m00b

The fix is much easier than that.

If you’re using memcached for object cache support, flush it and the upgrade message will be gone.

I have never had this problem with apc or xcache though.

Let me just start by saying that the memcached daemon does not require much cpu to run and that a load such as you have is perfectly normal. If you’re only using it for page caching then it won’t be using anywhere near the amount of memory you have specified anyway. FYI memcached GROWS to the size you set in the config, starting at 64mb, but won’t reach 512mb until that much is actually in its cache.

Experiance has shown that memcached is certianly not the best method to go with. For a start, every connection to the daemon has to run through a php process, which doesnt actually reduce load at all, it increases it, if anything.

I have a very similar setup to you, in that I loadbalance between two WWW servers. Except we only use memcached for minification.For page caching we setup an nfs share on the pgcache directory and then have nginx serve static w3tc files it creates in Disk (Enhanced) mode. We also use xcache on each of the www servers, although not in conjuction with w3tc, just as an opcode cacher. This setup has proven to be pretty fast and reliable.

To answer your question, after all the pointless crap I said above, I don’t believe there are options in memcached you can tweak to increase hitrate. You can of course get w3tc to do more caching with the daemon, object cache, db cache etc. From experiance however, this doesnt work well.

If you continue to use memcached, add OPTIONS=”-l ip.address” to its config and perhaps iptables rules to prevent unauthorized access to the daemon. Or run it on an internal ip.

Sorry if this is blindingly obvious to you, I don’t know how much previous experiance you have with memcached, but I hope some of this helps!

I found out the issue was with the memcached object cache plugin i was using (https://www.ads-software.com/extend/plugins/memcached/).

Just for future refference, although I think it was probably pretty obvious for users more experiance than I, you need to restart or flush your memcached backend to get over this issue.