raft_update
— State updates#
State changes or actions to handle after calling raft_step()
.
Data types#
-
struct raft_update#
The
raft_update
struct holds information about new state changes or actions that a user should handle after a call toraft_step()
, such as:New data to persist on disk (e.g. new entries or snapshot)
New messages to send to other Raft servers
New term, vote, commit index, etc
Users of the core
raft
struct are responsible for implementing I/O and application code that manages the above state updates.struct raft_update { unsigned flags; struct { ,.. } entries; struct { ,.. } snapshot; struct { ,.. } messages; struct };
Public members#
-
unsigned raft_update.flags#
Bit flags that indicate which particular state change or action should be processed:
#define RAFT_UPDATE_CURRENT_TERM 1 << 0 #define RAFT_UPDATE_VOTED_FOR 1 << 1 #define RAFT_UPDATE_ENTRIES 1 << 2 #define RAFT_UPDATE_SNAPSHOT 1 << 3 #define RAFT_UPDATE_MESSAGES 1 << 4 #define RAFT_UPDATE_STATE 1 << 5 #define RAFT_UPDATE_COMMIT_INDEX 1 << 6 #define RAFT_UPDATE_TIMEOUT 1 << 7
Current Term#
-
RAFT_UPDATE_CURRENT_TERM#
If this bit flag is on, the current term of
raft
struct has changed and must be durably persisted to disk. This has to be done before processing any other change or action (i.e. no messages must be sent until the new term has been persisted).User code can use
raft_current_term()
to get the new term that should be persisted.
Voted for#
-
RAFT_UPDATE_VOTED_FOR#
If this bit flag is on, the server that the
raft
struct has voted for has changed and must be durably persisted to disk. This has to be done before processing any other change or action (i.e. no messages must be sent until the new vote has been persisted).User code can use
raft_voted_for()
to get the server ID that should be persisted as new vote.
Entries#
-
RAFT_UPDATE_ENTRIES#
If this bit flag is on, a new batch of log entries should be persisted to disk, as described by the
raft_update.entries
field.
-
struct [anonymous] raft_update.entries#
Details about new entries to persist.
struct { raft_index index; /* Index of first entry in the batch */ struct raft_entry *batch; /* Array of entries to persist */ unsigned n; /* Number of entries in the array */ } entries;
Snapshot#
-
RAFT_UPDATE_SNAPOSHOT#
If this bit flag is on, a new snapshot chunk should be persisted to disk, as described by the
raft_update.snapshot
field.
-
struct [anonymous] raft_update.snapshot#
Details about new entries to persist.
struct { struct raft_snapshot_metadata metadata; /* Snapshot metadata */ size_t offset; /* Chunk offset */ struct raft_buffer chunk; /* Data chunk */ bool last; /* True if last chunk */ } snapshot;
Messages#
-
RAFT_UPDATE_MESSAGES#
If this bit flag is on, new messages should be sent, as described by the
raft_update.messages
field.
-
struct [anonymous] raft_update.messages#
Details about new entries to persist.
struct { struct raft_message *batch; /* Array of messages to send */ unsigned n; /* Number of messages in the array */ } messages;
State#
-
RAFT_UPDATE_STATE#
If this bit flag is on, the
raft_state
of theraft
struct has changed. The new state can be obtained withraft_state()
.
Commit index#
-
RAFT_UPDATE_COMMIT_INDEX#
If this bit flag is on, the commit index has changed. The new commit index can be obtained with
raft_commit_index()
.
Timeout#
-
RAFT_UPDATE_TIMEOUT#
If this bit flag is on, the time at which the next
RAFT_TIMEOUT
event should be fired has changed. The new time can be obtained withraft_timeout()
.