LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Re: Our LVS/DR backends freezes

To: Joseph Mack NA3T <jmack@xxxxxxxx>
Subject: Re: Our LVS/DR backends freezes
Cc: "LinuxVirtualServer.org users mailing list." <lvs-users@xxxxxxxxxxxxxxxxxxxxxx>
Cc: Olle �Vstlund <olle@xxxxxxxxxxx>
From: Horms <horms@xxxxxxxxxxxx>
Date: Wed, 29 Nov 2006 10:36:14 +0900
On Tue, Nov 28, 2006 at 09:45:22AM -0800, Joseph Mack NA3T wrote:
> On Tue, 28 Nov 2006, Olle �~Vstlund wrote:
> 
> >Look for yourself. These are figures from the director when all
> >realservers are open (weight = 10)
> >
> >IP Virtual Server version 1.2.0 (size=4096)
> >Prot LocalAddress:Port Scheduler Flags
> > -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
> >TCP  192.168.3.10:80 wrr
> > -> 192.168.3.105:80             Route   10     755        33987
> > -> 192.168.3.107:80             Route   10     842        35426
> 
> >These are figures from the director when the realserver with interfaces
> >192.168.3.5 and 192.168.3.105 have been closed (weight = 0) for 5
> >minutes:
> >
> >IP Virtual Server version 1.2.0 (size=4096)
> >Prot LocalAddress:Port Scheduler Flags
> > -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
> >TCP  192.168.3.10:80 wrr
> > -> 192.168.3.105:80             Route   0      623        33687
> > -> 192.168.3.107:80             Route   10     929        36252
> 
> >-----------------------
> >192.168.3.105  NONE            45 (max/mean expire time 59/31)
> >192.168.3.105  FIN_WAIT     15534 (max/mean expire time 113/29)
> >192.168.3.105  ESTABLISHED    623 (max/mean expire time 600/36)
> >192.168.3.105  CLOSE        18125 (max/mean expire time 59/29)
> >192.168.3.105  SYN_RECV        14 (max/mean expire time 59/24)
> -----------------------
> 
> 
> OK machine x.x.x.105 has its weight changed to 0 and the InActConn hardly 
> drops 
> in 5 minutes. The netstat figures on the realservers after 5 mins of 
> inactivity 
> still show a lot of CLOSE and FIN_WAIT. These should all be gone. Presumably 
> this is not happening with the standalone server?

Do the connections eventually go away, or is there a body of them that
stay forever. If it is the latter it sound s lot like a race condition.
And probably one that has been fixed in newer kernels.


-- 
Horms
  H: http://www.vergenet.net/~horms/
  W: http://www.valinux.co.jp/en/


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