HEARTBEAT API

Introduction:

This document describes the application program interface for heartbeat. The
following are the interfaces available for heartbeat.

 1. ll_cluster_new()
 2. singon()
 3. signoff()
 4. delete()
 5. set_msg_callback()
 6. set_nstatus_callback()
 7. set _ifstatus_callback()
 8. init_nodewalk()
 9. nextnode()
10. end_nodewalk()
11. init_ifwalk()
12. nextif()
13. end_ifwalk()
14. if_status()
15. sendclustermsg()
16. sendnodemsg()
17. inputfd()
18. msgready()
19. setmsgsignal()
20. rcvmsg()
21. readmsg()
22. setfmode()
23. get_parameter()
24. get_deadtime()
25. get_keepalive()
26. get_mynodeid()
27. get_logfacility()
28. get_resources()
29. errmsg()

Description:

The details of each of the api functions are as follows:

ll_cluster_new()

ll_cluster_t *ll_cluster_new(const char *type)

Parameter description:
    'type' identifies the service provider. It's value must currently be
    "heartbeat".
Description:
    Register with the heartbeat library and create a client instance of type
    "heartbeat"
Return Value:
    returns a handle to the heartbeat library instance. This handle is a opaque
    object with a set of member functions. The member functions provide the
    necessary mechanism to interact with the heartbeat daemon. Returns NULL on
    failure.
Observation:
    Why is 'type' parameter required when we know for sure that we are only
    interacting with the heartbeat daemon? Or is it that the library wants to
    be generic and wants to interact with any arbitrary daemon? Yes.  The
    latter.

signon()

int *signon(struct ll_cluster cinfo*, const char *service)

Parameter Description:
'cinfo'
    is the handle to the library instance got through ll_cluster_new()
    'service' is the type of service, which the calling process registers with
    the heartbeat daemon.
Description:
    This function sets up a communication link with the local heartbeat daemon
    and registers itself for the 'service' service. In the case of casual
    client the clientid is NULL. But in the case of non-casual client the
    clientid is non-NULL and it indicates the service to which this client
    wants to sign on for. This special communication link setup with the
    heartbeat daemon is a fifo. Non-casual clients interact using a fifo whose
    name is derived from the pid of the clients process. Casual clients
    interact with heartbeat daemon using the fifo whose name is derived from
    the 'service'string.

    The heartbeat daemon has no notion for 'service'. Only the clients
    registered with the heartbeat daemons for the same service, have the notion
    for that service.  'service' mostly is a mechanism of providing primitive
    group services to a set of non-casual clients registered for that service.

Return value:
    On success returns HA_OK
    On failure returns HA_FAIL

signoff()

int signoff(struct ll_cluster *cinfo)

Parameter Description:
'cinfo'
    is the handle to the library instance got through ll_cluster_new()
Description:
    This function deletes the communication link with the local heartbeat
    daemon and unsubscribe itself from being a heartbeat client.
Return Value:
    On success it returns HA_OK.
    On failure it returns HA_FAIL.

delete()


int delete(struct ll_cluster *cinfo)

Parameter Description:
cinfo'
    is the handle to the library instance got through ll_cluster_new()
Description:
    This operation is reverse of ll_cluster_new(). It unsubscribes itself from
    the heartbeat library and cleans up the corresponding datastructures.
Return Value:
    On success returns HA_OK
    On failure returns HA_FAIL

set_msg_callback()

int set_msg_callback(struct ll_cluster *cinfo,   const char *msgtype,  
llc_msg_callback_t callback, void * p)

Parameter Description:

cinfo'
    is the handle to the library instance got through ll_cluster_new()
'msgtype'
    is the type of message on receipt of which the 'callback' is called.
'callback'
    is the function to be called on receipt of a message of type 'msgtype'.
'p'
    is the parameter to be passed to the 'callback' function.
Description:
    This function registers a callback function along with its parameter. The
    callback function is called by the heartbeat library on receipt of a
    message of type 'msgtype' from the heartbeat daemon.
NOTE:
    Neither the heartbeat library nor heartbeat daemon interpret 'msgtype'. The
    'msgtype' is interpreted only by the client.  Clients can coin their own
    msgtype as long as all the clients signed up for the same service can
    interpret it.
Return Value:
    On success HA_OK is returned.
    On failure HA_FAIL is returned.

set_nstatus_callback()

int set_nstatus_callback(struct ll_cluster *cinfo, llc_nstatus_callback_t cbf,
void *p)

Parameter Description:
'cinfo'
    is the handle to the library instance got through ll_cluster_new
'cbf'
    is the function pointer to be called on change of status of any node in the
    cluster.
'p'
    is the parameter to be passed to the callback function.
Description:
    This function registers a callback function along with its parameter. The
    callback function is called by the heartbeat library on change of status of
    any node in the cluster.
Return Value:
    On success HA_OK is returned.
Observation:
    There is a bug in this function. It does not validate the 'cinfo'
    parameter'. It should return HA_FAIL on failure.

set_ifstatus_callback()

int set_ifstatus_callback(struct ll_cluster *cinfo, llc_ifstatus_callback_t
cbf, void *p)

Parameter Description:
'cinfo'
    is the handle to the library instance got through ll_cluster_new()
cbf'
    is the function pointer to be called on change of status of any interface
    in the cluster.
'p'
    is the parameter to be passed for the callback function.
Description:
    This function registers a callback function along with its parameter. The
    callback function is called by the heartbeat library on change of status of
    any network interface in the cluster in the view of the current cluster
    node.
Return Value:
    On success HA_OK is returned.

    Also there is a bug in this function. It does not validate the 'cinfo'
    parameter'. It should return HA_FAIL on failure.

init_nodewalk()


int init_nodewalk(struct ll_cluster_t *cinfo)

Parameter Description:
'cinfo'
    is the handle to the library instance got through ll_cluster_new
Description:
    This function talks to the heartbeat daemon and retrieves the set of nodes
    that are part of the cluster. If the retrieval is successful HA_OK is
    returned.
Return Value:
    On success HA_OK is returned
    On failure HA_FAIL is returned

nextnode()

char *nextnode(ll_cluster_t *cinfo)

Parameter Description:
'cinfo'
    is the handle to the library instance gotten through ll_cluster_new().
Description/Return Value:,
    This function returns the next node in the cluster, which is cached in the
    heartbeat library. If no more node is available it returns NULL.

end_nodewalk()


int end_nodewalk(ll_cluster_t *cinfo)

Parameter Description:
'cinfo'
    is the handle to the library instance got through ll_cluster_new().
Description:
    This clears the cluster node information cached in the heartbeat library.
Return Value:
    returns HA_OK on success  and HA_FAIL on failure.

init_ifwalk()

int init_ifwalk(struct ll_cluster_t *cinfo, char *host)

Parameter Description:
'cinfo'
    is the handle to the library instance got through ll_cluster_new
'host'
    is the node in the cluster whose network interfaces are of interest.
Description:
    This function talks to the heartbeat daemon and caches all the network
    interface information associated with 'host' node. This caching is done in
    the heartbeat library.
Return Value:
    If the network interface information can be successfully retrieved, HA_OK
    is returned. Failure returns HA_FAIL.

nextif()

  char * nextif(ll_cluster_t *cinfo)

Parameter Description:
'cinfo'
    is the handle to the library instance got through ll_cluster_new().
Description:
    This function returns the next network interface of the 'host' node in the
    cluster, which is cached in the heartbeat library.
NOTE:
    The 'host' is the node of the cluster specified during the call to
    init_ifwalk()
Return Value:
    If no more interfaces are available it returns NULL.

end_ifwalk()


int end_ifwalk(ll_cluster_t *cinfo)

Parameter Description:
'cinfo'
    is the handle to the library instance got through ll_cluster_new().
Description:
    This clears the network interface information cached in the heartbeat
    library.
Return Value:
    returns HA_OK on success and HA_FAIL of failure.

if_status()

char *  if_status(ll_cluster_t *cinfo, const char *host,
const char *ifname)

Parameter Description:
'cinfo'
    is the handle to the library instance got through ll_cluster_new().
'host'
    is the node name whose interface's status is queried.
'ifname' is the name of the interface whose status is queried.
Description:
    This function returns a string that specifies the state of the interface
    'ifname' on cluster node 'host'
Return Value:
    It returns "up" if the link is active and "dead" if the link is down. NULL
    if there was a error.

sendclustermsg()

int sendclustermsg(ll_cluster_t *cinfo, struct ha_msg* msg)

Parameter Description:
'cinfo'
    is the handle to the library instance got through ll_cluster_new().
'msg'
    is the message to be sent to all non-casual clients signed up for the same
    service as that of the calling client.
Description:
    This function broadcasts the message to all members of the group.
Return Value:
    This function returns HA_FAIL if called by a casual client or if there was
    some other failure.
    On success it returns HA_OK.

sendnodemsg()

int sendnodemsg(ll_cluster_t *cinfo, struct ha_msg* msg,
const char *nodename)

Parameter Description:
'cinfo'
    is the handle to the library instance got through ll_cluster_new().
'msg'
    is the message to be sent to a non-casual client signed on to the same
    service on node 'nodename'.
Description:
    This function sends the message to a non-casual client residing on node
    'nodename'. The client belongs to the same group as that of the calling
    client.
Return Value:
    This function returns HA_FAIL if called by a casual client or if there was
    some other failure.
    On success it returns HA_OK.

inputfd()

int inputfd(ll_cluster * cinfo)

Parameter Description:
'cinfo'
    is the handle to the library instance got through ll_cluster_new().

Description/Return Value:
    returns the file descriptor of the communication link setup with the local
    heartbeat daemon.
    On failure it returns -1.
Observation:
    This function is needed in order to be able to perform a select function
    which selects on incoming API messages. It should not be used for any other
    purpose.

msgready()

int msgready(ll_cluster_t *cinfo)

Parameter Description:
'cinfo'
    is the handle to the library instance got through ll_cluster_new().
Description:
    checks if there is any message to be read.
Return Value:
    If a message is ready to be read returns non zero value. Otherwise returns
    0.

setmsgsignal()

int setmsgsignal(ll_cluster_t *cinfo, int nsig)

Parameter Description:
'cinfo'
    is the handle to the library instance got through ll_cluster_new().
'nsig'
    is the signal to be used by heartbeat daemon to signal the client process.
Description:
    This function informs the heartbeat daemon send signal number 'nsig'
    whenever a message arrives for it.
Return Value:
    returns HA_OK on success
    and HA_FAIL on failure

rcvmsg()

int rcvmsg(ll_cluster_t *cinfo,  int blocking)

Parameter Description:
'cinfo'
    is the handle to the library instance got through ll_cluster_new().
'blocking'
    indicates whether to block for the receipt of a message or not.
Description/Return Value:
    This function returns true if a message for which a callback function was
    not registered is received.

    If blocking is set, then it waits till a message arrives. But it returns
    true only if that message does not have a callback registered.

NOTE:
    In all cases, received messages are deleted by this call.
Observation:
    This interface is provided as a convenience to allow for clients which only
    process certain message types through registered callback functions to
    never have to worry about the possibility of failing to destroy their
    messages. If you process all messages you handle with callbacks, then this
    function is a convenient choice. Otherwise, you likely need to use readmsg
    ().

readmsg()


struct ha_msg * readmsg(ll_cluster_t *cinfo, int blocking)

Parameter Description:
'cinfo'
    is the handle to the library instance got through ll_cluster_new().
'blocking'
    indicates whether to block for the receipt of a message or not.
Description/Return Value:
    This function is same as rcvmsg except that it returns messages without
    registered callbacks to the caller (and does not delete them). It returns
    the next message for which a callback is not registered.

    If 'blocking' is set to FALSE, it will not wait for a message to arrive. It
    returns NULL if there is not currently a message ready to read which is not
    intercepted by a callback function.

NOTE:
    The message returned must be disposed off by the caller by calling
    ha_msg_del().

setfmode()


int setfmode(ll_cluster_t *cinfo, int mode)

Parameter Description:
'cinfo'
    is the handle to the library instance got through ll_cluster_new().
'mode'
    indicates the type of messages to be delivered to the client and to filter
    out the rest.

    The modes have the following meanings:

    LLC_FILTER_DEFAULT
        All messages destined to this client are received, along with all that
        are addressed to all clients.
    LLC_FILTER_PMODE
        All messages, but filter out heartbeats that don't tell us anything
        new.
    LLC_FILTER_ALLHB
        All heartbeats including those that don't change status.
    LLC_FILTER_RAW
        All packets from all interfaces, even duplicates. Packets with
        authentication errors are still ignored.

Description:
    This function sets up the type of messages that the client is interested;
    to be received from the local heartbeat daemon. It is mainly used for
    debugging.
Return Value:
    returns HA_OK on success and HA_FAIL on failure


get_parameter()

    char* get_parameter(struct ll_cluster_t * cinfo, char *pname)
Parameter Description:
'cinfo'
    is the handle to the library instance got through ll_cluster_new
'pname'
    is the name of the parameter to retrieve
Description:
    This function retrieves the value of the named parameter from heartbeat.
    The parameters supported are listed in <hb_api.h>.
Return Value:
    If the network interface information can be successfully retrieved, a
    pointer to a malloced string is returned. NULL is returned on failure.
     This value is malloced, and the caller must free the return value is no
    longer needed.

get_deadtime()

    long get_deadtime(struct ll_cluster_t * cinfo)
Parameter Description:
'cinfo'
    is the handle to the library instance got through ll_cluster_new
Description:
    This function retrieves the deadtime value from heartbeat.  The value
    returned is in milliseconds.
Return Value:
    The return value is the deadtime, measured in milliseconds.

get_keepalive()

    long get_keepalive(struct ll_cluster_t * cinfo)
Parameter Description:
'cinfo'
    is the handle to the library instance got through ll_cluster_new().

Description:
    This function retrieves the keepalive value from heartbeat.  The value
    returned is in milliseconds.
Return Value:
    The return value is the keepalive interval, measured in milliseconds.

get_mynodeid()

const char * get_mynodeid(ll_cluster_t *cinfo)

Parameter Description:
'cinfo'
    is the handle to the library instance got through ll_cluster_new
Description/Return Value:
    This function returns the node id of the current cluster node.  It is
    generally (though not always) the same as the uname -n name of the current
    server.  The return value is a static value.


get_logfacility()

    long get_logfacility(struct ll_cluster_t * cinfo)
Parameter Description:
'cinfo'
    is the handle to the library instance got through ll_cluster_new().

Description:
    This function retrieves the logfacility value from heartbeat.  
Return Value:
    The return value is the logfacility currently being used by heartbeat.  If
    there is no log facility being used by heartbeat, then it will return -1.

get_resources()

const char * get_resources(ll_cluster_t *cinfo)

Parameter Description:
'cinfo'
    is the handle to the library instance got through ll_cluster_new
Description/Return Value:
    This function returns the current resource allocation as supported by the
    current resource management function.  The return value is a static string
    containing one of "local", "foreign", "all", "none" or "transition"..
     "local" means the local nodes is supporting only local resources.
     "foreign" means the local node is supporting only foreign resources (those
    belonging to the other node).  "all" means the local node is supporting all
    cluster resources.  "none" means the local node is not providing any
    cluster resources.  "transition" means that resources are currently in
    transition.   If heartbeat is not providing resource management, the return
    value will be NULL.


errmsg()

const char * errmsg(ll_cluster_t *cinfo)

Parameter Description:
'cinfo'
    is the handle to the library instance got through ll_cluster_new
Description/Return Value:
    This function returns the last set of error messages from the previous API
    call. This is not a very wonderful interface (according to alanr)


