Riak PG: Distributed Process Groups on Dynamo-style Distributed Storage

Slide 1

Slide 1 text

Riak PG Distributed Process Groups on Dynamo-style Distributed Storage Christopher Meiklejohn Basho Technologies, Inc. Erlang Workshop ’13 Wednesday, October 2, 13

Slide 2

Slide 2 text

The Goal Wednesday, October 2, 13

Slide 3

Slide 3 text

Build a highly-available, fault-tolerant registry. The Goal Wednesday, October 2, 13

Slide 4

Slide 4 text

Understand the tradeo!s. The Goal Wednesday, October 2, 13

Slide 5

Slide 5 text

The Problem Wednesday, October 2, 13

Slide 6

Slide 6 text

Highly-available distributed process groups. The Problem Wednesday, October 2, 13

Slide 7

Slide 7 text

Examples: pg2, gproc The Problem Wednesday, October 2, 13

Slide 8

Slide 8 text

Reappearing groups; synchronous global writes. The pg2 Problem Wednesday, October 2, 13

Slide 9

Slide 9 text

Election deadlocks; con"icts; dynamic clusters. The gproc Problem Wednesday, October 2, 13

Slide 10

Slide 10 text

Wednesday, October 2, 13

Slide 11

Slide 11 text

A more serious problem with using gen_leader is that it requires advance knowledge of all candidate nodes. “Extended Process Registry for Erlang”, Ulf T. Wiger, Erlang Workshop ’07 Wednesday, October 2, 13

Slide 12

Slide 12 text

Perhaps even more serious is gen_leader’s lack of support for dynamically recon#gured networks, and for de-con"icting the states of two leaders (which is presumably the most di$cult part of adding nodes on the "y). “Extended Process Registry for Erlang”, Ulf T. Wiger, Erlang Workshop ’07 Wednesday, October 2, 13

Slide 13

Slide 13 text

The Challenges Wednesday, October 2, 13

Slide 14

Slide 14 text

Dynamic addition and removal of nodes. The 3 Challenges Wednesday, October 2, 13

Slide 15

Slide 15 text

Coordination of state mutation. The 3 Challenges Wednesday, October 2, 13

Slide 16

Slide 16 text

Resolution of con"icting values. The 3 Challenges Wednesday, October 2, 13

Slide 17

Slide 17 text

Riak PG Wednesday, October 2, 13

Slide 18

Slide 18 text

Dynamic membership through virtual nodes. Riak PG Wednesday, October 2, 13

Slide 19

Slide 19 text

Replicated state; quorum reads and writes. Riak PG Wednesday, October 2, 13

Slide 20

Slide 20 text

Con"ict-free resolution with CRDTs. Riak PG Wednesday, October 2, 13

Slide 21

Slide 21 text

Eventually consistent; harvest vs. yield tradeo!. Riak PG Wednesday, October 2, 13

Slide 22

Slide 22 text

Eventual consistency is a consistency model used in distributed computing that informally guarantees that, if no new updates are made to a given data item, eventually all accesses to that item will return the last updated value. “Eventual Consistency”, Wikipedia Wednesday, October 2, 13

Slide 23

Slide 23 text

Our approaches tolerate partial failures by emphasizing simple composition mechanisms that promote fault containment, and by translating possible partial failure modes into engineering mechanisms that provide smoothly degrading functionality rather than lack of availability of the service as a whole. “Harvest, Yield, and Scalable Tolerant Systems”, Fox and Brewer Wednesday, October 2, 13

Slide 24

Slide 24 text

The Requirements Wednesday, October 2, 13

Slide 25

Slide 25 text

Structured names. The Requirements Wednesday, October 2, 13

Slide 26

Slide 26 text

Multiple non-unique names per process. The Requirements Wednesday, October 2, 13

Slide 27

Slide 27 text

Dynamic cluster membership. The Requirements Wednesday, October 2, 13

Slide 28

Slide 28 text

Partition tolerance and con"ict resolution. The Requirements Wednesday, October 2, 13

Slide 29

Slide 29 text

The Applications Wednesday, October 2, 13

Slide 30

Slide 30 text

Service lookup pattern; publish and subscribe. The Applications Wednesday, October 2, 13

Slide 31

Slide 31 text

Trade consistency for availability. The Applications Wednesday, October 2, 13

Slide 32

Slide 32 text

Riak Core; CRDTs The Background Wednesday, October 2, 13

Slide 33

Slide 33 text

Riak Core Wednesday, October 2, 13

Slide 34

Slide 34 text

Erlang implementation of Dynamo. Riak Core Wednesday, October 2, 13

Slide 35

Slide 35 text

Consistent hashing. Riak Core Wednesday, October 2, 13

Slide 36

Slide 36 text

Hash-space partitioning. Riak Core Wednesday, October 2, 13

Slide 37

Slide 37 text

Dynamic membership. Riak Core Wednesday, October 2, 13

Slide 38

Slide 38 text

Replication factor. Riak Core Wednesday, October 2, 13

Slide 39

Slide 39 text

Observed-Removed Set Wednesday, October 2, 13

Slide 40

Slide 40 text

CvRDT; bounded join-semilattice. Observed-Removed Set Wednesday, October 2, 13

Slide 41

Slide 41 text

Set; with merge function computing a LUB. Observed-Removed Set Wednesday, October 2, 13

Slide 42

Slide 42 text

Two G-Sets; preserves monotonicity. Observed-Removed Set Wednesday, October 2, 13

Slide 43

Slide 43 text

[ [{1, a}], [] ] [ [{1, a}], [] ] Wednesday, October 2, 13

Slide 44

Slide 44 text

[ [{1, a}], [] ] [ [{1, a}], [] ] [ [{1, a}, {2, b}], [] ] Wednesday, October 2, 13

Slide 45

Slide 45 text

[ [{1, a}], [] ] [ [{1, a}], [] ] [ [{1, a}, {2, b}], [] ] [ [{1, a}], [{1, a}] ] Wednesday, October 2, 13

Slide 46

Slide 46 text

[ [{1, a}], [] ] [ [{1, a}], [] ] [ [{1, a}, {2, b}], [] ] [ [{1, a}], [{1, a}] ] [ [{1, a}, {2, b}], [{1, a}] ] Wednesday, October 2, 13

Slide 47

Slide 47 text

[ [{1, a}], [] ] [ [{1, a}], [] ] Wednesday, October 2, 13

Slide 48

Slide 48 text

[ [{1, a}], [] ] [ [{1, a}], [] ] [ [{1, a}], [{1, a}] ] [ [{1, a}], [{1, a}] ] Wednesday, October 2, 13

Slide 49

Slide 49 text

[ [{1, a}], [] ] [ [{1, a}], [] ] [ [{1, a}], [{1, a}] ] [ [{1, a}], [{1, a}] ] [ [{1, a}, {2, a}], [{1, a}] ] Wednesday, October 2, 13

Slide 50

Slide 50 text

[ [{1, a}], [] ] [ [{1, a}], [] ] [ [{1, a}], [{1, a}] ] [ [{1, a}], [{1, a}] ] [ [{1, a}, {2, a}], [{1, a}] ] [ [{1, a}, {2, a}], [{1, a}] ] Wednesday, October 2, 13

Slide 51

Slide 51 text

The Implementation Wednesday, October 2, 13

Slide 52

Slide 52 text

Same as pg2; create, join, leave, and members. The Implementation Wednesday, October 2, 13

Slide 53

Slide 53 text

Extended with connected members. The Implementation Wednesday, October 2, 13

Slide 54

Slide 54 text

Membership vnode stores registrations. The Implementation Wednesday, October 2, 13

Slide 55

Slide 55 text

Con"ict-free resolution with OR-set. The Implementation Wednesday, October 2, 13

Slide 56

Slide 56 text

Process pruning; lack of monitors. The Implementation Wednesday, October 2, 13

Slide 57

Slide 57 text

Code examples. The Implementation Wednesday, October 2, 13

Slide 58

Slide 58 text

The Virtual Node Wednesday, October 2, 13

Slide 59

Slide 59 text

%% @doc Respond to a join request. handle_command({join, {ReqId, _}, Group, Pid}, _Sender, #state{groups=Groups0, partition=Partition}=State) -> %% Find existing list of Pids, and add object to it. Pids0 = pids(Groups0, Group, riak_dt_vvorset:new()), Pids = riak_dt_vvorset:update({add, Pid}, Partition, Pids0), %% Store back into the dict. Groups = dict:store(Group, Pids, Groups0), %% Return updated groups. {reply, {ok, ReqId}, State#state{groups=Groups}}; %% @doc Return pids from the dict. -spec pids(dict(), atom(), term()) -> term(). pids(Groups, Group, Default) -> case dict:find(Group, Groups) of {ok, Object} -> Object; _ -> Default end. riak_pg/src/riak_pg_memberships_vnode.erl Wednesday, October 2, 13

Slide 60

Slide 60 text

%% @doc Respond to a leave request. handle_command({leave, {ReqId, _}, Group, Pid}, _Sender, #state{groups=Groups0, partition=Partition}=State) -> %% Find existing list of Pids, and remove object from it. Pids0 = pids(Groups0, Group, riak_dt_vvorset:new()), Pids = riak_dt_vvorset:update({remove, Pid}, Partition, Pids0), %% Store back into the dict. Groups = dict:store(Group, Pids, Groups0), %% Return updated groups. {reply, {ok, ReqId}, State#state{groups=Groups}}; %% @doc Return pids from the dict. -spec pids(dict(), atom(), term()) -> term(). pids(Groups, Group, Default) -> case dict:find(Group, Groups) of {ok, Object} -> Object; _ -> Default end. riak_pg/src/riak_pg_memberships_vnode.erl Wednesday, October 2, 13

Slide 61

Slide 61 text

The Write Coordinator Wednesday, October 2, 13

Slide 62

Slide 62 text

%% @doc Execute the request. execute(timeout, #state{preflist=Preflist, req_id=ReqId, coordinator=Coordinator, group=Group, pid=Pid}=State) -> riak_pg_memberships_vnode:join(Preflist, {ReqId, Coordinator}, Group, Pid), {next_state, waiting, State}. %% @doc Attempt to write to every single node responsible for this %% group. waiting({ok, ReqId}, #state{responses=Responses0, from=From}=State0) -> Responses = Responses0 + 1, State = State0#state{responses=Responses}, case Responses =:= ?W of true -> From ! {ReqId, ok}, {stop, normal, State}; false -> {next_state, waiting, State} end. riak_pg/src/riak_pg_memberships_vnode.erl Wednesday, October 2, 13

Slide 63

Slide 63 text

The Read Coordinator Wednesday, October 2, 13

Slide 64

Slide 64 text

%% @doc Pull a unique list of memberships from replicas, and %% relay the message to it. waiting({ok, _ReqId, IndexNode, Reply}, #state{from=From, req_id=ReqId, num_responses=NumResponses0, replies=Replies0}=State0) -> NumResponses = NumResponses0 + 1, Replies = [{IndexNode, Reply}|Replies0], State = State0#state{num_responses=NumResponses, replies=Replies}, case NumResponses =:= ?R of true -> Pids = riak_dt_vvorset:value(merge(Replies)), From ! {ReqId, ok, Pids}, case NumResponses =:= ?N of true -> {next_state, finalize, State, 0}; false -> {next_state, waiting_n, State} end; false -> {next_state, waiting, State} end. riak_pg/src/riak_pg_members_fsm.erl Wednesday, October 2, 13

Slide 65

Slide 65 text

%% @doc Perform merge of replicas. merge(Replies) -> lists:foldl(fun({_, Pids}, Acc) -> riak_dt_vvorset:merge(Pids, Acc) end, riak_dt_vvorset:new(), Replies). riak_pg/src/riak_pg_members_fsm.erl Wednesday, October 2, 13

Slide 66

Slide 66 text

%% @doc Wait for the remainder of responses from replicas. waiting_n({ok, _ReqId, IndexNode, Reply}, #state{num_responses=NumResponses0, replies=Replies0}=State0) -> NumResponses = NumResponses0 + 1, Replies = [{IndexNode, Reply}|Replies0], State = State0#state{num_responses=NumResponses, replies=Replies}, case NumResponses =:= ?N of true -> {next_state, finalize, State, 0}; false -> {next_state, waiting_n, State} end. riak_pg/src/riak_pg_members_fsm.erl Wednesday, October 2, 13

Slide 67

Slide 67 text

%% @doc Perform read repair. finalize(timeout, #state{replies=Replies}=State) -> Merged = merge(Replies), Pruned = prune(Merged), ok = repair(Replies, State#state{pids=Pruned}), {stop, normal, State}. riak_pg/src/riak_pg_members_fsm.erl Wednesday, October 2, 13

Slide 68

Slide 68 text

%% @doc If the node is connected, and the process is not alive, prune %% it. prune_pid(Pid) when is_pid(Pid) -> lists:member(node(Pid), nodes()) andalso (is_process_alive(node(Pid), Pid) =:= false). %% @doc Remote call to determine if process is alive or not; assume if %% the node fails communication it is, since we have no proof it %% is not. is_process_alive(Node, Pid) -> case rpc:call(Node, erlang, is_process_alive, [Pid]) of {badrpc, _} -> true; Value -> Value end. %% @doc Based on connected nodes, prune out processes that no longer %% exist. prune(Set) -> Pids0 = riak_dt_vvorset:value(Set), lists:foldl(fun(Pid, Pids) -> case prune_pid(Pid) of true -> riak_dt_vvorset:update({remove, Pid}, none, Pids); false -> Pids end end, Set, Pids0). riak_pg/src/riak_pg_members_fsm.erl Wednesday, October 2, 13

Slide 69

Slide 69 text

The Evaluation Wednesday, October 2, 13

Slide 70

Slide 70 text

Partitions heal without con"icts. The Evaluation Wednesday, October 2, 13

Slide 71

Slide 71 text

Howl; CloudI Process Groups; Riak Pipe The Related Work Wednesday, October 2, 13

Slide 72

Slide 72 text

The Future Work Wednesday, October 2, 13

Slide 73

Slide 73 text

CRDT garbage collection. The Future Work Wednesday, October 2, 13

Slide 74

Slide 74 text

Active anti-entropy mechanism. The Future Work Wednesday, October 2, 13

Slide 75

Slide 75 text

The Conclusion Wednesday, October 2, 13

Slide 76

Slide 76 text

http://github.com/cmeiklejohn/riak_pg Thanks! Questions? Wednesday, October 2, 13