Hello,

I do not look for high availability memcached solution but I want to
know how to deal with memcached in our specific environment.

For each apps/server we have heartbeat and intention is to use memory
of the secondary server in HA cluster (around 300GB in total). During
fail over heartbeat simply kills memcached and sets it up later on new
secondary (which was primary host before fail over) so initial
memcached server list looks like:
appA-1
appA-2
appB-1
appB-2
...
appX-1
appX-2

The problem is that only secondary (app..-2) servers are initially
available and when fail over happens one secondary server disappears
and one primary server becomes available.

Our HA cluster has 3 IP addresses: one virtual and one address for
each node (one for app,,-1 and another for app...-2) thus I have
following choices:

- access memcached via virtual IP (not sure if it is possible or its
limitations)
this solution looks less complicated and I'm not aware of any
drawbacks.
- use all primary/secondary IPs
for this it's necessary to choose different hashing alg but I have no
idea which one could cope with removing and adding servers.

I do not care about data loss because it is possible to fetch data
from database but I need access cluster from PHP (must) and Perl (nice
to have) and memcached is 1.2.8 (from Fedora)

Thanks,
--
Ondrej

Reply via email to