1.4.8, memory_backend max_memory not honored?
landsidel.allen at gmail.com
Tue Jun 24 11:59:49 EDT 2014
Test cluster with 2 machines running 1.4.8 with the memory backend.
app_config has ring_creation_size 64, max_memory 4. If I understand the
documentation correctly this should result in a maximum memory footprint
(minus overhead) of 256M. Each node has 2G RAM, 2G of swap. After
running for a while constantly adding data, without ever explicitly
removing any, the memory footprint outstrips physical ram and will
eventually consume all available swap space and be killed.
Is this a known issue in 1.4.8 that's fixed in 1.4.9, or is there other
configuration settings that need adjusted to reduce the footprint?
Initially I had set it up to top out at 1GB but when it started
exhausting swap, I progressively reduced max_memory until I arrived
where I am now, with no difference at all.
More information about the riak-users