|
Just tried to use a standart Kernel with HZ=1000 and clock=pit, still hitting
ip_vs_async error.
-----Ursprüngliche Nachricht-----
Von: lvs-users-bounces@xxxxxxxxxxxxxxxxxxxxxx
[mailto:lvs-users-bounces@xxxxxxxxxxxxxxxxxxxxxx] Im Auftrag von Michael von
Guggenberg
Gesendet: Donnerstag, 30. August 2007 10:33
An: LinuxVirtualServer.org users mailing list.
Betreff: Re: [lvs-users] LVS Centos all in Vmware
Im using recompiled kernels with HZ=100, http://vmware.xaox.net/centos/4/, and
ntpd, so time isnt a problem in my VM's.
-----Ursprüngliche Nachricht-----
Von: lvs-users-bounces@xxxxxxxxxxxxxxxxxxxxxx
[mailto:lvs-users-bounces@xxxxxxxxxxxxxxxxxxxxxx] Im Auftrag von Graeme Fowler
Gesendet: Donnerstag, 30. August 2007 10:22
An: LinuxVirtualServer.org users mailing list.
Betreff: Re: [lvs-users] LVS Centos all in Vmware
On Thu, 2007-08-30 at 10:16 +0200, Michael von Guggenberg wrote:
> Ok tried the mentioned solutions, disabeld tx rx chechsums with
> ethtool, tried to echo 0 > /proc/sys/net/ipv4/tcp_sack on the involved
> servers and moved the RS to a diffrent vmware host, symptom remains the same.
> Now i have noticed that while one apache server is shutdown i get the
> ip_vs_send_async error on every second hit.
Are your guests running with any specific "clock=" bootloader options? It's
quite plausible that the sync traffic is getting mangled by the ability of
vmware guests to slip time back and forth.
Graeme
|
