Re: Streamline HA/LB Squid cluster first connection problem

Vic, Thank you again.


I did some tests and noted that the problem isn't a problem of cache for one
or another node, but the obsenceof  the arp entry to my virtual IP at the
first try of connect.


to make the tests i did these steps:

-- look at arp table: empty.
-- try to connect: Failury
-- look at arp table:

Interface: 172.16.8.80 --- 0x2
  Endereço IP           Endereço físico       Tipo
  172.16.8.78           00-0c-29-ce-45-50     dinâmico


-- Try to connect: Connection OK

--Look to arp table:

C:\Documents and Settings\admin>arp -a

Interface: 172.16.8.80 --- 0x2
  Endereço IP           Endereço físico       Tipo
  172.16.8.78           00-0c-29-ce-45-50     dinâmico
  172.16.8.83           00-0c-29-ce-45-50     dinâmico
  172.16.8.85           00-0c-29-70-98-b0     dinâmico

--Clear the real servers entry on arp table

-- Look at arp table:

Interface: 172.16.8.80 --- 0x2
  Endereço IP           Endereço físico       Tipo
  172.16.8.78           00-0c-29-ce-45-50     dinâmico

--Try to connect: OK

--clean the arp table.

-- ping on two nodes real servers.

--look at arp table:

Interface: 172.16.8.80 --- 0x2
  Endereço IP           Endereço físico       Tipo

  172.16.8.83           00-0c-29-ce-45-50     dinâmico
  172.16.8.85           00-0c-29-70-98-b0     dinâmico



--try to connect: Failury



--look at arp table:

Interface: 172.16.8.80 --- 0x2
  Endereço IP           Endereço físico       Tipo
  172.16.8.78           00-0c-29-ce-45-50     dinâmico
  172.16.8.83           00-0c-29-ce-45-50     dinâmico
  172.16.8.85           00-0c-29-70-98-b0     dinâmico



-- try to connect: OK

So I noted that when my Virtual IP is listed on my arp table its work fine,
if it's absent, even so the real ips are listeds, I can't connect at first
time.



**I tried that fix that u suggested decreasing the arp table expire time to
3 seconds but the same occurs





thanx

On 6/19/06, vic <[email protected]> wrote:

>   Hi Denis,
>
>
>
> > *From:* [email protected] [mailto:
> [email protected]] *On Behalf Of *Denis
> > *Sent:* Tuesday, June 20, 2006 5:53 AM
> > *To:* [email protected]
> > *Subject:* Re: Streamline HA/LB Squid cluster first connection problem
>
> >
>
> > Hello Vic,
>
> >
>
> > thanx for your help.
>
> >
>
> > what u said make sense, but if this is the real error, how could I solve
> it without reconfigure all workstations?
>
> > they are about 2 thousands...
>
>
>
> First of all, make sure if what I suggested fixes your problem. arp cache
> issue indeed became a problem for my LVS-DR (direct routing) setup.
> Pre-creating an arpcachelife.reg registry setting and informing my users
> via e-mail to install it, most of whom didn't even bother reading exactly
> what it is for, did not become a problem. And by the way, ArpCacheLife
> applies not only in M$ WinNT but also in M$ WinXP/XPE or other M$ Windowz
> variants that are based in NT technology. Arp cache didn't become an issue
> for Linux workstations.
>
>
>
>
>
> Best regards,
>
> Vic
>
>
>