LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Failover query

To: "'LinuxVirtualServer.org users mailing list.'" <lvs-users@xxxxxxxxxxxxxxxxxxxxxx>
Subject: Failover query
From: "Vic Berdin" <vic@xxxxxxxxxxx>
Date: Tue, 22 Nov 2005 10:45:53 +0800
Hi,

I'm doing LVS/HA Direct Routing using two old Linux-2.4.x boxes. My arp
block is implemented using the old "noarp" module. High Availability with
heartbeat and load balancing with ldirectord works well. Failover works ok
at some point. But what I wish to resolve with my setup are the unwanted
failovers.

The scenario is like this, once the current (default/configured) Master node
goes down, and the Secondary node takes over, the default Master node that
went down will still tend to grab Directorship status from the already
(currently) running Master node. This unwanted failover happens within a
_still_undetermined_ period after the newly raised node re-joins the
cluster. Sometimes this failover happens almost immediately, sometimes after
many hours (even after a few days) of run time.

I'm still using the old release heartbeat-1.2.3 and ldirectord that came
along with it.

My configurations for both boxes are as follows:

#-----------------------
# /etc/ldirectord.conf
#-----------------------
checktimeout=250
checkinterval=5
logfile="local0"
quiescent=yes
virtual=203.176.19.77:5432
        fallback=127.0.0.1:5432 gate
        service=none
        protocol=tcp
        checktype=connect
        real=203.176.19.75:5432 gate
        real=203.176.19.76:5432 gate
        scheduler=rr

virtual=203.176.19.77:5000
        fallback=127.0.0.1:5000 gate
        service=none
        protocol=tcp
        checktype=connect
        real=203.176.19.75:5000 gate
        real=203.176.19.76:5000 gate
        scheduler=rr


#-----------------------
# /etc/ha.d/ha.cf
#-----------------------
debugfile /var/log/ha-debug
logfile /var/log/ha-log
logfacility local0
keepalive 2
deadtime 5
udpport 694
auto_failback off
bcast eth1
mcast eth1 224.0.0.0 694 1 0
node cluster1
node cluster2

#-----------------------
# /etc/ha.d/haresources
#-----------------------
cluster1 IPaddr::203.176.19.77 ldirectord::/etc/ldirectord.cf noarp setstate
dhcpdsvc

During these unwanted failovers, I am not able to detect any abnormal, or
sudden "down-state" of the services that I'm performing LVS/HA with.

What could be wrong with my config settings, as I have "auto_failback" set
to "off"? My build environment restricts me from building/using the most
recent packages. Thus, I am tasked to (as much as possible) find a fix for
this issue w/o replacing (upgrading) my current packages.

Any suggestions/help/comments are greately appreciated.


Best regards,
Vic

-- 
No virus found in this outgoing message.
Checked by AVG Free Edition.
Version: 7.1.362 / Virus Database: 267.13.5/177 - Release Date: 11/21/2005
 


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