Generic functions for implementing snmp objects in a database
The module snmp_generic contains generic functions for implementing tables (and variables) using the SNMP built-in database or Mnesia. These default functions are used if no instrumentation function is provided for a managed object in a MIB. Sometimes, it might be necessary to customize the behaviour of the default functions. For example, in some situations a trap should be sent if a row is deleted or modified, or some hardware is to be informed, when information is changed.
The overall structure is shown in the following figure:
+---------------+
| SNMP Agent |
+- - - - - - - -+
| MIB |
+---------------+
|
Association file (associates a MIB object with
| snmp_generic:table_funct
| snmp_generic:variable_func)
+--------------------------------------+
| snmp_generic | Support for get-next,
| | RowStatus operations
+----------------------+---------------+
| snmpa_local_db | Mnesia | Database
+--------------+-------+---------------+
| dets | ets |
| (persistent) | |
+--------------+-------+
Each function takes the argument NameDb, which is a tuple {Name, Db}, to identify which database the functions should use. Name is the symbolic name of the managed object as defined in the MIB, and Db is either volatile, persistent, or mnesia. If it is mnesia, all variables are stored in the Mnesia table snmp_variables which must be a table with two attributes (not a Mnesia SNMP table). The SNMP tables are stored in Mnesia tables with the same names as the SNMP tables. All functions assume that a Mnesia table exists with the correct name and attributes. It is the programmer's responsibility to ensure this. Specifically, if variables are stored in Mnesia, the table snmp_variables must be created by the programmer. The record definition for this table is defined in the file snmp/include/snmp_types.hrl.
If an instrumentation function in the association file for a variable myVar does not have a name when compiling an MIB, the compiler generates an entry.
{myVar, {snmp_generic, variable_func, [{myVar, Db]}}.
And for a table:
{myTable, {snmp_generic, table_func, [{myTable, Db]}}.
In the functions defined below, the following types are used:
name_db() = {name(), db()}
name() = atom()
db() = volatile | persistent | mnesia
row_index() = [int()]
columns() = [column()] | [{column(), value()}]
column() = int()
value() = term()
row_index(): Denotes the last part of the OID which specifies the index of the row in the table (see RFC1212, 4.1.6 for more information about INDEX).
columns(): Is a list of column numbers in the case of a get operation, and a list of column numbers and values in the case of a set operation.
get_status_col(Name, Cols)
get_status_col(NameDb, Cols) -> {ok, StatusVal} | false
Types:
Name = name()
NameDb = name_db()
Cols = columns()
StatusVal = term()
Gets the value of the status column from Cols.
This function can be used in instrumentation functions for is_set_ok, undo or set to check if the status column of a table is modified.
get_index_types(Name)
Types:
Name = name()
Gets the index types of Name
This function can be used in instrumentation functions to retrieve the index types part of the table info.
get_table_info(Name, Item) -> table_info_result()
Types:
Name = name()
Item = table_item() | all
table_item() = nbr_of_cols | defvals | status_col | not_accessible | index_types | first_accessible | first_own_index
table_info_result() = Value | [{table_item(), Value}]
Value = term()
Get a specific table info item or, if Item has the value all, a two tuple list (property list) is instead returned with all the items and their respctive values of the given table.
This function can be used in instrumentation functions to retrieve a given part of the table info.
table_func(Op1, NameDb)
table_func(Op2, RowIndex, Cols, NameDb) -> Ret
Types:
Op1 = new | delete
Op2 = get | next | is_set_ok | set | undo
NameDb = name_db()
RowIndex = row_index()
Cols = columns()
Ret = term()
This is the default instrumentation function for tables.
*
The new function creates the table if it does not exist, but only if the database is the SNMP internal db.
*
The delete function does not delete the table from the database since unloading an MIB does not necessarily mean that the table should be destroyed.
*
The is_set_ok function checks that a row which is to be modified or deleted exists, and that a row which is to be created does not exist.
*
The undo function does nothing.
*
The set function checks if it has enough information to make the row change its status from notReady to notInService (when a row has been been set to createAndWait). If a row is set to createAndWait, columns without a value are set to noinit. If Mnesia is used, the set functionality is handled within a transaction.
If it is possible for a manager to create or delete rows in the table, there must be a RowStatus column for is_set_ok, set and undo to work properly.
The function returns according to the specification of an instrumentation function.
table_get_elements(NameDb, RowIndex, Cols) -> Values
Types:
NameDb = name_db()
RowIndex = row_index()
Cols = columns()
Values = [value() | noinit]
Returns a list with values for all columns in Cols. If a column is undefined, its value is noinit.
table_next(NameDb, RestOid) -> RowIndex | endOfTable
Types:
NameDb = name_db()
RestOid = [int()]
RowIndex = row_index()
Finds the indices of the next row in the table. RestOid does not have to specify an existing row.
table_row_exists(NameDb, RowIndex) -> bool()
Types:
NameDb = name_db()
RowIndex = row_index()
Checks if a row in a table exists.
table_set_elements(NameDb, RowIndex, Cols) -> bool()
Types:
NameDb = name_db()
RowIndex = row_index()
Cols = columns()
Sets the elements in Cols to the row specified by RowIndex. No checks are performed on the new values.
If the Mnesia database is used, this function calls mnesia:write to store the values. This means that this function must be called from within a transaction (mnesia:transaction/1 or mnesia:dirty/1).
variable_func(Op1, NameDb)
variable_func(Op2, Val, NameDb) -> Ret
Types:
Op1 = new | delete | get
Op2 = is_set_ok | set | undo
NameDb = name_db()
Val = value()
Ret = term()
This is the default instrumentation function for variables.
The new function creates a new variable in the database with a default value as defined in the MIB, or a zero value (depending on the type).
The delete function does not delete the variable from the database.
The function returns according to the specification of an instrumentation function.
variable_get(NameDb) -> {value, Value} | undefined
Types:
NameDb = name_db()
Value = value()
Gets the value of a variable.
variable_set(NameDb, NewVal) -> true | false
Types:
NameDb = name_db()
NewVal = value()
Sets a new value to a variable. The variable is created if it does not exist. No checks are made on the type of the new value.
Returns false if the NameDb argument is incorrectly specified, otherwise true.
The following example shows an implementation of a table which is stored in Mnesia, but with some checks performed at set-request operations.
myTable_func(new, NameDb) -> % pass unchanged
snmp_generic:table_func(new, NameDb).
myTable_func(delete, NameDb) -> % pass unchanged
snmp_generic:table_func(delete, NameDb).
%% change row
myTable_func(is_set_ok, RowIndex, Cols, NameDb) ->
case snmp_generic:table_func(is_set_ok, RowIndex,
Cols, NameDb) of
{noError, 0} ->
myApplication:is_set_ok(RowIndex, Cols);
Err ->
Err
end;
myTable_func(set, RowIndex, Cols, NameDb) ->
case snmp_generic:table_func(set, RowIndex, Cols,
NameDb),
{noError, 0} ->
% Now the row is updated, tell the application
myApplication:update(RowIndex, Cols);
Err ->
Err
end;
myTable_func(Op, RowIndex, Cols, NameDb) -> % pass unchanged
snmp_generic:table_func(Op, RowIndex, Cols, NameDb).
The .funcs file would look like:
{myTable, {myModule, myTable_func, [{myTable, mnesia}]}}.