Re: load balancing problem

Hi,
i've found a workaround to fix this problem :
ipvsadm --set 0 0 0
this will set expiry time to 0s for UDP packet, but ... when a real server
is offline the load balancer keeps sending incoming requests to it.
Thanks for your help.

Regards


On 5/21/07, Herta Van den Eynde <[email protected]> wrote:
>
> On 21/05/07, Mohamed HADROUJ (Gmail) <[email protected]> wrote:
> > Hi,
> > when  entring the following command : ipvsadm -L -c -nI I get this
> result :
> >
> > IPVS connection entries
> > pro expire state       source             virtual            destination
> > UDP 04:57  UDP         10.11.11.225:32875 10.11.11.230:1812
> > 10.11.11.19:1812
> >
> > Every request sets the counter to 5:00
> > Thank you for your help
> >
> >
> >  On 5/21/07, Mohamed HADROUJ (Gmail) <[email protected]> wrote:
> > > I think also it's a persistence issue, because when i tried ipvsadm
> after
> > few minutes, the InActConn is initialized to 0. then the load balancer
> sends
> > the incoming request to the other server. here is an example :
> > > -I try to send several request : they all get answers but from the
> same
> > server
> > >    -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
> > >  UDP  10.11.11.230:radius wrr
> > >    -> 10.11.11.20:radius           Masq    1      0          0
> > >    -> 10.11.11.19:radius            Masq    1      0           1
> > >
> > > after few minutes :
> > >    -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
> > >  UDP  10.11.11.230:radius wrr
> > >    -> 10.11.11.20:radius            Masq    1      0          0
> > >    -> 10.11.11.19:radius           Masq    1      0          0
> > >
> > > and then all the requests coming to the load balancer are routed to
> the
> > other server :
> > >    -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
> > >  UDP  10.11.11.230:radius wrr
> > >    -> 10.11.11.20:radius           Masq    1      0          1
> > >    -> 10.11.11.19:radius           Masq    1      0          0
> > >
> > > How to fix this behaviour ?
> > >
> > > Regards,
> > >
> > >
> > >
> > >
> > > On 5/21/07, Herta Van den Eynde <[email protected]> wrote:
> > > > On 21/05/07, Mohamed HADROUJ (Gmail) <[email protected]> wrote:
> > > > > Hi All,
> > > > > I have some load balancing troubles using lvs/piranha, here is the
> > > > > description :
> > > > > when i use the ipvsadm command to view the statistics here what i
> get
> > :
> > > > > IP Virtual Server version 1.0.8 (size=65536)
> > > > > Prot LocalAddress:Port Scheduler Flags
> > > > >   -> RemoteAddress:Port           Forward Weight ActiveConn
> InActConn
> > > > > UDP  10.11.11.230:radius wrr
> > > > >   -> 10.11.11.20:radius           Masq    1      0          0
> > > > >   -> 10.11.11.19:radius           Masq    1      0          1
> > > > >
> > > > > the load balancer keeps forwarding the request to only one server.
> in
> > > > > addition of that the counter of the connection established is
> always
> > equal
> > > > > to 1.
> > > > > i've checked also in /var/log/messages, both servers are seen as
> > available.
> > > > > here is my config file :
> > > > >
> > > > > backup_active = 1
> > > > > backup = 10.11.11.229
> > > > > backup_private = 10.11.11.121
> > > > > heartbeat = 1
> > > > > heartbeat_port = 539
> > > > >  keepalive = 6
> > > > > deadtime = 18
> > > > > network = nat
> > > > > nat_router = 10.11.11.122 bond0.499:1
> > > > > debug_level = NONE
> > > > > virtual gw_radius {
> > > > >         active = 1
> > > > >         address = 10.11.11.230 bond0.498:1
> > > > >         port = 1812
> > > > >         send_program = "radpwtst -s %h -noacct -noauth -auth_port
> 1812
> > > > > -secret mysecret -status"
> > > > >         expect = "OK"
> > > > >         use_regex = "0"
> > > > >         #load_monitor = uptime
> > > > >         scheduler = wrr
> > > > >         protocol = udp
> > > > >         timeout = 6
> > > > >         reentry = 15
> > > > >         quiesce_server = 0
> > > > >         server GW_1 {
> > > > >                 address = 10.11.11.19
> > > > >                 active = 1
> > > > >                 weight = 1
> > > > >         }
> > > > >         server GW_2 {
> > > > >                 address = 10.11.11.20
> > > > >                 active = 1
> > > > >                  weight = 1
> > > > >         }
> > > > >         server GW_3 {
> > > > >                 address = 10.11.11.21
> > > > >                 active = 0
> > > > >                 weight = 1
> > > > >         }
> > > > > }
> > > > >
> > > > >
> > > > >
> > > > > Thank you for your help
> > > > > Regards
> > > >
> > > > Might be a persistence issue.  Did you try connecting from 2
> different
> > > > nodes/PCs?
> > > >
> > > > Kind regards,
> > > >
> > > > Herta
>
> Mohamed, it looks like you posted an extract from the lvs.cf instead
> of the full file.  Would you please look whether you have an entry
> "persistent = some value" in the "virtual gw_radius" part?  If not,
> try adding a "persistent = 0".
>
> Note that depending on what's behind the load balancer, persistence
> may actually be what you need (in which case you could add e.g.
> "persistent = 18000", where the value equals 18000 seconds, or 5
> hours).
> Also, in real life, when using persistence, ipvsadm may not show
> perfect balancing (e.g. people behind proxy servers will all connect
> from the same address, and therefore get connected to the same server,
> or some people may stay connected far longer than others).  As long as
> the connections are reasonably spread out, you should be OK.
>
> Kind regards,
>
> Herta
>
> _______________________________________________
> Piranha-list mailing list
> [email protected]
> https://www.redhat.com/mailman/listinfo/piranha-list
>

_______________________________________________
Piranha-list mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/piranha-list