I have a scenario where client wants to use the 10gig interfaces from node1 and configuration configuration is as belows
pool0 for nfs - dynamic
pool1 for smb - static
The node1 10gig infterfaces are kept in pool0 doing LACP. Can we use the same 10gig interfaces from node1 in pool1 not using LACP, just adding them as 10gig-1 and 10gig-2. I am not really sure if it can work that way with no issues are we have different pool config.
So just to clarify, you are looking to use (from node1):
a) 10gige-agg-1 (which represents the aggregate of: 10gige-1 and 10gige-2) with a link aggregration protocol of LACP in pool0?
- Of course, the switch will be configured to support this
b) Then, add 10gige-1 and 10gige-2 *individually* (no link aggregation) so that they each get a separate IP in pool1?
While you can definitely use the same interface (from the same node) in more than one pool, you should consider the following:
1) If in pools from different subnets, use the same MTU
In fact if you were to assign for example
a) MTU of 1500 in subnet0 and assign 10gige-1 to subnet0:pool0
b) MTU of 9000 in subnet1 and assign 10gige-1 to subnet1:pool1,
you'd get the following error message:
"When the MTU for the subnet does not match the MTU in use on the designated interface, the lower MTU will be used and packets may be dropped."
Note though, if the same MTU were used in the different subnets (and added the same interfaces to multiple pools) you would not get this error.
2) To use the same interfaces in different subnets the only way this could work is using VLAN Tagging.
Simply added here to remind you to define the VLAN ID in the subnet properties within OneFS.
Specific to your inquiry and proposed setup though
3) You should consider using the same link aggregation for the same interfaces when used in multiple pools.
Remember, the ports will have a matching switch configuration so if the port group is defined on the switch with LACP, you will want the same on the setup for the corresponding Isilon interfaces.
I understood the environment, I think you are able to use other ports for pool 1. For example,
If you have four external ports on node 1, port1 and port2 bundled the one pair into an aggregated set using LACP for pool 0, so port1 and port2 are unavailable for pool1. However, port3 and port4 can be used for pool1.
They wanted to do it in the same subnet, but they also have plans on creating multiple subnets as well. So how should i go about if it is same and with different subnets, how does it make a difference.
I checked on my virtual cluster now and i am able to add the same ports in multiple pools. I am not really sure if it is going to be different on real cluster. I will try out tomorrow and let you know.
should not make any difference, i just tried on real cluster and was able to get to the screen where i can select the same nodes even though they are already part of another pool. I don't have a set of available IP address to complete that wizard but i don't see why it would not work. Do test SmartConnect functionality after the fact.
A node's external interfaces cannot by an IP address in both an aggregated configuration and as individual interfaces. You must remove a node's individual interfaces from all pools before configuring an aggregated NIC. Also, you must select the same aggregation method for all participting devices.
As a general best practice, If aggregated interfaces are used in one pool, Then the individual members of the same aggregated interfaces should not be used in another pool.
I have confirmed the information on the offical training material of Isilon diagnostics and troubleshooting, I don't find any problem on my reply. Thanks.
I will try to use 10gig-agg interfaces across pools. I think that whats Christopher says in his 3rd point.
Thank you all.
Yes, you are correct, that is what point 3 above is stating. Using your proposed configuration as an example, you can *not* use (node1) 10gige-agg-1 in pool0 but then add them individually in pool1.
*Either* you use 10gige-agg-1 in both pools (with the respective LACP/port aggregation configuration on the switch) or you use them individually in both pools (no LACP configuration on the switch).
christopher_ime
4 Operator
•
2K Posts
0
December 26th, 2013 22:00
So just to clarify, you are looking to use (from node1):
a) 10gige-agg-1 (which represents the aggregate of: 10gige-1 and 10gige-2) with a link aggregration protocol of LACP in pool0?
- Of course, the switch will be configured to support this
b) Then, add 10gige-1 and 10gige-2 *individually* (no link aggregation) so that they each get a separate IP in pool1?
While you can definitely use the same interface (from the same node) in more than one pool, you should consider the following:
1) If in pools from different subnets, use the same MTU
In fact if you were to assign for example
a) MTU of 1500 in subnet0 and assign 10gige-1 to subnet0:pool0
b) MTU of 9000 in subnet1 and assign 10gige-1 to subnet1:pool1,
you'd get the following error message:
"When the MTU for the subnet does not match the MTU in use on the designated interface, the lower MTU will be used and packets may be dropped."
Note though, if the same MTU were used in the different subnets (and added the same interfaces to multiple pools) you would not get this error.
2) To use the same interfaces in different subnets the only way this could work is using VLAN Tagging.
Simply added here to remind you to define the VLAN ID in the subnet properties within OneFS.
Specific to your inquiry and proposed setup though
3) You should consider using the same link aggregation for the same interfaces when used in multiple pools.
Remember, the ports will have a matching switch configuration so if the port group is defined on the switch with LACP, you will want the same on the setup for the corresponding Isilon interfaces.
Jeffey1
4 Operator
•
2.8K Posts
0
December 26th, 2013 18:00
Hi Rdamal,
May I know your OneFS version? do you have smartpools advanced licensed? Thanks.
Rdamal
2 Intern
•
165 Posts
0
December 26th, 2013 18:00
Hi Jeffey,
Its a cluster with oneFS 7.0.2.4 and yes, we do have smartpools advanced licensed
Jeffey1
4 Operator
•
2.8K Posts
1
December 26th, 2013 19:00
Hi Rdamal,
I understood the environment, I think you are able to use other ports for pool 1. For example,
If you have four external ports on node 1, port1 and port2 bundled the one pair into an aggregated set using LACP for pool 0, so port1 and port2 are unavailable for pool1. However, port3 and port4 can be used for pool1.
Rdamal
2 Intern
•
165 Posts
0
December 26th, 2013 20:00
Hi Dynamox
They wanted to do it in the same subnet, but they also have plans on creating multiple subnets as well. So how should i go about if it is same and with different subnets, how does it make a difference.
dynamox
9 Legend
•
20.4K Posts
0
December 26th, 2013 20:00
same subnet ?
Rdamal
2 Intern
•
165 Posts
0
December 26th, 2013 20:00
I checked on my virtual cluster now and i am able to add the same ports in multiple pools. I am not really sure if it is going to be different on real cluster. I will try out tomorrow and let you know.
dynamox
9 Legend
•
20.4K Posts
0
December 26th, 2013 21:00
should not make any difference, i just tried on real cluster and was able to get to the screen where i can select the same nodes even though they are already part of another pool. I don't have a set of available IP address to complete that wizard but i don't see why it would not work. Do test SmartConnect functionality after the fact.
Jeffey1
4 Operator
•
2.8K Posts
1
December 27th, 2013 03:00
A node's external interfaces cannot by an IP address in both an aggregated configuration and as individual interfaces. You must remove a node's individual interfaces from all pools before configuring an aggregated NIC. Also, you must select the same aggregation method for all participting devices.
dynamox
9 Legend
•
20.4K Posts
0
December 27th, 2013 05:00
Jeffey,
your reply is contradicting Christopher's reply, which EMC employee is correct ?
ACHOUKEKAR
22 Posts
1
December 27th, 2013 07:00
As a general best practice, If aggregated interfaces are used in one pool, Then the individual members of the same aggregated interfaces should not be used in another pool.
Rdamal
2 Intern
•
165 Posts
0
December 27th, 2013 07:00
Jeffey and Christopher
I will try to use 10gig-agg interfaces across pools. I think that whats Christopher says in his 3rd point.
Thank you all.
dynamox
9 Legend
•
20.4K Posts
0
December 27th, 2013 07:00
but it is possible correct ? I see possible use cases where that is needed.
Jeffey1
4 Operator
•
2.8K Posts
0
December 27th, 2013 08:00
Hi Dynamox,
I have confirmed the information on the offical training material of Isilon diagnostics and troubleshooting, I don't find any problem on my reply. Thanks.
christopher_ime
4 Operator
•
2K Posts
0
December 27th, 2013 09:00
Yes, you are correct, that is what point 3 above is stating. Using your proposed configuration as an example, you can *not* use (node1) 10gige-agg-1 in pool0 but then add them individually in pool1.
*Either* you use 10gige-agg-1 in both pools (with the respective LACP/port aggregation configuration on the switch) or you use them individually in both pools (no LACP configuration on the switch).