LAN Routing
cancel
Showing results for 
Search instead for 
Did you mean: 

Load Balancing High Availability Configuration

robyaps
Advisor

Load Balancing High Availability Configuration

We have 4 Switch HP A5120 48G-EI stack IRF switch and 2 Switch HP 3500yl-24G switch. 3500yl are configured in cluster, 2 A5120 form an IRF fabric an the other 2 A5120 form another IRF fabric. We divided 4 A5120 so we can upgrade IRF fabrics without complete interruption in case of upgrade problems. Having 2 separate IRF with differente ESXi hosts connected allow to move VMs on host not connected with in upgrade phase IRF Fabric. At the moment we connect IRF fabrics to 3500yl cluster using Distributed Trunk links on 3500yl side and link aggregation links on IRF side for the LAN vlan only (vlan1) [ (IRF1) --trunk--> (3500yl-cluster) <--trunk-- (IRF2) ]. Can someone help to understand and make the best solution to connect also the 2 IRF fabrics togheter to have redundancy connectivity ? We want to transmit also other vlans on the actual trunks links and on the new redundant connection between IRFs avoiding loops and if possible balancing vlan traffic ?

(IRF1) <--trunk (all vlan)--> (3500yl-cluster) <--trunk (all vlan)--> (IRF2)

+

(IRF1) <--trunk (all vlan)--> (IRF2)

We read about RRPP for A5120 that probably allow such a ring configuration, but 3500yl does not seem to support this protocol, RRPP can be used also with intermediate switches not supporting RRPP ?

We need info about protocols and configuration to get the best solution for resiliency and fast convergence. Also, what about Smart link technology in this ring-like topology ?

 

7 REPLIES
Vince-Whirlwind
Honored Contributor

Re: Load Balancing High Availability Configuration

"We want to transmit also other vlans on the actual trunks links and on the new redundant connection between IRFs avoiding loops and if possible balancing vlan traffic"

If I've understood this correctly then you probably want to make use of separate instances using MST.

You'll have instance "2" which can be the old setup, with VLANs 100-199 only.

Instance "3" can be the new trunk between the two IRF stacks, with VLANs 200-299 only.

If you configure the VLANs everywhere, then your important design decisions will revolve around where each VLAN is to be routed.

robyaps
Advisor

Re: Load Balancing High Availability Configuration

I read some specs of protocols suitable for Redundancy/Load Balancing loop connection topology. for MST you mean MSTP ? We have to read about MSTP, but probably other protocols have best convergence time (Smart Link on which we go deeper to see if appropriate solution). I'try to explain best the configuration topology:

The intended configuration is think as 3 switch interconnected in a ring topology with 2 o 4 port for each switch linked together to have more capacity and resiliency also at link level. The complex part is that 3 switch fabric are of different technology also if all HP switch. One switch fabric is formed by 2 HP Procurve 3500yl switch (clustered together, this is not a true cluster like IRF, but Distributed Trunking come to help). A second switch fabric is formed by 2 HP A5120 switch configured in true with IRF tech), the same configuration for the third switch fabric (2 HP A5120 switch in IRF). We divided 4 HP A5120 IRF in 2 IRF after an IRF upgrade that blocked many systems connetcet to the 4 Switch IRF. So Now we have 3 Switch fabric to connect together in a risience configuration. IRF fabric can be connected with 3500 Distributed trunk with multiple links (we just use this connection, this works well), but we want to connect also the 2 IRF switch fabric togheter so we'll have a ring topology.

If possible we search a solution that allow to use one ring direction for some vlans and other ring direction for other vlans so we reach resiliency and load palancing instead of aving one link taken down in standby mode. Smart Link tech seems to allow load balancing with smart link group and different protected VLAN for each group, but we so not know yet if this tech is applicable to link-aggregated "link" or on single switch port.

As you know MSTP is applicable for the objective we want to reach ?

robyaps
Advisor

Re: Load Balancing High Availability Configuration

With MST you mean  MSTIs (activating MST region for use with smart link group protected VLAN) ! We do not have MST nor smart link active at moment. We are in study phase, reading about smart link.

We read about configuration example on a5120 "High Availability - Configuration Guide". In the guide we read 2 conflicting sentence:

1. Configuring a smart link device -> Configuration prerequisites -> ... "Disable STP and RRPP on the ports that you want to add to the smart link group, and make sure that the ports are not member ports of any aggregation group"

2. Syntax for configuring smart link "In port view" -> ... " Enter Ethernet port view or layer 2 aggregate port view"

So... in 2. seems that Smart Link can be used for aggregated port, part of a link aggregation. What about sentence 1 speaking about member of an aggregate group ?

 

Also, in Procureve 3500yl "Advanced Traffic Management Guide" for K.15.15 switch software, we do not see "smart-link flush enable control-vlan" command for associated device specificed in a5120 documentation, so how  Enable flush message receiving can be configured on 3500yl as associated smart link device ?

 

 

Mike_ES
Valued Contributor

Re: Load Balancing High Availability Configuration

Hi,

Procurve Smart-Link example:

Smart-Link

Michal

 

 

 

robyaps
Advisor

Re: Load Balancing High Availability Configuration

I read about smartlink tech but this is not an option in a ring topology with 3 switch as any smartlink group must have two links (one "active" and one "inactive"). For our purpose the only solution seem to be using STP (MSTP), changing the actual Distributed Trunking - IRF connections in normal Link Aggegation connections and configuring an MSTI with all VLANs or using different MSTIs for VLANs (load balancing), in this example we want IRF1 (5120) and IRF2 (5120) to communicate on direct link leaving link between IRFs passing trhough 3500 only for Redundancy, so we do not need 2 MSTI as now.

Procurve does not support RRPP as 5120 does and Distributed Trunking does not support STP on DT links as 5120 does for IRF Link Aggreggations. For this reasons we'll loose DT-IRF load balancing properties to achieve HA topology. As soon as we use these two HP tech not perfectly speaking one anoher...

robyaps
Advisor

Re: Load Balancing High Availability Configuration

Probably one possible alternative be to configure smartlink with a smartink group on 3500 using as active/standby links to IRF1 and IRF2, configuring IRF switches only as associated devices (passing smartlink messages). Also if probably the MSTP solution is more supported, We'll try smartlink solution as depicted to be sure. 

robyaps
Advisor

Re: Load Balancing High Availability Configuration

If one try to configure a DT trunk as smart-link master or salve the 3500 switch returns this warning:

     The port cannot be added to Smart link group when Distributed Trunking is enabled.

So no STP and no Smart-Link can be used with DT !

Solution:

1. no DT-IRF links between 2 procurve 3500 and IRF1=(2 x 5120) and IRF2=(2 x 5120)

2. HA using MSTP (connection: IRF1 to 3500_1 with Lacp Link Aggregation + IRF2 to 3500_2 with Lacp Link Aggregation + 3500_1 to 3500_2 using ISC or normal Link if DT is not used at all

3. Using MSTP to block redundant path, port priority to decide best traffic path in normal state and if necessary multiple MSTIs to make load balance on vlan

DT seems a great tecnology, but it is not so useful because so many restrictions (MSTP, Smart Link, Mesch, etc.)