Effect of loss of link IOM UCS on CF in Port-channel mode
Hello
We recently discovered that when UCS Server Ports are port-channel mode, loss of a member link causes CF traffic disruption in a VMware environment. Host VMware never sees a path down and his guest enter a time-out period of IO wait. Once expired the time-out of comments, a SCSI reset restores connectivity of storage. We responded to several customers and said that this is the expected behavior. This is of concern to us and how it affects the design decisions.
Any contribution is appreciated.
Thank you.
It is true, that you expect the SCSI timeout, and then the operating system must redo a flogi. Of course, I guess that you have CF multipathing installation and work; Therefore, no disruption to traffic.
Without port-channel, the vhba on the failing tissue will go down; and the management of multi-channel FC is the only solution.
Tags: Cisco DataCenter
Similar Questions
-
Hello
I hope someone can help with this,
I have a Setup with 2 Nexus of 5548 connected via a link of vpc peers and a keepalive, these two nexus switch have a link of standard VPC to a 6500 upstream. It works fine,
Then, there are several native FC connections of the 5548 to of 6248. I know that there is the possibility to configure the channel port of CF on the UCS 2.0 but it is possible to link 5548 CF interfaces in a port channel san through a vpc between the nexus of 5548.
so, for example, I can port-channel which follows through a vpc or is a channel of port san, limited to the same nexus only switch ports.
Fc1/31 & fc1/32 fabric sent to-> NX5K - 1 Port fc1/31 & NX5K - 2 Port fc1/32 (san port-channel 10 on a VPC)
Fabric B fc1/31 & fc1/32 passed to-> NX5K - 1 Port fc1/32 & NXFK - 2 Port fc1/31 (san port-channel 11 on a VPC)
Thank you
Ray.
It is not possible. VPC is the only Layer 2 ethernet port channel. Port channel San cannot be built through vpc
Sent by Cisco Support technique iPhone App
-
Setting port channel between UCS - FI and MDS 9124 (Mode F)
Dear team,
We tried to create the channel of port between UCS FI and MDS 9124
But the port channel do not take action in mode F on MDS 9124
FI is in host FC end Mode
We have allowed FC uplink on FI trunking
We have activated NPIV on MDS
We have activated the MDS trunk
FI and MDS in default VSAN
To check that we have changed the way FI FC channels mode and switch port became active, but in E mode
When we enabled CF trunking of uplink on the port mode FI channels and FC Switching became active in mode TE
but in both cases above, showflogi database shows WWPN of SAN alone does not have the any fi.
How to achieve this?
Have read that no need to change the mode switching mode of CF swicthing and keep FC Endhost way
SO how to channel ports with mode F MDS and FI (Display Mode as NProxy)
What is it has nothing to do with the MDS NX - OS version? (https://supportforums.cisco.com/thread/2179129)
If yes how to put as license for ports came with the camera and we don't have any CAP/PAK or license file as she came
with license
Also, we saw 2 files available for download (m9100-s2ek9-kickstart - mz.5.2.8 .bin and m9100-s2ek9 - mz.5.2.8 .bin b b) to use
Thanks and greetings
Jose
Hi Jo Bo.
What version of the software if your MDS race?
On your UCS do connect nxos and show ficelleStringString ficelleT inteface and find the mac address.
It is possible that you could be hitting the bug below. If this is the case, you may need to update the firmware on your MDS.
Add MAC YES '002a6a', '8c604f', '00defb' for 5 k/UCS-FI
http://Tools.Cisco.com/support/BugToolKit/search/getBugDetails.do?method=fetchBugDetails&bugId= CSCty04686
Symptom:
Link change cannot connect any other Nexus or other Cisco Switch in mode VAN with a port channel F. Question can be viewed in older versions of 5.1
5.1.3.N1.1a
but not later
5.1.3.N2.1c
Release. Question is also found in
5.2 (1) N1 (1)
and
6.0 (2) N1 (1)
and later versions.
Conditions:
Nexus configured for the link SAN PortChannels or NPIV Nexus mode connected to the UCS via regular F port channel where UCS VAN VAN edge mode switch: YES switch manufactured FI or another Cisco UCS Port WWN: xx:xx:00:2 has: 6a: xx:xx:xx or xx:xx:8 c: 60:4f:xx:xx:xx
Workaround solution:
Turn-off on Nexus 5 k TF-port question link mode does not happen with standard F-PORT SAN to remove Portchannel config
Other Description of the problem:
To check question collect please see the flogi-event history internal errors whenever the port is attempted OLS, AMENDMENTS, PBA counters will increment. This can be determined via the following output, view port internal info to see all the internal-historic port of error events
-
Hello
One of my customers wonder about linking the FI 2 L1/L2 loss.
In my mind, this case should not be here, why we have 2 links but I don't know what happened exactly in this case... the 2 FI become master of the UCS cluster...
any idea or already Tester?
THX
The L1/L2 are by data management engine (DME) to synchronize the config only - there is no path of data on these links.
In case these links both go down, there is a rear mechanism built into the system to prevent split brain. Each frame has a SEEPROM (flash). Within the SEEPROM, there is a shared area each FI can read & write to and two subzones; an inmate by each FI in which only FI can write to. The subareas are used to store information about the config DB fails to a bank owner in the same, changes are made and it is reintroduced. You will avoid 'stale' config overwrite any changes made during a FI was disconnected.
If there is a failure in the cluster llink (L1/L2), IF read and write small counters in the shared area. This serves as a secondary heartbeat to warn the other FI is always active. Meanwhile, the cluster starts in a State of 'Failed-Link' and prevents any election primary/secondary. Everything else as-if - primary primary, secondary rest is secondary.
Kind regards
Robert
-
After effects CS 5 Download link
Need the download link for after effects CS 5
CS5 - http://helpx.adobe.com/creative-suite/kb/cs5-product-downloads.html
-
Loss of links HTML and Word to PDF conversion
Hi, when I convert a Web page or a word to PDF document, the links show as blue underlined fonts, but the URL is missing, and so the link is not active. I checked that "add links" is checked in my conversion settings. The same thing happens when I print to Adobe PDF, or when I select the menu "convert to PDF".
I use Adobe Pro X and Word 2003. My OS is Windows XP. Any advice would be appreciated.
Thanks, Paul
First of all, by using the Adobe PDF printer to create PDF won't get you links.
Second, try to drag your HTML files directly in Acrobat to see what is happening. If you aren't yet getting links then I would repair or reinstall Acrobat X
-
IOM pinning via the port-channel
Hello
It seems to me that, with UCS manager 1.4 and higher, it is possible to setup the pinning of the IOM through the port-1-channel.
I couldn't find any configuration guide for this kind of configuration.
Can you help me or inform me on this subject?
Thanks in advance.
Nicolas.
Hi Nicolas,.
+ The feature is only available with the 2.0 and 2.0 material, i.e. a combination of FI and IOM 22xx 62xx.
the option is available on the same page as the page of chassis discovery strategies.
-> Hardware tab select hardware-> policies-> global Polocies-> preferences to group link (right)
. / Afonso
-
Hi all
I tried to connect the FI FC port to the Nexus 5 K. YEW has been configured in host end (VAN) mode, while the N5K has been configured as NPIV switch (NPV core).
After trying many things, the SAN PortChannel link on both FI and N5K cannot grew up, stuck in the initializing state.
If the N5K and FI was not grouped in PortChannel, connection set successfully.
In addition, I found and who issues the FI when running in NPV mode, the trunk cannot be configured as trunk mode on (if configured, so the connection will get stuck in the initialization of State too).
Attached are the files of configuration on FI and N5K
Also, I tried to connect the FI to MDS 9222I and had the same problems that connect to N5K.
Can anyone help?
Comparation while I connect the Nexus 5 K as NPV switch and MDS 9211i as mode NPIV, the PortChannel SAN and circuits (N5K as TNP) and MDS TF was successfully raise.
Thank you
GOFI
Gofi Hello,
As interface comes from the individual links and N5K running version 5.1 (3) N1 (1), he may run into the following faults
http://Tools.Cisco.com/support/BugToolKit/search/getBugDetails.do?method=fetchBugDetails&bugId= CSCty04686
If so, N5K software must be upgraded to a version that has the fix for this flaw.
HTH
Padma
-
UCS to SAN Brocade FC Switch mode
Is it possible to connect a network SAN FC Brocade for interconnections of fabric once to have them put in place mode Switch CF?
Thank you
Hello Dani
No! We do not support interop modes!
the only solution is running FI in NPV mode and Brocade in Access Gateway (NPIV)
Why is this not possible? We have many installations of UCS, connection to a cloth of Brocade with NPV mode
Walter.
-
Why the pdf links don't work for Muse in preview mode?
I want to add PDFs on my Muse site, but they are not opening in preview mode. I have the PDF files loaded as files in my property and make them related to the text on the page. But, when I hit the preview mode, the links do not work. Any tips? They were always open when the site is published?
They will not work if you use the file > Preview in browser or file Page > Site overview in the browser. They do not work for the preview window integrated into the Muse.
-
Curious to know if the cutting/possible ovevolting on one of the analog channels would affect the performance of other similar channels. I have two systems, one where signals don't clip and the other channels are still operable. The other has some problems of cutting (which fade eventually btw) on some of the ins anlog. One with the cutting on other channels are apparently unusable while those who did no cropping entry seems to be well.
I seem to be affected by a situation in which hanging on an analog channel makes so that other analog channels do not work.
Curious to know if I can see what is expected?
Anyone have an overview here?
This is how it is supposed to work?
We do not guarantee that all channels will work normally when you're in a State of fault/over-voltage.
Since it is a MUX architecture'ed, that the channels are analyzed in order in the amplifier of shared instrumentation and ADC. If the instrumentation amplifier is on track for channel 0 (for example), it will take some time to recover from this error after the next channel is enabled in. This can cause a number of incorrect following channels (possibly indicated as being on track).
Of course, we do not recommend to use this device in an error condition. When you said "analog channels does not" what do you mean? The data on track?
Some things that might help:
-Slower sample which allows more time to settle the instrumentation amplifier
-Skip the chain incriminated in the scan list
-Make sure that the offending channel is set to the widest range of voltage (+/-10 v)
-
Often I want to record a software simulation in demonstration, training and evaluation, or at least in two modes.
After registration, I have to make some changes, for example, I want to change "click field Contact must be completed" to "in the Contact field, click the fill button. So if I want to use videos as a demo, training and evaluation, I have to review even three times. Is there an easier way to do it?
N ° you must edit each output separately.
-
10G between adpater and blades for VIC1240 + 1280
As my other discussion suggested, I VIC 1240 and 1280 cards installed on the B200 M3 blade.
UCS automatically creates 4x20G port-channel between the adapter and the blades. We have therefore 80 G total n/b to the INSTITUTION of the blade. Side has 40 G and 40 B side
I have a delicate question: if I could change the 4 ports-channels to 8 individual links while I have granular control to 10G does level not 20G
Can anyone help to answer the question?
No, these are internal pc, automatically. Be aware that a single stream is limited to 10G, more if you have not configured a pc between the IOM and fi, you are limited by this public static pinning (IOM - FI) to 10G anyway. PC between IOM and fi requires the second generation material, for example. 62xx and 22xx.
-
I have two hosts (vhost1 = ESX 4.0, vhost2 = 5.5 ESX) both are configured for the Group and both connected to a switch of 1544 Adtran L3 (named sw01). vhost1 has been in production for several years with the config below and has been solid as a rock. I'm trying to reproduce the installation of grouping/collage on vhost2 and it does not work when are associated but will work when you are using a single NIC (with or without aggregation).
To be more specific about what does not work, two virtual servers have "Network server" (vlan 8) defined as a group of ports. Both physical and virtual servers on the same vlan cannot ping or resolve queries arp for virtual machines on vhost2. Everything that is sent from a different subnet can connect to the virtual machines on each virtual server.
So the question is this: why not ESX 5.5 works as ESX 4.0 when the configs are also near identical, as I can make them and they are connected to the same switch?
Thanks for any input you can provide!
S
This is the configuration for vhost1
switch 1 (sw01)
SW01 #sho run int port 3
Building configuration...
!
!
interface port-channel 3
Description LAG for vhost1
no downtime
switchport mode trunk
switchport trunk allowed vlan 1.8
!
end
SW01 #sho run int 0/23 concert
Building configuration...
!
!
0/23 gigabit-switchport interface
Vhost1 vmnic2 vswitch1 description
no downtime
channel-group mode 3 on
!
end
SW01 #sho run int concert 0/24
Building configuration...
!
!
0/24 gigabit-switchport interface
Vhost1 vmnic1 vswitch1 description
no downtime
channel-group mode 3 on
!
end
vhost1
[root@vhost1 ~] # esxcfg - NICS - l
Name PCI Driver link speed Duplex MAC address MTU Description
vmnic0 03:04.00 tg3 up to 1000Mbps Full 78:e7:d1:5f:01:f4 1500 Broadcom Corporation NC326i PCIe Dual Port Gigabit Server Adapter
vmnic1 03:04.01 tg3 up to 1000Mbps Full 78:e7:d1:5f:01:f5 1500 Broadcom Corporation NC326i PCIe Dual Port Gigabit Server Adapter
[root@vhost1 ~] # esxcfg - vmknic - l
Interface Port Group/DVPort IP IP family address Netmask Broadcast MAC address MTU TSO MSS active Type
vmk1 VMkernel - Server NetIPv4 10.1.8.31 255.255.255.0 10.1.8.255 00:50:56:78:9e:7e 1500 65535 true STATIC
VMkernel - SAN Net IPv4 10.1.252.20 vmk0 255.255.255.0 10.1.252.255 00:50:56:7 c: d8:7e 9000 65535 true STATIC
[root@vhost1 ~] # esxcfg - vswif - l
Port Group/DVPort IP IP family name address Netmask Broadcast Enabled TYPE
Service vswif1 - NetIPv4 management 10.1.1.12 console 255.255.255.0 10.1.1.255 true STATIC
[root@vhost1 ~] # esxcfg - vswitch - lSwitch name Num used Ports configured Ports MTU rising ports
32 4 32 1500 vSwitch0
Name PortGroup VLAN ID used rising Ports
0 3 VM network
Switch name Num used Ports configured Ports MTU rising ports
vSwitch1 64 12 64 1500 vmnic1, vmnic0
Name PortGroup VLAN ID used rising Ports
Server network 8 7 vmnic1, vmnic0
Console - management service vmnic1 Net0 1, vmnic0
VMkernel - server 1 vmnic1 Net8, vmnic0
On the Load Balancing vSwitch it Hash IP value.
This is the configuration for vhost2
switch 1
SW01 #sho run int 4 port
Building configuration...
!
!
interface port-channel 4
LAG description for vhost2
no downtime
switchport mode trunk
switchport trunk allowed vlan 1.8
!
end
SW01 #sho run int concert 0/17
Building configuration...
!
!
0/17 gigabit-switchport interface
Description vhost2
no downtime
channel-group mode 4 on
!
end
SW01 #sho run int concert 0/18
Building configuration...
!
!
interface gigabit-switchport 0/18
Description vhost2
no downtime
channel-group mode 4 on
!
end
vhost2
~ # esxcfg - NICS - l
Name PCI Driver link speed Duplex MAC address MTU Description
vmnic0 e1000e up to 1000Mbps Full 00:25:90:e7:0e:9 0000:08:00.00 1500 c Intel Corporation 82574 L Gigabit Network Connection
vmnic1 0000:09:00.00 e1000e up to 1000Mbps Full 00:25:90:e7:0e:9 d 1500 Intel Corporation 82574 L Gigabit Network Connection
~ # esxcfg - vmknic - l
Interface Port Group/DVPort IP IP family address Netmask Broadcast MAC address MTU TSO MSS active Type
Management 10.1.1.15 IPv4 network vmk0 255.255.255.0 10.1.1.255 c 1500 65535 real STATIC 00:25:90:e7:0e:9
~ # esxcfg - road - l
VMkernel itineraries:
Interface of network gateway subnet mask
10.1.1.0 255.255.255.0 subnet local vmk0
by default 0.0.0.0 10.1.1.1 vmk0
~ # esxcfg - road - n
Expiration type neighbor Interface MAC address
10.1.1.1 00:a0:c8:8 has: ff: 3 b vmk0 19m13s unknown
~ # esxcfg - vswitch - l
Switch name Num used Ports configured Ports MTU rising ports
vSwitch0 4352 7 128 1500 vmnic0, vmnic1
Name PortGroup VLAN ID used rising Ports
Server network 8 1 vmnic0, vmnic1
Management network 0 1 vmnic0, vmnic1
On the Load Balancing vSwitch it Hash IP value.
Hello
Do you share the vmnic for ESXi management and network of the VM?
Try to follow the steps in this KB: VMware KB: NICs using EtherChannel grouping allows intermittent network connectivity in ESXi
Trying to team adapters using EtherChannel, network connectivity is disrupted on an ESXi host. This problem occurs because grouping NIC properties will not propagate to the network management portgroup in ESXi.
When you configure the ESXi host for grouping of NETWORK cards in attributing to the road based on ip hashload balancing, this configuration is not propagated to the portgroup management network.
Workaround
Note: to reduce the possibility network loss please change the route based on the hash of IP using this method of balancing load vSwitch
- Close all ports in the physical switch team leaving a single port as active.
- Change the Route based on ip hash vSwitch and load balancing management Portgroup.
- Set up the physical switch port channel.
- Activate the physical switch ports.
Notes: Using this method will avoid a loss of connection when you switch to a port channel. Also make sure that under the properties for vmk portgroup # used for network management you tick the box "Enabled".
-
Hello
I have a pair of Nexus 5 K in an area of the VPC and few 2960's as members of the VPC, with a port to the domain channel.
Topology is:
5K 1 and 5 K 2 in the area of the VPC
VPC 5 1 K and 5 K 2 to 2960
2960 a gi0/1 and gi0/2 in 1 port channel
Gi0/1 5 k 1, gi0/2 5 k 2
I know what I'm going to ask can be completely against the purpose of the VPC, but I'm looking for a way to promote gi0/1 for traffic, rather than balancing on gi0/1 and gi0/2. The reaon for this is I want to take advantage of the absence of loop that provides mail ORDER, but would also like to have a link primary and secondary, as the majority of traffic should in fact go through 5K 1, rather than 5 K 2.
Any suggestion is welcome.
Thanks in advance
Anthony
Hi Anthony,.
The Cisco NX - OS software load balance traffic across all operational interfaces in a portchannel by chopping the addresses in a numeric value that selects one of the links in the channel. Port channels provides default load balancing. It uses IP addresses, MAC addresses or layer port numbers 4 to select the port-channel-balancing link of the load. The port-channel load balancing uses the source or destination or port addresses, or both the source and ports or destination addresses.
You can configure the mode of load balancing to be applied to all port channels that are configured on all of the device, or the specified modules. Each module configuration takes precedence over the configuration of load balancing for the entire device. You can configure a single mode of load balancing for the whole device, a different mode for specified
modules and another mode of the other specified modules. You can not configure the port-channel load balancing method.
You can configure the type of balancing algorithm used. You can choose the balancing algorithm that determines which Member port to select for the evacuation traffic by looking at the fields of the frame.
Note: The mode of load balancing by default for layer 3 interfaces is the IP source and destination address, and the mode of the load balancing by default for non - IP interfaces is the source MAC address and destination.
The configuration mode, you can try different method of load-balancing,
port-channel - the balance of the load {dest-ip-port | dest-ip-port-vlan |}
destination-ip-vlan. destination-mac | destination-port | source-dest-ip-port | source-dest-ip-port-vlan. source-dest-ip-vlan. source-dest-mac | source-dest-port | Source-ip-port | Source-ip-port-vlan. Source-ip-vlan. source-mac | source-port} [number of the module]
To sum up: I can't tell which port would be chosen, it depends purely on image type you send with the combination method of load balancing.
After you change you know also from the command which link takes the traffic.
NEXUS2-SPAN # show the port-channel - the balance of the charge-transfer port-channel 71 src - ip 1.1.1.1 route dst - ip 2.2.2.2 VLANs 51 2 interface module
Module 2: Missing params will be substituted by 0.
Load balancing algorithm: src - dst ip-l4port
RBH: port id out 0xb0: Ethernet8/8
We can try also to rework the same NLB on the 2960 also. Is it purely depends on the algorithm of load balancing. Below is for 2960 balancing tweaking,
Even after doing this that I wouldn't say 100 percent, he would choose a link.
I hope this helps!
Thank you
Richard.
* Rate if this is useful
Maybe you are looking for
-
Address Book G4 Intel Snow Leopard Leopard
I want to get my contacts from address book to my Mini G4 under Leopard on my Intel Mini under Snow Leopard. While I know that there is an address book import feature, I don't know what to do exactly. What file/folder on the Snow Leopard address bo
-
Update Satellite Pro 420 CDT to Windows 98
I have an old 420CDT Satellite Pro I want to test my wireless router. It was delivered pre-loaded with Windows 95, but I need to upgrade to Windows 98 minimum. When I run the installer of Windows 98 (upgrade mode) he told me that I should update the
-
Still signed in even though closed Skype is closed and completely disconnected
Hey I seem to have a problem for months, I don't know if it of a known issue or not, but some friends have that mine have told me that I'm still registered on Skype, even though I have disconnect manually. Sometimes I'm quite 'Skype' and the program
-
Hello I posted this question in another room but thought I would repost it here. I bought a HP Officejet Pro 8600 Plus e-all-in-one... When I added it to the selection of the connected HP device it shows my printer and it is online but under it, it d
-
Hello. Anyone know how to display three color polygon in a graph only? Thank you.