Mcc (Mnesia Cluster Cache)
mcc is Mnesia Cluster Cache, which support expiration and cluster.
usage
There are some steps to use mcc:
- define table
- define model
- start expiration process
define table
Mcc support expiration and each table need one assistant table to store expiration information.
use Mcc.Model.Table,
table_opts: [
type: :set,
ram_copies: [node()],
storage_properties: [
ets: [:compressed, read_concurrency: true]
]
],
expiration_opts: [
expiration_table: AccountExp,
main_table: MccTest.Support.Table.Account,
size_limit: 100,
# 300M
memory_limit: 300
]
defstruct([:id, :user_profile], true)
Like above table definition, the table_opts is from mnesia create table. And expiration_opts has several option:
expiration_table which table to store expiration information
main_table current table
size_limit the table size limit
memory_limit the memory usage limit, unit is
mega byte
After define the table, need define the fields for the table, user can use defstruct/1,2 to define the table’s fields. Multi-fields use [] to organize, and fields should use atom. The second parameter for defstruct is represent if use cache expiration mechanism.
For the assistant table, the definition will be like:
defmodule MccTest.Support.Table.Account.Exp do
@moduledoc """
Table for expiration Account table.
"""
use Mcc.Model.Table,
table_opts: [
type: :ordered_set,
ram_copies: [node()],
storage_properties: [
ets: [:compressed, read_concurrency: true]
]
]
defstruct [:key, :value]
def put(key, value) do
put(%__MODULE__{key: key, value: value})
end
enddefine model
After define table, user need to define model, it’s not difficult:
defmodule MccTest.Support.Cache do
@moduledoc """
Definition for cache test.
"""
use Mcc.Model
import_table(MccTest.Support.Table.Account.Exp)
import_table(MccTest.Support.Table.Account)
enddefine model is import the tables which defined before.
start expiration process
mcc use one long-lived process to scan the mnesia table for expiration. User(application) need to start the process explicitly:
MccTest.Support.Cache.start_expiration_process()Generally speaking, it will be put into application start.
[One complete example could be found in the test cases]
Change log
[v1.0.1]
-
Fixed bug when put with ttl and remove
delete_object/1
-
Fixed bug when put with ttl and remove
[v1.0.2]
- try to connect kernel optional nodes before join cluster
- added log after joined in cluster and create/copy table
[v1.0.3]
- support different level for operation consistency
-
remove
delete/2function, will delete expiration table automatically -
need not pass
cache_timewhen put record, will calculatenowreal-time