Configure Options on Managed Devices for Better SNMP Response Time
The following sections contain information about configuration options on the managed devices that can enhance SNMP performance:
Enable the stats-cache-lifetime Option
Junos OS provides you with an option to configure the length of time (in seconds)
the interface stats are cached. If the NMS queries again for the same interface
within the cache time, the same data is returned. If the NMS queries after the
cache time, the cache is no longer valid, fresh data is fetched from the lower
layers, and the cache timestamp is updated. The default
stats-cache-lifetime
is 5 seconds. This can be tuned as per
the polling frequency.
Reducing the value of the stats-cache-lifetime option results in more queries and can impact performance. To get the live uncached statistics, set the value of the stats-cache-lifetime option to 0. However, this is not recommended since it completely disables the caching feature and impacts performance.
Filter Out Duplicate SNMP Requests
If a network management station retransmits a Get
,
GetNext
, or GetBulk
SNMP request too
frequently to a device, that request might interfere with the processing of
previous requests and slow down the response time of the agent. Filtering these
duplicate requests improves the response time of the SNMP agent. The Junos OS
enables you to filter out duplicate Get
,
GetNext
, and GetBulk
SNMP requests. The
Junos OS uses the following information to determine if an SNMP request is a
duplicate:
-
Source IP address of the SNMP request
-
Source UDP port of the SNMP request
-
Request ID of the SNMP request
By default, filtering of duplicate SNMP requests is disabled on devices running the Junos OS.
To enable filtering of duplicate SNMP requests on devices running the Junos OS,
include the filter-duplicates
statement at the [edit
snmp]
hierarchy level:
[edit snmp] filter-duplicates;
Exclude Interfaces That Are Slow in Responding to SNMP Queries
An interface that is slow in responding to SNMP requests for interface statistics can delay the kernel responses to SNMP requests. You can review the mib2d log file to find out how long the kernel takes to respond to various SNMP requests. For more information about reviewing the log file for the kernel response data, see “Checking Kernel and Packet Forwarding Engine Response” under Monitoring SNMP Activity and Tracking Problems That Affect SNMP Performance on a Device Running Junos OS.
If you notice that a particular interface is slow in responding and think that it
is slowing down the kernel from responding to SNMP requests, exclude that
interface from the SNMP queries to the device. You can exclude an interface from
the SNMP queries either by configuring the filter-interface
statement or by modifying the SNMP view settings.
The following example shows a sample configuration for excluding interfaces from
the SNMP Get
, GetNext
, and
Set
operations:
[edit] snmp { filter-interfaces { interfaces { # exclude the specified interfaces interface1; interface2; } all-internal-interfaces; # exclude all internal interfaces } }
The following example shows the SNMP view configuration for excluding the interface with an interface index (ifIndex) value of 312 from a request for information related to the ifTable and ifXtable objects:
[edit snmp] view test { oid .1 include; oid ifTable.1.*.312 exclude; oid ifXTable.1.*.312 exclude }
Alternatively, you can take the interface that is slow in responding offline.