|
Hi folks,
Got a bunch of problems I am unable to track down here. These didn't
become apparent on our production servers until we took them down for
imaging. Totally sucked to find them out at 3:00am on the morning we
went live. grrrr
1) LVS caches dead routes even when restarted. (Sometimes requires a
complete reboot to get rid of them though twice a reboot STILL didn't
get rid of the cached routes.
2) Must totally kill heartbeat and ldirectord on both nodes for any
configuration change to take place. A /etc/rc.d/init.d/heartbeat restart
doesn't do it.
3) Must reapply ipchains rulesets completely after restart on BOTH
nodes.
4) Must restart DNS services on both nodes with Primary node being first
for DNS to work right, even though that is not a ldirectord'd service.
5) On system restart, heartbeat installs multiple copies (3 to 4) of the
default route in the route table (found by running the route command).
6) If you hit the LVS IP with the IP itself with multiple vhosts defined
in apache, it does not shunt the request to the default domain as it
should. But if you come with the domain name(s) it works right.
7) Does not correctly remove dead subnodes from the availability list
requiring the need to restart the LVS nodes.
Using kernel version 2.2.14-5.0.14.um.3smp and IP Virtual Server version
0.9.11 and matching heartbeat.
I would like to upgrade to the current version of the LVS software but
am loathe to do it due to the down time required to switch over.
--
David D.W. Downey RHCE, UNIX/Linux/Win 9x Administrator
Linux Systems Administrator Member OSWG, LPI, SAGE, HTML Writers
Guild
QIXO, Inc. Certified Internet Security
Specialist
http://www.QIXO.com W: (408) 514-6400 F: (408) 516-9090
|
