|
Julian,
Thanks again for this -- the delay is now down to 4.5seconds per connect,
after I cleaned up the DNS stuff. The only challenge I seem to have left
involves packets looking like this:
19:18:52.050236 192.168.0.1.netbios-ns > lvs.monkey.com.netbios-ns:
>>> NBT UDP PACKET(137): QUERY; REQUEST; BROADCAST
TrnID=0xB303
OpCode=0
NmFlags=0x1
Rcode=0
QueryCount=1
AnswerCount=0
AuthorityCount=0
AddressRecCount=0
QuestionRecords:
Name=* NameType=0x00 (Workstation)
QuestionType=
19:18:52.050364 lvs.monkey.com > 192.168.0.1: icmp: lvs.monkey.com udp port
netbios-ns unreachable [tos 0xc0]
192.168.0.1 is the client machine, lvs.monkey.com is the director. The
client sends off 3 of these packets (at ~1.5s delay apiece) -- is this an
issue with the app itself (MSSQL), or is this a networking situation
involving windows netbios broadcasts? (or other?)
Any insight greatly appreciated.
Best,
- Mike
----- Original Message -----
From: "Julian Anastasov" <ja@xxxxxx>
To: "Michael Brannigan" <goombah@xxxxxxxxx>
Cc: <lvs-users@xxxxxxxxxxxxxxxxxxxxxx>
Sent: Thursday, February 28, 2002 1:41 AM
Subject: Re: Help -- confirm correct LVS-NAT packet stream?
>
> Hello,
>
> On Wed, 27 Feb 2002, Michael Brannigan wrote:
>
> > Julian,
> >
> > Thanks for this - I isolated it to some SQL flakiness, and I now have
> > LVS/NAT directing 5 MSSQL boxes on Windows 2000,
successfully....however...
> > the connection times are very slow! if I have a realserver connect to
> > another realserver, the connection happens in sub-second time.. however,
a
> > client machine on the other side of the LVS director experiences ~30
second
> > connect times. What gives?
>
> I'm not an expert in these apps but in the LVS history
> the only known delays are from DNS resolutions and IDENT protocol
> performed from the real servers.
>
> 30 seconds is an usual time for delay caused from IDENT.
> OTOH, IDENT usually works with NAT if you correctly NAT all ports,
> not only the port for the real service.
>
> > RSs can ping client
> > Client cannot ping RSs
> > Director can ping both
> > RSs trace to client via director
>
> looks good
>
> > Director kernel is 2.2.19, running ipvs 1.0.8
> >
> > Any ideas of things I should check for? Is my above scenario correct?
I've
>
> Looks correct
>
> > followed the "quick" troubleshooters pretty carefully, but I'm still a
linux
> > newbie, so everytime I've twiddled things, I've broken the config and
had to
> > fallback on the configure script. Nothing is being saturated on the
> > machine - bandwidth, ram, cpu -- all practically idle.
>
> Try to trace all communications and to spot the problem,
> may be there should be other traffic that is not handled.
>
> > Thanks in advance,
> >
> > - Mike
>
> Regards
>
> --
> Julian Anastasov <ja@xxxxxx>
>
>
> _______________________________________________
> LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
> Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
> or go to http://www.in-addr.de/mailman/listinfo/lvs-users
_________________________________________________________
Do You Yahoo!?
Get your free @yahoo.com address at http://mail.yahoo.com
|
