|
On Wed, Nov 15, 2006 at 11:09:43AM -0800, Joseph Mack NA3T wrote:
> >Hi there,
> >i have a problem loadbalancing DNS udp/53 via LVS DR.
> >Whenever i have a small number of clients, in the example below it is DNS
> >realtime blackhole lists used by 10 mailservers.
> >Redundancy is working fine but granularity seems to be an issue and there
> >is
> >times when i end up with something like this:
> >#ipvsadm --list --rate --numeric
> >UDP 1.2.3.4:53 0 1889 0 173253
> >0
> > -> 10.1.53.15:53 0 0 0 18
> > 0
> > -> 10.1.53.9:53 0 1889 0 173235
> > 0
>
> I don't know, but I thought I'd pop up and say something as
> I expect the answer isn't well known. This used to work in
> the early days.
>
> UDP schedules by packet (or small groups of packets). The
> only thing I can think of is that the director sees all your
> clients as being the one client. However I notice your
> persistence_granularity is one machine, so that shouldn't be
> a problem. Does the persistence_granularity actually get to
> the director? (I don't know where it goes - do you see it in
> the output of ipvsadm, is it in /proc somewhere?).
Yes i found the ipvsadm --list --persistent-conn command which shows
UDP 1.2.3.4 wrr
-> 10.1.53.15:domain 1 2 0 6
-> 10.1.53.9:domain 1 1 0 17
i think that explains a lot.
UDP 1.2.3.4:53 0 1676 0 153676 0
-> 10.1.53.15:53 0 0 0 30 0
-> 10.1.53.9:53 0 1676 0 153646 0
It does additionally show the persistence template timeout when configured so
keepalived actually gets it in the kernel.
What i don't get is why this is listed as one persistent connection.
All i can gather from documentation is that default granularity is
255.255.255.255 which is per client.
I noticed is that as all our mailservers are running a local BIND process to
forward those RBL questions to us and apparently they tend to use the same
outgoing port when (re)started close to each other.
In the short tcpdump i did i ended up with 7 of 11 total IPs with the same
src port but this should give 11 persistent connections not just 1 right?
Does anybody know any way to totally get rid of persistence?
(I don't really want to munch around in the source doing things i only get
half the truth of. ;-)
> >I noticed there apparently are some slewing algorythms
> >doing things like a slow start whenever i add a realserver
> >but i never stumbled over any documentation, can you give
> >me a hint on where to look for this?
>
> It's in the HOWTO as slow start and thundering herd. I
> thought it was in the code from the early days, but
> apparently it wasn't and I was mistaken. It's been talked
> about enough now that I can't remember if it ever went in or
> not.
Well from what i am seeing it is definitely in now.
Stefan
--
panic ("No CPUs found. System halted.\n");
2.4.3 linux/arch/parisc/kernel/setup.c
|
