LACP configuration not work correctly

[Bab]

Dear all

Hi

i have a problem with LACP and not sure what is my problem

i have added 2 pnic to my vss and both of them are ok

and create a lag with 2 uplink for my LACP (vmnic4, vmnic6)

my team failing is on ip hash

and lag is my active uplink

now my problem is while down vmnic4 will get time out and will be switch to other vmnic 2-3 min later what is my problem ?

BR

[daphnissov]

LACP does not work with vSS, only static LAGs. Regardless, why do you want to use a LAG? In almost every case, if you have an entitlement to use a vDS, you don't need any LAGs and can achieve more capabilities. Please explain your use case and needs.

[Bab]

thanks

i ran lacp on dvs

vss has 2 pnic and use for my management

i just use lacp for virtual machine traffics.

you asked why want use lag : i want use LACP thus had to create a lag and i create a lag with 2 uplinks now my main question is why my lacp not work correctly ?

[daphnissov]

You're not explaining what you're trying to do. LACP is a specific form of LAG. vSS can only do a static LAG whereas LACP is dynamic. But you didn't answer ​why​ you want to use LACP in the first place.

[Bab]

i ran LACP just for use link aggregation .

i have configure lacp on dvs not vss

[daphnissov]

Quoting from my previous reply

But you didn't answer why you want to use LACP in the first place.

[Bab]

i want use more bandwidth thus want use LACP

in the first step select my vds and choose lacp such as pic and then select migration network traffic to link aggregation group  and select manage distributed port groupds such as pic1

then select teaming and failover and select all portgroups that previously created

and according to pic3 in teamin g and failover select ip hash and lag1 for active adapters

according to pic4 next step select manage network host and assign vmnic4 to lag1-0 and vmnic6 to lag1-1

and last step such as step1 another manage distributed port grroup such as pic5 and finish

but that is so strange for me why another click on manage network traffic lags it not show

my previous configurations and had to do all of them another.

[daphnissov]

i want use more bandwidth thus want use LACP

And this is a common misconception. LACP is not the only way to achieve this and you can get this effect without using LACP by using load-based teaming in the vDS. LACP is not needed to utilize multiple uplinks simultaneously.

[Bab]

none of teaming and failover in vds can not aggregate two bandwith's link all of them are use for load balancing and redundancy

[daphnissov]

Yes it can, load-based teaming does this. You need to learn how LACP actually works and how LBT works. LACP doesn't just evenly "spray out" data across both links equally. That isn't how it's designed and what it does.

[Bab]

would you please say which type of teaming and failover can work such as LACP and aggregate two active uplinks?

[daphnissov]

I have told you multiple times now. ​LOAD-BASED TEAMING (LBT)

[Bab]

All of the teaming type has been show in attach pic i cannot find LBT

[daphnissov]

Route based on physical nic load = load-based teaming

[Bab]

but all documents mention these about LBT

The truth about LBT is that it select uplinks the same way as Route Based on Originating Virtual Port ID initially. When a VM boots, the vNICs are assigned to a dvPort. That port is used to determine which uplink the traffic will use (Figure 3). The LBT mechanism comes into play every 30 seconds, when it polls the uplinks. If an uplink is more than 75 percent utilized during that polling period, LBT will move that dvPort to a less utilized uplink.

https://virtualizationreview.com/articles/2015/03/26/~/media/ECG/virtualizationreview/Images/2015/03/0315vrm_GreenvSwitc…

1. LBT does have awareness of the link utilization and ensures that no link is utilized more than 75 percent before all the others are, as well.

2. LBT does not evenly balance traffic across uplinks when saturation is not occuring. This may explain the confusion for some folks looking at ESXTOP metrics. It will only move an assignment to another uplink once saturation occurs.

[daphnissov]

Yes, these are all true. Did you know that LACP also doesn't "balance" traffic across uplinks? Did you know that LACP (depending on what workloads you have) will also only use one link in most cases? Unless you have done the analysis on your workloads and understand their traffic patterns, you're doing a disservice to use LACP over LBT. The reality is LACP, in most cases, results in no measurable benefit versus LBT and LBT actually has the ability to look at link utilization and move traffic accordingly–something LACP cannot do.

[Bab]

you asked : Did you know that LACP also doesn't "balance" traffic across uplinks ? 

As I know LACP will aggregate bandwidth of two pnic in my LAG

Did you know that LACP (depending on what workloads you have) will also only use one link in most cases ?

in teaming and failover of each portgroup for LACP i am  using IP Hash , will it use just one pnic in this mode ?

you said : LACP, in most cases, results in no measurable benefit versus LBT and LBT actually has the ability to look at link utilization and move traffic accordingly–something LACP cannot do..

All of documents that I was read , did not mention about aggregate bandwidth between two PNIC in LBT mode .

all of them said. in LBT mode we have best load balancing in first step it will be use PNIC1 and if the traffic on this uplinc meet 75% for 30 second ,other traffics from now will be pass through from PNIC2 . and that's means it will be use just one pnic in all of times

Now I'm so confused :

1- will LACP aggregate two pnic bandwith ?

2 - willl it aggregate two pnic's bandwidth all of the time or it will be do that if need more bandwidth ?

BR

[daphnissov]

As I know LACP will aggregate bandwidth of two pnic in my LAG

You need to read about what the word "aggregate" means in this context. A LAG will utilize both uplinks simultaneously. LBT will do the same thing. In fact, regular old "route based on originating virtual port ID" will do the same thing. Aggregate means "to combine in parallel" which is exactly what both algorithms do.

in teaming and failover of each portgroup for LACP i am  using IP Hash , will it use just one pnic in this mode ?

No, you need to read how the various LACP algorithms work. None of them involving using just a single pNIC. Please go do some research on this networking topic. Use Google for help.

All of documents that I was read , did not mention about aggregate bandwidth between two PNIC in LBT mode .

all of them said. in LBT mode we have best load balancing in first step it will be use PNIC1 and if the traffic on this uplinc meet 75% for 30 second ,other traffics from now will be pass through from PNIC2 . and that's means it will be use just one pnic in all of times

Then you're not understanding what you're reading. Load balancing means that it looks at the vmnic (pNIC) load and moves ​some​ traffic around. That doesn't mean only a single vmnic is ever utilized until it reaches 75% and then all gets switched over. What would ever be the point of that? It'd be useless.

[Bab]

but now i need a feature that supply for me two options:

1- aggregate uplink's bandwidth for two pnics

2- best load balance between two pnics

what sort of teaming and failover type can provide these options ?

[daphnissov]

Load based teaming, without LACP

[Bab]

now, How can check link aggregation for LBT teaming and failover mode in my lab ?

[Duncan Epping]

just use ESXTOP and look at the various metrics for the NICs.

But as Chip said: LBT and LACP typically won't load balance a workload across both NICs at the same time, only when you have many source IP and destination IPs and the hash value outcome would result in something different would the same VM be able to send packets across two links. In most cases this is not true. And to be honest, there aren't many workloads that would have this high of a load anyway to require this.

[Bab]

You said there is not link aggregation in LBT and just is in LACP but hese said aggregation also is in LBT

Re: LACP configuration not work correctly

[Duncan Epping]

I suggest you re-read his comments, it is pretty clear.

[Bab]

i asked this question :

but now i need a feature that supply for me two options:

1- aggregate uplink's bandwidth for two pnics

2- best load balance between two pnics

what sort of teaming and failover type can provide these options ?

and get follow answer :

Load based teaming, without LACP

[Duncan Epping]

And that answer is correct.

[Bab]

As I understood in LBT we have aggregate bandwidths and utlize load balance but in LACP just can aggregate bandwidths

Is that correct ?

[Duncan Epping]

Re: LACP configuration not work correctly

[daphnissov]

Thanks for the backup, Duncan. I feel like there's an echo in this room :smileyhappy: