N3K异常重启（案例）

在实际的情况下，有时候会遇到设备无故重启的问题，这个时候，我们需要判断一下重启的根本原因是什么，是否有规避的方法等。

这里记录了几个N3K异常重启的问题。

案例1：

设备型号：N3K-C3048TP-1GE

OS版本：6.0(2)U4(4)

阶段1：采集信息。我们需要采集的信息如下：

1、收集core文件

N3K# show core

2、收集tech文件

#tac-pac bootflash:

如此会在bootflash自动创建一个名为“show_tech_out.gz”的文件，然后可以通过ftp或者tftp方式导出

阶段2：分析（这里是一个故障的实例）

========

问题描述

========

N3K unexpectedly reboot in 2016 Dec 10 08:23

2016 Dec  10 08:23:20.318 N3K-01 %$ VDC-1 %$
%SYSMGR-2-SERVICE_CRASHED: Service "bcm_usd" (PID 3257) hasn't caught
signal 6 (core will be saved).
2016 Dec  10 08:23:20.336
N3K-01 %$ VDC-1 %$ %SYSMGR-2-HAP_FAILURE_SUP_RESET:
System reset due to service "bcm_usd" in vdc 1 has had a hap failure

========

分析过程

========

1. 重启原因
`show system reset-reason`
----- reset reason for Supervisor-module 1 (from Supervisor in slot 1) ---
1) At 948672 usecs after Thu Dec 10 08:23 2016
    Reason: Reset triggered due to HA policy of Reset
    Service: bcm_usd hap reset
    Version: 6.0(2)U1(2)

2. 启动前后
Software
  BIOS:      version 1.2.0
  loader:    version N/A
  kickstart: version 6.0(2)U4(4)
  system:    version 6.0(2)U4(4)
  Power Sequencer Firmware:
             Module 1: version v4.4
  BIOS compile time:       08/22/2014
  kickstart image file is: bootflash:///n3000-uk9-kickstart.6.0.2.U4.4.bin
  kickstart compile time:  3/12/2015 0:00:00 [03/12/2015 17:34:09]
  system image file is:    bootflash:///n3000-uk9.6.0.2.U4.4.bin
  system compile time:     3/12/2015 0:00:00 [03/12/2015 20:28:55]

Hardware
  cisco Nexus 3048 Chassis ("48x1GE + 4x10G Supervisor")
  Intel(R) Celeron(R) CPU        P450 with 3665256 kB of memory.
  Processor Board ID FOCXXXXXXXX

Device name: N3K-01
  bootflash:    2007040 kB

Kernel uptime is 0 day(s), 08 hour(s), 37 minute(s), 21 second(s)

Last reset at 948672 usecs after  Thu Dec 10 08:23 2016

Reason: Reset triggered due to HA policy of Reset
  System version: 6.0(2)U1(2)
  Service: bcm_usd hap reset

我们可以看到系统重置是由bcm_usd hap重置引起的。 如果可以解码core文件，我们可以发现6.0（2）U1（2）这个版本也命中bug，bug ID：CSCuo88390（Nexus 3k bcm_usd crash after parity error）的错误。 从show version，我们看到当前版本是6.0.2.U4.4，这是修复了bug的版本。

bug的详细信息如下：

Nexus 3k bcm_usd crash after parity error

CSCuo88390

 

Description

Symptom:
Nexus 3000 switch may reboot unexpectedly. Last reset reason is recorded as 'bcm_usd hap reset'.

`show system reset-reason`
----- reset reason for Supervisor-module 1 (from Supervisor in slot 1) ---
1) At 369448 usecs after Tue May 6 16:49:37 2014
Reason: Reset triggered due to HA policy of Reset
Service: bcm_usd hap reset
Version: 6.0(2)U1(1a)

Conditions:
The crash occurs while recovering from a parity error. This bug will not be seen if a parity error does not occur first.

Workaround:
Not known at this time.

Further Problem Description:

=======

建议

=======

如果该错误在短时间内持续发生，建议直接更换设备，该错误可能是由于硬件导致，如观察段时间内并未发生，则可以忽略这个问题。

案例2：

1、采集信息

采集的信息还是和案例1中一直，查看core文件，采集tech文件

2、分析

异常log如下：

2016 Dec 11 01:23:18.782 N3K-01 %$ VDC-1 %$ %BCM_USD-2-BCM_USD_NOTIFICATION_IMP: FAST REBOOT DISABLED

2016 Dec 11 01:23:18.782 N3K-01 %$ VDC-1 %$ %USER-2-SYSTEM_MSG: CLIS: loading cmd files begin  - clis

2016 Dec 11 01:23:18.782 N3K-01 %$ VDC-1 %$ %USER-2-SYSTEM_MSG: CLIS: loading cmd files end  - clis

2016 Dec 11 01:23:18.782 N3K-01 %$ VDC-1 %$ %USER-2-SYSTEM_MSG: CLIS: init begin  - clis

2016 Dec 11 01:23:18.782 N3K-01 %$ VDC-1 %$ %USER-0-SYSTEM_MSG: Starting bcm_attach - bcm_usd

2016 Dec 11 01:23:18.782 N3K-01 %$ VDC-1 %$ %USER-0-SYSTEM_MSG: Finished bcm_attach... - bcm_usd

2016 Dec 11 01:23:18.782 N3K-01 %$ VDC-1 %$ %VDC_MGR-2-VDC_ONLINE: vdc 1 has come online

2016 Dec 11 01:23:18.782 N3K-01 %$ VDC-1 %$ %ETHPC-2-PORTS_UP:

其次，从show core中没有看到有core文件

再看重启原因：

`show system reset-reason`

----- reset reason for Supervisor-module 1 (from Supervisor in slot 1) ---

1) No time

Reason: Unknown

Service:

Version: 6.0(2)U4(4)

2) No time

Reason: Unknown

Service:

Version: 6.0(2)U4(4)

3) No time

Reason: Unknown

Service:

Version: 6.0(2)U4(4)

Mon Dec 11 01:23:18 2016: Card Uptime Record

----------------------------------------------

Uptime: 28938, 0 days 6 hour(s) 3 minute(s) 12 second(s)

Reset Reason: Unknown (0)

Reset Reason SW: Unknown (0)

Reset Reason (HW): Unknown

Card Mode..........................: Runtime

这个问题很可能是和PSU（Power Supply Unit）的的关系，超出规定范围的电源电压值，可能是电源相关程序触发设备重启，以防止进一步的损坏。“Unknown”的重置原因表面设备由于电源序列发生器组件重新加载：由于电源不足或PSU出现问题。

可以看到，目前设备的PS2没有up.

Power Supply:

Voltage: 12 Volts

-----------------------------------------------------------

PS Model Input Power Current Status Type (Watts) (Amps)

-----------------------------------------------------------

1 N2200-PAC-400W-B AC 396.00 33.00 ok

2 -- -- -- -- fail/not-powered-up

由此，这个重启问题可能是由于电源硬件问题引起的。建议检查外部电源是否有问题，是否有其他的设备和该N3K连接相同的PDU，还可以检查次设备日志，来判断当时是否有电源问题。如果没有外源的问题，那建议更换电源。

案例3：

1、采集信息：

采集的信息和案例1,2中提到的一样。

2、分析：

2.1 先检查logging，观察FAN，PS是否log报错。

2.2 设备重启原因是 I2C bus 拥塞，导致设备认为 FAN 被拔出，这种应该不是硬件问题。 I2C 通道负责在主板、电源、风扇、SFP 等之间的通信；如果 I2C 通道拥堵，可能导致设备不能正确读取 FAN 状态。

%KERN-2-SYSTEM_MSG: [23949956.295550] clk_flush: Couldn't Clear Bus – kernel << I2C bus 拥塞

CST: %PFMA-0-SYS_SHUTDOWN_FAN_REMOVAL: System shutdown in 120 seconds due to fan missing or failed

CST: %NOHMS-2-NOHMS_ENV_ERR_FAN_READ: System minor alarm in fan tray 1: failed to read TACH << 风扇读取失败

Wed Oct 10 15:10:38 2018: Card Uptime Record

----------------------------------------------

Uptime: 3674, 0 days 1 hour(s) 1 minute(s) 14 second(s)

Reset Reason: Unknown (0)

Reset Reason SW: Powered-down due to fan policy trigger (104)

Reset Reason (HW): Unknown

Card Mode..........................: Runtime

N3K 重启，启动以后，风扇和电源状态都正常：

Fan:

------------------------------------------------------

Fan Model Hw Status

------------------------------------------------------

Fan-1 NXA-FAN-30CFM-B -- ok

Fan-2 NXA-FAN-30CFM-B -- ok

Fan-3 NXA-FAN-30CFM-B -- ok

Fan-4 NXA-FAN-30CFM-B -- ok

PS-1 N2200-PAC-400W-B -- ok

PS-2 N2200-PAC-400W-B -- ok

Power Supply:

Voltage: 12 Volts

-----------------------------------------------------------

PS Model Input Power Current Status Type (Watts) (Amps)

-----------------------------------------------------------

1 N2200-PAC-400W-B AC 396.00 33.00 ok

2 N2200-PAC-400W-B AC 396.00 33.00 ok

建议：无已知的 bug匹配，建议升级推荐版本。如果再次遇到同样问题，可以更换机箱。

https://www.cisco.com/c/en/us/td/docs/switches/datacenter/nexus3000/sw/recommended_release/b_Minimum_and_Recommended_Cisco_NX-OS_Releases_for_Cisco_Nexus_3000_Series_Switches.html

其他log分析：N3K 重启之前，有一些其他 kernel 和电源相关的 log，是显示性问题，实际上没有影响

%KERN-3-SYSTEM_MSG: [23949949.272393] PFM: Error while trying to service 0x2aa, err=-14 - kernel

https://bst.cloudapps.cisco.com/bugsearch/bug/CSCur58188/?reffering_site=dumpcr

CST: %PFMA-2-PS_FAIL: Power supply 1 failed or shutdown (Serial number DCA00000000)

CST: %NOHMS-2-NOHMS_DIAG_ERR_PS_FAIL: System minor alarm on power supply 2: failed or not powered up

https://bst.cloudapps.cisco.com/bugsearch/bug/CSCva03834/?reffering_site=dumpcr

https://bst.cloudapps.cisco.com/bugsearch/bug/CSCux99539/?reffering_site=dumpcr

设备一直报错：

%$ VDC-1 %$ Oct 10 19:36:06 %KERN-2-SYSTEM_MSG: [ 5375.470604] clk_flush: Couldn't Clear Bus - kernel

最终该问题摘要：

1、有两台N3K,当重启Device2时，Device1无法转发信息，当Device2恢复后，一切正常了。

2、Device2重启时异常：流量从一个接口进来，但是没有从对应的接口出去。在冲Device1之后，问题得到解决。

3、建议：目前使用的6.0(2)A6(1a)版本太低，建议升级

https://www.cisco.com/c/en/us/td/docs/switches/datacenter/nexus3000/sw/recommended_release/b_Minimum_and_Recommended_Cisco_NX-OS_Releases_for_Cisco_Nexus_3000_Series_Switches.html?dtid=osscdc000283