LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Re: ipvs-0.8.0 available

To: Julian Anastasov <ja@xxxxxx>
Subject: Re: ipvs-0.8.0 available
Cc: lvs-users@xxxxxxxxxxxxxxxxxxxxxx
From: Radu-Adrian Feurdean <raf@xxxxxxxx>
Date: Wed, 9 May 2001 16:21:51 +0200 (CEST)
On Wed, 9 May 2001, Julian Anastasov wrote:

> 
>       Hello,
> 
> On Wed, 9 May 2001, Radu-Adrian Feurdean wrote:
> 
> > We have encountered crashes on 2.4 + ipvs-0.2 when using SMP kernels
> > (including the latest versions) when we started to load-balance high-traffic
> > servers (3 real servers each with ~380 active connections, total traffic of
> > about 24Mbps). It crashed systematically after about 4 hours of
> > working. However, it works fine with UP kernel for one week. The conenction
> > table bits was (and still is) set to 18 (256K simultaneous connections).
> 
>       What is the evidence this is a LVS problem? Any crash report?
> Is Netfilter running on the same box?

We're not 100% sure it's ipvs causing the problem. We have on identical
hardware configurations 2.4.2 running packet filtering, apache and a
mutant of SuSe ftp-proxy and they are running fine, even at high traffic.
The ipvs director only runs packet filtering and ipvs, and at high traffic
crashes after a few hours.

No crash-logs (they don't arrive to syslog), however, there's plenty of:

IPVS: did not set timer with refcnt=2, called from e0946f84

(address varies). Turning off debugging in ipvs at compile-time didn't help.

And yes, there is netfilter (aka firewall + MASQUERADE) on the same box.
(here are some other minor problems, like packets - all of them - that don't
pass through chain OUTPUT, table mangle).


Radu-Adrian Feurdean
mailto: raf@xxxxxxxx
-------------------------------------------------------------------
"If the night is silent enough you can hear a Windows NT rebooting"



<Prev in Thread] Current Thread [Next in Thread>