Support Support Downloads Knowledge Base Case Manager My Juniper Community

Knowledge Base

Search our Knowledge Base sites to find answers to your questions.

Ask All Knowledge Base Sites All Knowledge Base Sites JunosE Defect (KA)Knowledge BaseSecurity AdvisoriesTechnical BulletinsTechnotes Sign in to display secure content and recently viewed articles

[MX] FPC buffer memory increases significantly without a noticeable trigger

0

0

Article ID: KB36963 KB Last Updated: 28 May 2021Version: 1.0
Summary:

FPC buffer (LAN buffer) utilization can increase due to heavy traffic exchange between FPC/RE. This does not always indicate any issues as the implementation of the PFE memory management allows LAN buffer memory increase for a prolonged amount of time.

Symptoms:

Utilization of the buffer in the output of 'show chassis fpc' can increase significantly (from 25% to 55% for example) without any noticeable trigger.

Example:

user@router1-re0> show chassis fpc
                     Temp  CPU Utilization (%)   CPU Utilization (%)  Memory    Utilization (%)
Slot State            (C)  Total  Interrupt      1min   5min   15min  DRAM (MB) Heap     Buffer
  0  Online            37     26          0       25     25     26    3584       27         26
  1  Online            38     26          0       24     23     23    3584       28         55
Cause:

The LAN buffer is used to transfer packets between RE and FPC. The LAN buffer is divided into several sheaves for packets with different lengths. Each sheaf gets initial memory allocation from the LAN buffer.

Once there are a lot of packets coming towards the RE from FPC (or from RE to FPC) and the current value of the buffer for a particular packet size (sheaf) is not enough, then more memory is provided to this sheaf from LAN buffer to manage all the packets (the value of the current sheaf buffer is tracked by Total). The size of each sheaf can be seen from FPC shell with the following command:

NGMPC5(router1-re0 vty)# show packet
Packet Sheaves:
 
ID Total(Init, High)  Free( Min,  Max)  Nall/Nfre/Nemp Name
  6  1536(1536, 1536)  1535(1024, 2048)  1536/   0/   0 packet(100)
  5  1074( 768, 1079)  1023( 512, 1024)  2040/ 966/   0 packet(600)
  4  14948(6144, 14948)  13324(3072, 24576)  14948/   0/   0 packet(1500)
  3   513( 384,  513)   513( 384, 5184)   513/   0/   0 packet(5000)
  2     6(   6,    6)     6(   4,    8)     6/   0/   0 packet(10000)
  1     2(   1,    2)     2(   1,    3)     2/   0/   0 packet(18000)
  7     0(   0,    0)     0(   0,    0)     0/   0/   0 packet(misfits)
  8   256( 256,  256)   256( 128, 1024)   256/   0/   0 packet(clones)


When the number of packets in the system is decreased, the previously allocated memory is not just released to the initial level because memory allocation/releasing is an expensive operation and can lead to memory fragmentation. Instead, the memory is released if the number of free packets for a particular sheaf is more than the Free Max value (so too many not-used memory in a particular sheaf). So the memory utilization can go up due to some burst of traffic and it does not go down later on.

Solution:

This behavior is FAD (functioning as designed) and can be ignored. In many cases, the buffer memory increase should be investigated if the utilization crosses 80% and the percentage grows monotonically over time. It may be possible to check which packets caused the LAN buffer memory increase. In that case, contact your JTAC Representative to open a service request.

Comment on this article > Affected Products Browse the Knowledge Base for more articles related to these product categories. Select a category to begin.

Getting Up and Running with Junos

Getting Up and Running with Junos Security Alerts and Vulnerabilities Product Alerts and Software Release Notices Problem Report (PR) Search Tool EOL Notices and Bulletins JTAC User Guide Customer Care User Guide Pathfinder SRX High Availability Configurator SRX VPN Configurator Training Courses and Videos End User Licence Agreement Global Search