|
On Tue, 3 May 2005, Graeme Fowler wrote:
> On Tue 03 May 2005 18:56:04 BST , Mark de Vries
> <markdv.lvsuser@xxxxxxxxxx> wrote:
> > ANY help welcome... Right now I'm rebooting boxes every 24-48 hours... not
> > exactly what I had in mind when I thought LVS would help in creating a
> > high-available service... :(
>
> Hrm... as Joe said, maybe this is UDP related; that would be rather
> unusual though.
>
> I have a three-node DNS cluster using DR running a custom-compiled
> 2.4.23 kernel, but it isn't getting the query rate you're stating. I
> have no problems with it whatsoever. For comparison can you do:
I'm on 2.6.10 (kernel.org)
> ipvsadm -Ln
> ipvsadm -Ln --rate
>
> sanitising the output to protect the innocent (or guilty!), obviously ;-)
I have 9 VIPs, each with the same 4 realservers so thats a lot of lines...
I'll leave some stuff out...
> [root@dns02 root]# ipvsadm -Ln
> IP Virtual Server version 1.0.10 (size=65536)
> Prot LocalAddress:Port Scheduler Flags
> -> RemoteAddress:Port Forward Weight ActiveConn InActConn
> FWM 5 wlc
> -> 192.168.80.4:53 Route 10 0 17287
> -> 192.168.80.3:53 Route 10 0 17285
> -> 192.168.80.2:53 Local 10 0 17286
Only the ones with the highest numbers:
-> 10.31.66.181:53 Route 10 0 69327
-> 10.31.66.180:53 Route 10 0 67688
-> 10.31.66.179:53 Route 10 0 70264
-> 10.31.66.178:53 Route 10 0 70084
-> 10.31.66.181:53 Route 10 0 16394
-> 10.31.66.180:53 Route 10 0 15003
-> 10.31.66.179:53 Route 10 0 17054
-> 10.31.66.178:53 Route 10 0 16129
-> 10.31.66.181:53 Route 10 0 22597
-> 10.31.66.180:53 Route 10 0 22259
-> 10.31.66.179:53 Route 10 0 22664
-> 10.31.66.178:53 Route 10 0 22653
-> 10.31.66.181:53 Route 10 0 15913
-> 10.31.66.180:53 Route 10 0 15967
-> 10.31.66.179:53 Route 10 0 16100
-> 10.31.66.178:53 Route 10 0 16228
> [root@dns02 root]# ipvsadm -Ln --rate
> IP Virtual Server version 1.0.10 (size=65536)
> Prot LocalAddress:Port CPS InPPS OutPPS InBPS OutBPS
> -> RemoteAddress:Port
> FWM 5 127 312 0 22546 0
> -> 192.168.80.4:53 42 107 0 7811 0
> -> 192.168.80.3:53 40 101 0 7237 0
> -> 192.168.80.2:53 45 104 0 7498 0
Only the VIPs:
ipvsadm -Ln --rate | grep UDP
UDP :53 98 244 0 16069 0
UDP :53 266 724 0 46439 0
UDP :53 4 13 0 861 0
UDP :53 309 705 0 45634 0
UDP :53 460 1015 0 65995 0
UDP :53 24 93 0 6094 0
UDP :53 6 31 0 2090 0
UDP :53 15 25 0 1579 0
UDP :53 34 80 0 5458 0
I would say my rates are significantly higher... at peak hour the total is
even ~800 InPPS more...
> Have you fiddled with the bucket size, BTW?
I'm going to compile a new kernel tomorrow and raise that one... Someone
else compiled this one and I see it's still at the default 4k. Hope that
helps. I think if it does it's still a bug though. Larger bucket size
should only make finding connections more efficient (less CPU cycles)...
Rgds,
Mark.
|
