Lazy flush doesn't work
I have a cluster consisted of 2 servers. There is a membase bucket which resides on both servers. My client is using memcache protocol and it is connected to the bucket via stand-alone moxi. The client is supposed to store an infinite numbers of keys. Before the bucket hits its RAM quota, it manages to store around 3M keys. Then the client sends a flush_all command and it properly invalidates all of the keys. When i run the client again, it manages to store only 100k keys a then it throws an error:
[NOTICE@1291817675.774581] mcm_storage_cmd():3344: unable to store value: Interrupted system call: temporary failure
I think, that there is a problem with that lazy flushing. It looks like that membase doesn't free the memory at all... When I restart the server the bucket is empty again.
This problem also appears on server-side moxi.
There is no problem with memcache bucket. Flushing works fine there...
Membase server is 220.127.116.11 64bit and running on Debian GNU/Linux 5.0
Moxi is 18.104.22.168 64bit
Client is using memcache C API 1.4.0.rc2-1
Can someone point me to the right direction please?