Description

The following syslog message is seen on the QFX device acting as a leaf in the EVPN/VXLAN deployment: 

rpdd l2ald_client_send: RPDF L2ALD pipe blocked

Symptoms

The following messages are seen in the syslog continuously:

Dec  7 04:02:13.867  Device l2ald[6203]: rpdd:pipe is unblocked
Dec  7 04:02:13.867  Device l2ald[6203]: rpdd: pipe is now flow blocked
Dec  7 04:02:17.258  Device l2ald[6203]: rpdd:pipe is unblocked
Dec  7 04:02:59.466  Device l2ald[6203]: rpdd: pipe is now flow blocked
Dec  7 04:03:00.906  Device l2ald[6203]: rpdd:pipe is unblocked
Dec  7 04:03:00.908  Device l2ald[6203]: rpdd: pipe is now flow blocked
Dec  7 04:03:04.057  Device l2ald[6203]: rpdd:pipe is unblocked
Dec  7 04:03:08.727  Device l2ald[6203]: rpdd: pipe is now flow blocked
Dec  7 04:03:08.988  Device l2ald[6203]: rpdd:pipe is unblocked
Dec  7 04:03:50.977  Device l2ald[6203]: rpdd: pipe is now flow blocked
Dec  7 04:03:51.254  Device l2ald[6203]: rpdd:pipe is unblocked
Dec  7 04:03:51.255  Device l2ald[6203]: rpdd: pipe is now flow blocked

The output of the command ' show system process extensive' : 

 PID USERNAME  THR PRI NICE   SIZE    RES STATE   C   TIME    WCPU COMMAND
 6509 root        1  52    0   923M   430M kqread  0  29:17  80.18% rpd  
 6519 root        1  90    0   525M 65956K CPU1    1 113:15  61.96% l2ald   
 3825 root        1  20    0   463M  8860K select  1 601:40   1.37% ppmd
 3843 root        1  20    0   463M  5912K select  1 695:07   1.07% bfdd
 3829 root        2 -26  r26   539M 28960K nanslp  1 390:21   0.88% chassisd
 3857 root        1  20    0 89960K   672K RUN     1 316:21   0.39% fxpc
 6510 root        1  20    0   528M  6208K select  1  82:47   0.00% rmopd
 6526 root        1  20    0   504M 23168K select  0  36:32   0.00% pfed

Solution

The above error messages can be seen during high CPU utilization scenarios specifically in the VXLAN/EVPN environment. The error indicates the communication between the l2ald (layer 2 address learning daemon) is not able to communicate with the RPD (routing daemon).

This error can be seen either if RPD is very high (around 70-80%) and at the same time the l2ald is trying to send the MAC learning information to RPD for learning in the EVPN table. Either of the two daemons, if busy, can result in pipe block messages.

Use the following KB articles to troubleshoot high CPU.

In the above scenario, we need to troubleshoot the cause of the high CPU on the device. Once the CPU utilization (rpd and l2ald) comes down, the above error messages will stop appearing.

Contact your JTAC Representative for further assistance.