some issues are not immediately evident from the log files though, such as the following:
•Failure to save or reload Cisco Prime AR after a configuration change
•Cisco Prime AR is not listening on the correct UDP ports for RADIUS requests
Other Troubleshooting Techniques and Resources
aregcmd Stats Command
The aregcmd command stats provides statistics on request processing.
--> stats
Global Statistics for Radius:
serverStartTime = Tue Oct 2 10:28:02 2008
serverResetTime = Tue Oct 2 20:25:12 2008
totalPacketsInPool = 1024
totalAccessChallenges = 0
totalAccountingRequests = 0
totalAccountingResponses = 0
totalStatusServerRequests = 0
totalAscendIPAAllocateRequests = 0
totalAscendIPAAllocateResponses = 0
totalAscendIPAReleaseRequests = 0
totalAscendIPAReleaseResponses = 0
totalUSRNASRebootRequests = 0
totalUSRNASRebootResponses = 0
totalUSRResourceFreeRequests = 0
totalUSRResourceFreeResponses = 0
totalUSRQueryResourceRequests = 0
totalUSRQueryResourceResponses = 0
totalUSRQueryReclaimRequests = 0
totalUSRQueryReclaimResponses = 0
totalPayloadDecryptionFailures = 0
Core Files
A core file in the Cisco Prime AR installation directory is an indication that Cisco Prime AR has crashed and restarted. Check that the radius server process generated the core file using the UNIX file command:
> file core
core: ELF 32-bit MSB core file SPARC Version 1, from 'radius'
Check the timestamp on the core file and look for corresponding log messages in the name_radius_1_log file in $INSTALL/logs. The word assertion commonly appears in core messages. Try to establish what caused the problem and contact Cisco TAC.
radclient
The Cisco Prime AR package provides a utility called radclient that allows RADIUS requests to be generated. Use radclient to test configurations and troubleshoot problems.
Cisco Prime Access Registrar Replication
For more information about using Cisco Prime AR replication, see Chapter 12 "Using Replication."
Checking AR Server Health Status
To check the server's health, use the aregcmd command status. The following issues decrement the server's health:
•Multiple occurances of Access-Request rejection
Note One of the parameters in the calculation of the Cisco Prime AR server's health is the percentage of responses to Access-Accepts that are rejections. In a healthy environment, the rejection percentage will be fairly low. An extremely high percentage of rejections could be an indication of a Denial of Service attack.
•Configuration errors
•Running out of memory
•Errors reading from the network
•Dropping packets that cannot be read (because the server ran out of memory)
•Errors writing to the network.
Cisco Prime AR logs all of these conditions. Sending multiple successful responses to any packet, increments the server's health.