Risk of atom table exhaustion from dynamic "maint" node naming [JIRA: TOOLS-149]

[angrycub]

When nodetool makes a VM to perform an operation against a running Erlang application, it includes the PID of that specific script invocation:

https://github.com/basho/node_package/blob/develop/priv/base/nodetool#L191

This has the unintended consequence of potentially exhausting the atom table on a long-running node, or a node upon which nodetool-backed maintenance activities are run frequently.

[smarj]

This looks fixed on node_package 4.0 and later, including develop. Riak 2.0.8, Riak 2.2.1 and later use 4.0.1, so they no longer exhibit this behavior. There is no safe version of 2.1.x.