> I tried running the mc_conn_tester script, but wasn't able to trigger the > problem with that. > > The interesting thing is that even if I close down all apps that are > communicating with memcached, eg. Apache (due to the pecl memcached modules), > the memcached daemon still has thousands of CLOSE_WAIT connections. There > are no other outstanding apps or connections > talking to it, yet there are thousands of sockets stuck in a CLOSE_WAIT state. > > Shouldn't memcached be cleaning those up as the sockets hard close?
Is memcached handling any traffic at all when you see it in this state? If you strace it is it still reading/writing/anything happening? It sort of sound like it's hardlocking and your clients are trying to close off (not persistent or whatever), but it's just not doing anything at all. Which means it could be related to the aforementioned bug. If existing connections do still work, it's possibly something else entirely. You can see this any number of ways: does bandwidth to the memcached box drop to nothing, if you write a test script that connects and does sets/fetches once per second, does it hang? etc. -- --- You received this message because you are subscribed to the Google Groups "memcached" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/groups/opt_out.
