May 19 00:31:37:N:System: Module 1 powered off
May 19 00:31:37:I:System: Interface L2-SW04-e2/3/1 1/8, state down – card down
May 19 00:31:37:I:System: Interface L2-TSW-E24 1/7, state down – card down
May 19 00:31:37:I:System: Interface L2-VDX-TE1/0/54 1/6, state down – card down
May 19 00:31:37:N:System: Module down in slot 1, reason CARD_DOWN_REASON_POWERED_OFF_SYS_MONITOR. Error Code 0
May 19 00:31:37:D:System: TM Health Monitoring detects an issue in slot 1 ppcr 0 Reg Offset 00000800 Value 00000005
May 19 00:31:37:D:System: Module reset in slot 1, triggered by TM Health Monitoring
May 19 00:31:36:A:System: LP1/TM0: All ports down due to dram crc
May 19 00:31:36:I:System: Interface L2-SW03-e1/3/1 1/3, state down – ingress dram crc
May 19 00:31:36:I:System: Interface L2-TSW-E24 1/2, state down – ingress dram crc
May 19 00:31:36:I:System: Interface L2-VDX-TE1/0/54 1/1, state down – ingress dram crc

I’m running NOS 4.0.1, tried to create a port-channel on this VDX then it crashed.

Found this in the syslog ras

2015/02/17-09:17:36, [SSMD-1001], 504432, DCE, ERROR, VDX, Failed to allocate memory: (ssm_nsm_link_attr_update).

2015/02/17-09:17:35, [SSMD-1001], 504429, DCE, ERROR, VDX, Failed to allocate memory: (ssm_ha_send_event).

It wouldn’t join back into the cluster because of mismatched configuration. I had to change it to default config then rejoin the cluster.

Today I had to troubleshoot packet loss on a VMware virtual machine on Cisco UCS. I went through to the port group and saw the teaming was set to route based on IP Hash. They had set it this way for some Microsoft Load Balancing. If you take a look at the KB on VMware site, it states that the physical switch ports be combined into an EtherChannel. Since the ports used in the port group were on two different Fabrics and I didn’t know of a way to setup Etherchannel on UCS on two different fabrics to a server, I ended up just changing the second port to standby.

Quick Summary:

Problem: Packet loss using IP hasing Load balancing because can’t create etherchannel to server from UCS with two different fabrics.

Kind of a solution: change the 2nd port to standby.