Re: [lvs-users] LVS on localhost

To: Julian Anastasov <ja@xxxxxx>
Subject: Re: [lvs-users] LVS on localhost
Cc: " users mailing list." <lvs-users@xxxxxxxxxxxxxxxxxxxxxx>
From: Tim Vaillancourt <tim@xxxxxxxxxxxxxxxx>
Date: Mon, 28 Mar 2011 15:55:13 -0700
Interesting, thanks.

I better take a step back and get some more background on what I am 
doing. I assumed this very, very simple at first but am finding myself 
missing background on LVS as this discussion progresses, and am a bit 
lost now. I see where you're going with translation though.

I will take a much more comprehensive look at what I am doing and try 
and reduce the noise on the list I am causing. You're help and work on 
LVS is much, much appreciated!



On 24/03/11 03:59 PM, Julian Anastasov wrote:
>     Hello,
> On Thu, 24 Mar 2011, Joseph Mack NA3T wrote:
>> On Thu, 24 Mar 2011, Julian Anastasov wrote:
>>>> The ipvsadm syntax I've tried:
>>>> ipvsadm -A -t -s rr
>>>> ipvsadm -a -t -r -m
>>>> ipvsadm -a -t -r -m
>>>     LocalNode is rewritten in 2.6.37. The above commands
>>> should work with NFCT enabled or disabled:
>> I didn't know about this. Sorry Tim, looks like I've been
>> misleading you.
>> thanks. Are there instructions on setting up LocalNode and
>> using ipvsadm as a result of these changes?
>     The above example is a combination of local
> client and local real server and requires NAT due to
> difference in ports. LocalNode was known as a method
> that installs NULL transmitter, i.e. delivers the
> traffic for VIP to the local stack without any
> address/port manipulations.
>     The user visible changes from these patches
> are as follows (2.6.37+):
> - 
>         This patch allows NAT/DR/TUN real servers
>     to fallback to local-RIP mode at any time. It can
>     happen if the real IP is added/removed from server
>     or even for complex setups with master and backup
>     where the backup is a real server in the master
>     and when backup takes control we must assume
>     local mode for the synced connections that have
>     BACKUP_IP as RIP. Note that every sync-ed connection
>     comes with its forwarding method.
>         Another addition is the ability to use
>     NAT method for local real server. Before now
>     it was only LocalNode possible. Now you can
>     translate to different address and port as the
>     above example for
>         Now when adding real server you can not
>     see type Local if the RIP is a local address.
>     The forwarding type remains as configured, so that it
>     can be synced. As result, DR and TUN can fallback
>     to the known LocalNode mode. OTOH, NAT has
>     the option to translate to different local IP
>     and port, i.e. it is not exactly the LocalNode
>     as we know it.
> - 
>     Now local client can work with NF conntracks.
>     To summarize, the rules are as before: if we
> do not want translation we can use DR, if we require
> translation we should use NAT.
> Regards
> -- 
> Julian Anastasov <ja@xxxxxx>

Please read the documentation before posting - it's available at: mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
or go to

<Prev in Thread] Current Thread [Next in Thread>