This post is more than 5 years old
143 Posts
0
51142
How to verify link-level flow control is working on EqualLogic?
I recently needed to disable DCB on various switches that did not fully support DCB. I now want to ensure that the EqualLogic arrays are properly using link-level flow control. I remember seeing a message in the event log a while back that said that flow control was enabled on all interfaces, but I am wondering if there is an easier and more definitive way for me to check whether flow control is in fact enabled and working on the EqualLogic arrays. Firmware version is 7.0.9. I have DCB enabled in the EqualLogic group (as recommended by Dell) and disabled on the Ethernet switches. Thanks.
Anonymous
5 Practitioner
5 Practitioner
•
274.2K Posts
0
December 11th, 2014 09:00
Hello,
You need to check the switches, they should show it being enabled or not. EQL arrays will never SEND a pause frame to the switch, it will however, respond to flowcontrol if sent to it from the switch.
Note: Not all switches support send flowcontrol.
Regards,
boyler05
143 Posts
0
December 11th, 2014 09:00
Thanks Don. Is it a correct statement that if the EQL array says "DCB: off" as a result of the command "member select XXXX eth select X show", it means that traditional 802.3x flow control messages will be correctly responded to by the EQL? Second, on the switch side, the switch port physically connected to the EQL should at a minimum say "flowcontrol tx on", correct? If it says "flowcontrol rx on tx on", that would be fine, too, I presume, but if it says "flowcontrol rx on tx off", that would be the opposite of what we would want, correct?
Anonymous
5 Practitioner
5 Practitioner
•
274.2K Posts
0
December 11th, 2014 09:00
Re:1 Yes. With DCB off, pause frames will be honored by the array. There's no setting to enable or disable it.
Re: 2. Correct. If the switch can't send flowcontrol it will show as off.
However, if you have a good switch for iSCSI, it's very rare for a switch to send a pause frame. Pause frames are not end-to-end, only point to point. Typically it is the host that needs it on reads. The array/switch on sustained READS can usually overwhelm a host. It then sends a pause frame to the switch port its connected to. Then only that port or port(s) are affected. When you send a pause frame that entire port is shutdown, not just the sessions for that host. So if you pause an array port, means multiple servers are affected.
Regards,
boyler05
143 Posts
0
December 11th, 2014 11:00
Thanks Don, that really helps. So in the case of a VMware host, it would be fairly likely for the VMware host to SEND a pause frame to the switch, which means the switch port connected to the VMware host would need "receive" flowcontrol enabled (flowcontrol rx on), correct? Likewise, if it is rare for the switch to send a pause frame to the VMware host, then it would be OK if the switch port has "transmit" flowcontrol disabled (flowcontrol tx off), correct?
Conversely, since the EQL will never SEND a pause frame, the switch port connected to the EQL could have "flowcontrol rx off" and "flowcontrol tx on", correct? In other words the flow control settings for the VMware host port and the EQL port would essentially be opposite settings?
Anonymous
5 Practitioner
5 Practitioner
•
274.2K Posts
0
December 11th, 2014 11:00
That's correct. However, don't be surprised if EQL reports TX and RX as ON. It responds that it does, even though it will never actually send a PAUSE frame.
harlan.stanley
2 Posts
0
January 13th, 2015 09:00
Hi, I have the following configuration, and read through this post wondering if I could replicate, However, I can't!?
On My Force10 S50n:
interface GigabitEthernet 0/23
description EQL iSCSI Port
no ip address
mtu 9216
switchport
flowcontrol rx on tx on
spanning-tree rstp edge-port
no shutdown
On the Equallogic Sumo
(member_Sumo2-96 eth_1)> lldp show
____________________________ Eth LLDP Information _____________________________
Name: eth1 LLDP-State: active
System-Name: force10S50n HardwareAddress: 00:01:**:**:16:**
System-Description: Dell Force10 Real Chassis-Id: 00:01:**:D6:16:**
Time Operating System Software. Dell Port-Id: GigabitEthernet 0/23
Force10 Operating System Version: Port-Description: EQL iSCSI Port
1.0. Dell Force10 Application
Software Version: 8.4.2.7. Copyright
(c) 1999-2012 Dell Inc. All Rights
Reserved.Build Time: Thu Sep 27
14:03:07 PDT 2012
I'm getting many messages like this on all my vHosts:
Device naa.6090a05840a81742cc2985a411124123412 performance has
deteriorated. I/O latency increased from average value of 3451
microseconds to 69409 microseconds.
warning
1/13/2015 11:16:26 AM
*******@****.com
Device naa.6090a05840a81742cc2985a411124123412 performance has
improved. I/O latency reduced from 69409 microseconds to 13608
microseconds.
info
1/13/2015 11:16:31 AM
*******@****.com