Procurve 5406ZL - Trunk question 10GB and 1 GB mixed

[M.Breitenbach]

hello together,

we got an phenomen,

we have an Trunk with 3 x 1 GBIT copper and 1 10GBIT FC between 2 Procurve 5406zl Switches

The Traffic uses at first the 3 Copper -> when they are full, the 10 GB will be used. ( we see this in our monitoring tool)

The Problem is, that we got errors from monitoring and systems, cause the 3 1 GBIT are full...

??

Any ideas how to fix (without removing the 3 x 1 GBIT copper) ?

Thanks and regards

[parnassus]

Do you mean you have a single LAG (Link Aggregation Group) which members are:

• 3 Gigabit Ethernet ports
• 1 TenGigabit Ethernet (via SFP+ Transceiver) port

or that you have two separate LAGs:

• 1st LAG with those three Gigabit Ethernet ports above
• 2nd LAG with that single TenGigabit Ethernet (via SFP+ Transceiver) port above

or what else?

AFAIK a LAG, indipendently from being set to use the LACP type (IEEE 802.3ad Protocol) or the Trunk type (No Protocol), doesn't admit to have member ports of different media types and speeds.

Documentation is quite clear about that (see any Software Management and Configuration Guide for K-branch based Switches):

• For proper trunk operation, all ports in a trunk must have the same media type and mode (such as 10/100TX set to 100FDx, or 100FX set to 100FDx.) The flow control settings must also be the same for all ports in a given trunk.
• You can configure the trunk group with up to eight ports per trunk. If multiple VLANs are configured, all ports within a trunk will be assigned to the same VLAN or set of VLANs. (With the 802.1Q VLAN capability built into the switch, more than one VLAN can be assigned to a trunk.

and again:

• The switch does not support port trunking through an intermediate, non-trunking device such as a hub, or using more than one media type in a port trunk group. Similarly, for proper trunk operation, all links in the same trunk group must have the same speed, duplex and flow control.

That's valid also if you are going to create a LAG set to LACP which permits to define one or more Standby Ports (Standby Link Capability): those ports must be of the same media type, must have the same speed, duplex and flow control as the primary ports of the same LAG.

Which type of trunk you're referring to?

[M.Breitenbach]

hello and thank you very much

we are using Trunks - Standard - and now only with the same media type / speed etc.

We have done a mistake i think :-)

Thanks! and i hope that now everything is working fine

[parnassus]

Yep, anyway...if between your two HPE 5406 zl you still have the Fiber Optic link and both your Switches are still equipped with SFP+ transceiver...it's somewhat better to use a single 10G uplink (no LACP, no trunking...it's just a single uplink) than using a LAG made of 3 x 1G (so port trunking with LACP). Better a fat uplink than many tiny ones (up to 8, aggregated)...yes, you will lose resiliency but you will gain in data transfer throughput especially when the traffic happens, between both Switches, from-many-to-many.

[M.Breitenbach]

yes,

the 3 x 1 GBIT Copper is an very old configuration. The 5406 Switche are more than 6 years old.

Now we have 10 GB modules and put them into the existing trunks ... failure :-)

[parnassus]

Well...the HPE 5400 zl Switch series is still actually good (especially if you can equip it with v2 zl Modules instead of using only old v1 zl Modules).

A 10G ports oversubscription consideration:

Don't know what is your exact 10G SFP+ Module configuration (which version [*] it is, if it is v1 zl or v2 zl) nor what is going to be its complete purpose (not only for a future LAG I suppose) but, related to creating a future LAG of two 10G SFP+ ports, let's say you have 4 10G SFP+ ports v1 zl Module, keep in mind to aggregate together 10G SFP+ ports 1+2 or 3+4 within the same 10G SFP+ Module especially if then you plan to use remaining 2 10G SFP+ ports for other usage (as example: 10G Hosts connections or just another two 10G ports LAG) since the ports mapping that shares a single 14.4Gbps Channel (of a total of two single 14.4Gbps Channels, for a grand total of 28.8Gbps of aggregated bandwidth across all 4 linked ports) is statically assigned respectively on ports pair 1+4 and ports pair 2+3.

[*] As example you can use the HP 4-port 10GbE SFP+ zl Module (J9309A) - which is a v1 zl - or the HP 8-port 10GbE SFP+ v2 zl Module (J9538A).

[ptwog]

Hi Parnassus,

I have (4) HP5412zl with J8707A (X2 10gbe) modules. I was planning to use LACP/Trunking for all (4) X2 interfaces into a single trunk, which will uplink to a Fortinet (FortiSwitch). Am I unable to aggregate all 4 interfaces or is it more useful to do them in pairs as you've described:  10G SFP+ ports 1+2 or 3+4 within the same 10G SFP+ Module. It appears that I'll only get a total of 28.8 gbps- which makes me wonder if a trunk of 3 of the 4 interfaces makes more sense- but again, there may be an issue with the 14.4 gbps limit per channel.

If I make two aggregated links, I would need to separate out my vlan trunking as well to prevent a loop, correct? In other words, if I have 10 vlans, rather than having one (4) interface trunk with the default vlan untagged and all other vlans tagged, I'll need to split those out, right?

I also have some J9546A modules, but with less available ports. The J9546A per port cost is much higher and 10-GBASE-T Copper SFP+ transceivers are much more expensive than 10GBE Fiber SR SFP+ transceivers. Thanks for any input.