Cluster Configuration

The cluster-config configuration element manages basic SLB interaction with clustered OCSBCs — it contains a set of global parameters that define the management of the RFC 2003 IP-in-IP tunnels that connect the SLB to clustered OCSBCs, and the details of rebalance operations. In addition, cluster-config provides for the creation of a list of service interfaces (signaling addresses) that are advertised to endpoints comprising the user access population.

Use the following procedure to perform required cluster-config configuration.

  1. From superuser mode, use the following ACLI command sequence to access cluster-config configuration mode. While in this mode, you configure the cluster-config configuration element. 
    ORACLE# configure terminal
ORACLE(configure)# session-router
ORACLE(session-router)# cluster-config
ORACLE(cluster-config)# ?

state                         cluster control state
log-level                     configure log level
auto-rebalance                Auto-rebalance cluster on new SD availability
source-rebalance-threshold    Percentage of advertised registration capacity
dest-rebalance-threshold      Percentage of advertised registration capacity
dest-rebalance-max            Percentage of advertised registration capacity
tunnel-check-interval         How often an SD's tunnels are checked
tunnel-fail-interval          Time for which no messages have been received
rebalance-request-delay       Delay between subsequent rebalance requests
session-multiplier            ratio of users (endpoints to sessions)
atom-limit-divisor            ratio of atoms (e.g. contacts to endpoints)
rebalance-skip-ahead          Skip endpoints refreshing sooner than
rebalance-max-refresh         Skip endpoints refreshing later than
ignore-tgt-svcs-on-rebalance  When selecting source SDs during rebalancing
rebalance-del-app-entries     Delete Application endpoint Data
inactive-sd-limit             Duration no SD control messages received
                              (seconds)
red-port                      redundant mgcp sync port
red-max-trans                 max redundant transactions to keep
red-sync-start-time           redundant sync start timeout
red-sync-comp-time            redundant sync complete timeout
service-ports                 configure service ports
select                        select cluster config
no                            delete cluster config
show                          show cluster config
done                          save cluster config information
exit                          return to previous menu
  2. Use the state parameter to enable or disable the SLB software.

    The default setting, enabled, enables SLB functionality; disabled renders the SLB inoperable. 

    ORACLE(cluster-config)# state enabled
ORACLE(cluster-config)#
  3. Use the log-level parameter to specify the contents of the SLB log.

    Log messages are listed below in descending order of severity.

    • emergency — the most severe
    • critical
    • major (error)
    • minor (error)
    • warning
    • notice
    • info — (default) the least severe
    • trace — (test/debug, not used in production environments)
    • debug — (test/debug, not used in production environments)
    • detail — (test/debug, not used in production environments)

    In the absence of an explicitly configured value, log-level defaults to critical, meaning that log messages with a severity of critical or greater (emergency) are written to the SLB log. 

    ORACLE(cluster-config)# log-level critical
ORACLE(cluster-config)#
  4. Use the auto-rebalance parameter to specify SLB behavior when a new OCSBC joins an existing cluster.

    With this parameter enabled, the default setting, the SLB redistributes endpoints among cluster members when a new member joins the cluster. Refer to the Rebalancing section for operational details.

    With this parameter disabled, the alternate setting, pre-existing OCSBCs retain their endpoint populations, and the SLB directs all new endpoints to the newly active OCSBC until that OCSBC reaches maximum occupancy. 

    ORACLE(cluster-config)# auto-rebalance enabled
ORACLE(cluster-config)#
  5. If auto-rebalance is set to enabled, use the source-rebalance-threshold and dest-rebalance-threshold parameters to specify threshold settings that identify existing cluster SBCs as either endpoint sources or endpoint destinations during the rebalance operation. Use the dest-rebalance-max parameter to specify the occupancy for the new cluster member. Refer to the Balancing section for details on occupancy and its calculation.

    If auto-rebalance is set to disabled, these three parameters can be ignored.

    Parameter values are numeric percentages within the range 0 through 100.

    source-rebalance-threshold specifies the minimum occupancy percent that identifies a clustered OCSBC as a source of endpoints during a rebalance operation. For example, using the default value of 50 (percent), any clustered SBC with an occupancy rate of 50% or more sheds endpoints during a rebalance. The SLB assigns these endpoints to the new cluster member.

    dest-rebalance-threshold specifies the maximum occupancy percent that identifies a clustered OCSBC as a destination for endpoints during a rebalance operation. Note that the default setting of 0 (percent), ensures that no pre-existing OCSBC gains endpoints during a rebalance.

    dest-rebalance-max specifies the maximum occupancy percent that the SLB transfers to the new cluster member during a rebalance operation. The default setting is 80 (percent). Should this threshold value be attained, the SLB distributes remaining endpoints to those OCSBCs identified as endpoint destinations by their dest-rebalance-threshold settings. 

    ORACLE(cluster-config)# source-rebalance-threshold 50
ORACLE(cluster-config)# dest-rebalance-threshold 40
ORACLE(cluster-config)# dest-rebalance-max 75
  6. If auto-rebalance is set to enabled, you can optionally use four additional parameters to fine-tune rebalance operational details.

    If auto-rebalance is set to disabled, these four parameters can be ignored.

    rebalance-request-delay specifies the interval (in milliseconds) between endpoint request messages sent from the SLB to a clustered OCSBC. As explained in the Rebalancing section, these messages request a list of endpoints that will be redistributed from the OCSBC to a new cluster member.

    By default, this parameter is set to 500 milliseconds.

    Setting this parameter to a higher value results in longer times for the completion of rebalancing; however longer durations provide more time for cluster member processsing of SIP traffic.

    rebalance-skip-ahead restricts the target set of OCSBC endpoints registration eligible for rebalancing to those whose re-registration is not imminent — that is, the registration is not scheduled within the number of milliseconds specified by the parameter setting. Setting this parameter to a non-zero value mitigates against the possibility of a race condition precipitated by a simultaneous endpoint removal generated by the OCSBC and the arrival of endpoint signalling on an SLB service port. The default setting (0 milliseconds) effectively makes the entire OCSBC endpoint set eligible for rebalancing.

    rebalance-max-refresh restricts the target set of OCSBC endpoints eligible for rebalancing to those whose re-registration is no further in the future than the time period (milliseconds) specified by this parameter— for example, assuming a parameter value of 6000, the target endpoint set is restricted to those whose re-registration is scheduled within the next 6 seconds.

    Because a re-balancing operation necessarily introduces a small window of unreachability for re-balanced endpoints, this parameter provides users with some degree of control over the period of time that a re-balanced endpoint may be unreachable.

    The default setting (0 milliseconds) effectively makes the entire SBC endpoint set eligible for rebalancing.

    rebalance-del-app-entries specifies when cached SIP entries for rebalanced endpoints are removed from the clustered OCSBC. The default setting (enabled) specifies that the SBC removes cached registration entries at the completion of the rebalance operation. When set to disabled, this parameter specifies that cached entries are retained after a rebalance operation, and subsequently removed from the cache by standard time-out procedures. 

    ORACLE(cluster-config)# rebalance-request-delay 750
ORACLE(cluster-config)# rebalance-skip-ahead 100
ORACLE(cluster-config)# rebalance-max-refresh 1000
ORACLE(cluster-config)# rebalance-del-app-entries enabled
  7. Three parameters, tunnel-fail-interval, tunnel-check-interval, and inactive-sd-limit maintain and monitor the IP-in-IP tunnels established between the SLB and clustered SBCs.

    tunnel-fail-interval specifies the interval (in milliseconds) between periodic keepalive messages sent from a clustered OCSBC to the SLB. If the SLB fails to receive a keepalive message within the specified period, it flags the tunnel as dead. By default, this parameter is set to 10000 milliseconds.

    tunnel-check-interval specifies the interval (in milliseconds) between SLB tunnel audits. During a tunnel audit, the SLB checks the status of each tunnel and removes all tunnels flagged as dead. By default, this parameter is set to 15000 milliseconds.

    If you change default settings for either parameter, ensure that the setting for tunnel-check-interval is greater than the tunnel-fail-interval setting.

    inactive-sd-limit specifies the maximum silent interval (defined as the absence of heartbeat traffic from any tunnel) seconds) before the SLB flags a cluster member as dead, and removes that OCSBC from the cluster. By default, this parameter is set to 1800 seconds (30 minutes). supported values are integers within the range 0 through 31556926 (365 days). 

    ORACLE(cluster-config)# tunnel-fail-interval 10000
ORACLE(cluster-config)# tunnel-check-interval 15000 
ORACLE(cluster-config)# inactive-sd-limit 900
  8. Use the session-multiplier and atom-limit-divisor parameters to specify optional, user-configurable numeric factors used in occupancy and occupancy rate calculations.

    session-multiplier provides a factor that when multiplied by an OCSBC’s licensed session limit, determines the maximum number of endpoints that the OCSBC can support (that is, its maximum occupancy).

    The default setting is 10; valid settings include any integer values within the range 1 through 100.

    Using the default setting, an OCSBC licensed for 32,000 concurrent sessions has a maximum theoretical occupancy of 320,000 endpoints.

    atom-limit-divisor provides another factor that can be used in occupancy and occupancy percent calculations. By default, occupancy calculations are based on endpoints (IP addresses), and do not take into account the fact that the same IP address can represent multiple users.

    The default setting is 1, which assumes a conservative 1-to-1 correlation between endpoints and users; valid settings include any integer values within the range 1 through 1000.

    Note:

    The SLB initially calculates a tentative maximum occupancy value, expressed as a number of endpoint addresses, for each clustered OCSBC. SLB calculations are based upon the licensed capacity of each cluster member, and the values assigned to the session-multiplier and atom-limit-divisor parameters. After calculating the tentative maximum occupancy value, the SLB compares this value to the value of the registration-cache-limit parameter as defined on the clustered OCSBC. If the value of registration-cache-limit is either 0, or greater than the tentative maximum occupancy value, the calculated value is retained as the occupancy ceiling. However, if the registration-cache-limit value is greater than 0, but less than the tentative calculation, the value of registration-cache-limit is used as the occupancy ceiling.

    Once an OCSBC has reached its maximum number of endpoints, the SLB removes it from the load balancing algorithm. These parameter settings should be changed only after careful examination of network conditions and behavior. 

    ORACLE(cluster-config)# session-multiplier 10
ORACLE(cluster-config)# atom-limit-divisor 1
  9. The ignore-tgt-svc-on-rebalance parameter is not currently supported, and can be safely ignored.
  10. Retain default settings for the red-port, red-max-trans, red-sync-start-time, and red-sync-comp-time parameters.
  11. Use done, exit, and verify-config to complete cluster configuration.