This post is more than 5 years old
2 Intern
•
359 Posts
0
3624
redundancy for celerra connection to network
Currently our Celerra is connected to one Cisco 3750G 24 port switch. Our VMWare (3) servers are also connected to this. There's 3 VLANs,
1 LAN
10 SAN
20 DMZ
This switch is trunked to the core.
Now we added a second switch. Any best practices how to make it redundant, so if one switch goes down?
on both sides in the back of the celerra there's
cge0, cge1, cge2, cge3 - cge0, cge1, cge2, cge3
Any link to documents would be helpful.
Storage traffic is on its on vlan. The VMWare ESX hosts have interfaces on the storage vlan to talk to the celerra via NFS.
dynamox
2 Intern
2 Intern
•
20.4K Posts
0
February 11th, 2013 11:00
you are very close
1) connect cge0,1,2 to switch 1. Create portchannel on the switch side that consists of the 3 ports (tag for VLAN id you will need), then create trunk on the Celerra side.
2) connect cge3 to swtich2 (tag for VLAN id you will need).
3) create fail-safe-network (FSN) devices that consists of trunk created in step1 and cge3 devices. When you create FSN device you will specify which device is the active one, in this case you would want your trunk to be the active device and cge3 standby.
4) create interfaces on Celerra and specify VLAN ID that you want to use for that interface. Like you said, you could have interface named "vmware" with VLAN ID 30 and then another interface named "cifs" with VLAN ID 40.
in order for server_3 to work correctly in failover situation it needs to be cabled and configured identically from switch perspective, you don't have to do any logical configuration from Celerra side because when server_2 fails over to server_3 it will assume its identity (MAC address, IP ..etc)
dynamox
2 Intern
2 Intern
•
20.4K Posts
0
February 11th, 2013 06:00
from Celerra configuraiton perspective, have you seen this document ?
https://support.emc.com/docu10545_Configuring-and-Managing-Celerra-Network-High-Availability.pdf?language=en_US
etaljic81
1K Posts
0
February 11th, 2013 06:00
If you want to protect yourself against a switch failure then you need to create an FSN (fail safe network). You can create an FSN across two cge ports or you can create two LACP trunks and create an FSN on top of those two.
Example 1: no trunking
Step1: cge0 - connect to switch1
Step2: cge1 - connect to switch 2
Step3: create FSN using cge0 and cge1. You'll be able to choose which link the traffic should go through. If that link dies FSN will use the other link
Example 2: trunking
Step1: cge0 and cge1 - connect to switch1
Step2: create LACP/etherchannel trunk using cge0 and cge1 ports (let's call this LACP0)
Step3: cge2 and cge3 - connect to switch2
Step4: create LACP/etherchannel trunk using cge2 and cge3 ports (let's call this LACP1)
Step5: create FSN using LACP0 and LACP1
You don't need to do anything special on the switches to make an FSN work.
The standby datamover is not being used unless there's a datamover failure. You have to connect the cge ports on the standby datamover the same way that you connected the ports on the active datamover. For example, if cge0 and cge1 are connected to switch1 on the active datamover and those two switch ports on switch1 are configured in VLAN10, you have to connect cge0 and cge1 on the standby datamover to switch1 and configure those to be in VLAN10 as well.
Hope this helps.
etaljic81
1K Posts
0
February 11th, 2013 07:00
Correct. You will still be able to manage the array.
etaljic81
1K Posts
0
February 11th, 2013 07:00
The only thing with this config is that the cifs and iscsi traffic will be going through the same trunk.
dynamox
2 Intern
2 Intern
•
20.4K Posts
0
February 11th, 2013 07:00
here is another option for you:
switch 1: cge0,1,2 - create LACP trunk on Celerra (trk0)
switch 2 - cge3
combine trk0 and cge3 to create fsn0. Assumption is that when switch1 is up and running you have 3 x 1G connections available, when switch1 is down you are running in "degraded" mode as you only have one 1G link available.
kjstech
2 Intern
2 Intern
•
359 Posts
0
February 11th, 2013 07:00
It doesn't support 802.1q trunking?
Like on my switch to trunk various vlans over its switchport trunk encapsulation dot1q, then on the interfaces belonging to that port channel you add that statement as well.
That way multiple VLANs traverse the trunk and they can be seperated out on a per port basis, or another switch / router / device that can read and talk the correct vlan.
kjstech
2 Intern
2 Intern
•
359 Posts
0
February 11th, 2013 07:00
That document does help, thanks!
Ernes, that is helpful, thank you.
I did not configure this the first time around. All I have to go by is looking at the current active switch config. I have no idea why ports 10 and 14 are explicitly shut down and no port channel defined.
interface Port-channel2
interface Port-channel3
switchport access vlan 10
interface Port-channel4
switchport access vlan 10
interface GigabitEthernet1/0/1
description server_2-cge3
channel-group 2 mode active
spanning-tree portfast
interface GigabitEthernet1/0/3
description server_2-cge1
channel-group 2 mode active
spanning-tree portfast
interface GigabitEthernet1/0/5
description server_3-cge2
switchport access vlan 10
channel-group 3 mode active
spanning-tree portfast
interface GigabitEthernet1/0/7
description server_2-cge0
switchport access vlan 10
channel-group 4 mode active
spanning-tree portfast
interface GigabitEthernet1/0/8
description server_3-cge0
switchport access vlan 10
channel-group 3 mode active
spanning-tree portfast
interface GigabitEthernet1/0/10
description server_3-cge3
shutdown
spanning-tree portfast
interface GigabitEthernet1/0/12
description server_2-cge2
switchport access vlan 10
channel-group 4 mode active
spanning-tree portfast
interface GigabitEthernet1/0/14
description server_3-cge1
shutdown
spanning-tree portfast
Now when I log into Unisphere on this
lan ip - cifs - server_2 - cifs_trk
lan ip - replication-int - server_2 - cifs_trk
san ip - iscsi - server_2 - iscsi_trk
128.221.252.2 - el30 - server_2 - mge0
128.221.252.3 - el30 - server_3 - mge0
128.221.253.2 - el31 - server_2 - mge1
128.221.253.3 - el31 - server_3 - mge1
dynamox
2 Intern
2 Intern
•
20.4K Posts
0
February 11th, 2013 07:00
yes, you can create VLAN tagged interfaces
kjstech
2 Intern
2 Intern
•
359 Posts
0
February 11th, 2013 07:00
Oh I'm also wondering if I delete the interfaces and recreate them, I won't lose connection to the celerra in unisphere right? Because thats connected to the control station on a different IP address (and NIC going to the control station), but it can still talk to the datamovers and update said configuration no problem due to the default 128.221.25*.* IP's correct?
etaljic81
1K Posts
0
February 11th, 2013 07:00
Click on the devices tab in Unisphere. You listed the interfaces created (cifs_trk) but we need to know how that cifs_trk is configured. Go to the devices tab on the same screen where you obtain the interfaces info and post the output here
dynamox
2 Intern
2 Intern
•
20.4K Posts
0
February 11th, 2013 07:00
fatter trunk
, less resources are sitting in standby mode.
kjstech
2 Intern
2 Intern
•
359 Posts
0
February 11th, 2013 07:00
Would bandwidth be reduced since cge1 and cge3 would be on two separate switches?
I'm almost thinking we could just put all of the interfaces from server_3 on the second switch and like you said fail over the datamover if there was a switch failure.
I will have to discuss with the I.T. team. downtime would surely need to be scheduled and off hours.
You have been really great and thanks for looking over my config so far.
etaljic81
1K Posts
0
February 11th, 2013 07:00
You will only be using one link instead of two in the LACP trunk. You are correct. So you have to make a decision as far as what is more important. Redundancy or bandwdith.
You can connect the cge ports on server_3 and perform a manual failover, that is definitely an option. Probably the easiest option at this point, as long as you are OK with performing a manual failover of the datamover once you get a call that switch1 is down, is to leave the configuration as is.
Message was edited by: Ernes Taljic
etaljic81
1K Posts
0
February 11th, 2013 07:00
One more thing. In your switch config, can you clarify what is on switch1 and what is on switch2?