2 // Replicated state machine implementation with a primary and several
3 // backups. The primary receives requests, assigns each a view stamp (a
4 // vid, and a sequence number) in the order of reception, and forwards
5 // them to all backups. A backup executes requests in the order that
6 // the primary stamps them and replies with an OK to the primary. The
7 // primary executes the request after it receives OKs from all backups,
8 // and sends the reply back to the client.
10 // The config module will tell the RSM about a new view. If the
11 // primary in the previous view is a member of the new view, then it
12 // will stay the primary. Otherwise, the smallest numbered node of
13 // the previous view will be the new primary. In either case, the new
14 // primary will be a node from the previous view. The configuration
15 // module constructs the sequence of views for the RSM and the RSM
16 // ensures there will be always one primary, who was a member of the
19 // When a new node starts, the recovery thread is in charge of joining
20 // the RSM. It will collect the internal RSM state from the primary;
21 // the primary asks the config module to add the new node and returns
22 // to the joining the internal RSM state (e.g., paxos log). Since
23 // there is only one primary, all joins happen in well-defined total
26 // The recovery thread also runs during a view change (e.g, when a node
27 // has failed). After a failure some of the backups could have
28 // processed a request that the primary has not, but those results are
29 // not visible to clients (since the primary responds). If the
30 // primary of the previous view is in the current view, then it will
31 // be the primary and its state is authoritive: the backups download
32 // from the primary the current state. A primary waits until all
33 // backups have downloaded the state. Once the RSM is in sync, the
34 // primary accepts requests again from clients. If one of the backups
35 // is the new primary, then its state is authoritative. In either
36 // scenario, the next view uses a node as primary that has the state
37 // resulting from processing all acknowledged client requests.
38 // Therefore, if the nodes sync up before processing the next request,
39 // the next view will have the correct state.
41 // While the RSM in a view change (i.e., a node has failed, a new view
42 // has been formed, but the sync hasn't completed), another failure
43 // could happen, which complicates a view change. During syncing the
44 // primary or backups can timeout, and initiate another Paxos round.
45 // There are 2 variables that RSM uses to keep track in what state it
47 // - inviewchange: a node has failed and the RSM is performing a view change
48 // - insync: this node is syncing its state
50 // If inviewchange is false and a node is the primary, then it can
51 // process client requests. If it is true, clients are told to retry
52 // later again. While inviewchange is true, the RSM may go through several
53 // member list changes, one by one. After a member list
54 // change completes, the nodes tries to sync. If the sync complets,
55 // the view change completes (and inviewchange is set to false). If
56 // the sync fails, the node may start another member list change
57 // (inviewchange = true and insync = false).
59 // The implementation should be used only with servers that run all
60 // requests run to completion; in particular, a request shouldn't
61 // block. If a request blocks, the backup won't respond to the
62 // primary, and the primary won't execute the request. A request may
63 // send an RPC to another host, but the RPC should be a one-way
64 // message to that host; the backup shouldn't do anything based on the
65 // response or execute after the response, because it is not
66 // guaranteed that all backup will receive the same response and
67 // execute in the same order.
69 // The implementation can be viewed as a layered system:
70 // RSM module ---- in charge of replication
71 // config module ---- in charge of view management
72 // Paxos module ---- in charge of running Paxos to agree on a value
74 // Each module has threads and internal locks. Furthermore, a thread
75 // may call down through the layers (e.g., to run Paxos's proposer).
76 // When Paxos's acceptor accepts a new value for an instance, a thread
77 // will invoke an upcall to inform higher layers of the new value.
78 // The rule is that a module releases its internal locks before it
79 // upcalls, but can keep its locks when calling down.
82 #include "rsm_client.h"
87 rsm_state_transfer::~rsm_state_transfer() {}
89 rsm::rsm(const string & _first, const string & _me) : primary(_first)
91 cfg = unique_ptr<config>(new config(_first, _me, this));
94 // Commit the first view here. We can not have acceptor::acceptor
95 // do the commit, since at that time this->cfg is not initialized
98 rsmrpc = cfg->get_rpcs();
99 rsmrpc->reg(rsm_client_protocol::invoke, &rsm::client_invoke, this);
100 rsmrpc->reg(rsm_client_protocol::members, &rsm::client_members, this);
101 rsmrpc->reg(rsm_protocol::invoke, &rsm::invoke, this);
102 rsmrpc->reg(rsm_protocol::transferreq, &rsm::transferreq, this);
103 rsmrpc->reg(rsm_protocol::transferdonereq, &rsm::transferdonereq, this);
104 rsmrpc->reg(rsm_protocol::joinreq, &rsm::joinreq, this);
106 // tester must be on different port, otherwise it may partition itself
107 testsvr.reset(new rpcs((in_port_t)std::stoi(_me) + 1));
108 testsvr->reg(rsm_test_protocol::net_repair, &rsm::test_net_repairreq, this);
109 testsvr->reg(rsm_test_protocol::breakpoint, &rsm::breakpointreq, this);
116 thread(&rsm::recovery, this).detach();
119 // The recovery thread runs this function
120 void rsm::recovery() {
125 while (!cfg->ismember(cfg->myaddr(), vid_commit)) {
126 // XXX iannucci 2013/09/15 -- I don't understand whether accessing
127 // cfg->view_id in this manner involves a race. I suspect not.
128 if (join(primary, ml)) {
130 commit_change(cfg->view_id(), ml);
133 std::this_thread::sleep_for(milliseconds(3000)); // XXX make another node in cfg primary?
137 vid_insync = vid_commit;
138 LOG << "sync vid_insync " << vid_insync;
139 if (primary == cfg->myaddr()) {
140 r = sync_with_backups(ml);
142 r = sync_with_primary(ml);
146 // If there was a commited viewchange during the synchronization, restart
148 if (vid_insync != vid_commit)
152 myvs.vid = vid_commit;
154 inviewchange = false;
156 LOG << "go to sleep " << insync << " " << inviewchange;
157 recovery_cond.wait(ml);
161 bool rsm::sync_with_backups(lock & rsm_mutex_lock) {
162 rsm_mutex_lock.unlock();
164 // Make sure that the state of lock_server is stable during
165 // synchronization; otherwise, the primary's state may be more recent
166 // than replicas after the synchronization.
167 lock invoke_mutex_lock(invoke_mutex);
168 // By acquiring and releasing the invoke_mutex once, we make sure that
169 // the state of lock_server will not be changed until all
170 // replicas are synchronized. The reason is that client_invoke arrives
171 // after this point of time will see inviewchange == true, and returns
174 rsm_mutex_lock.lock();
175 // Start accepting synchronization request (statetransferreq) now!
177 cfg->get_view(vid_insync, backups);
178 backups.erase(std::find(backups.begin(), backups.end(), cfg->myaddr()));
179 LOG << "backups " << backups;
180 sync_cond.wait(rsm_mutex_lock);
186 bool rsm::sync_with_primary(lock & rsm_mutex_lock) {
187 // Remember the primary of vid_insync
189 while (vid_insync == vid_commit) {
190 if (statetransfer(m, rsm_mutex_lock))
193 return statetransferdone(m, rsm_mutex_lock);
198 // Call to transfer state from m to the local node.
199 // Assumes that rsm_mutex is already held.
201 bool rsm::statetransfer(const string & m, lock & rsm_mutex_lock)
203 rsm_protocol::transferres r;
205 LOG << "contact " << m << " w. my last_myvs(" << last_myvs.vid << "," << last_myvs.seqno << ")";
208 rsm_mutex_lock.unlock();
209 cl = rpcc::bind_cached(m);
211 ret = cl->call_timeout(rsm_protocol::transferreq, milliseconds(100),
212 r, cfg->myaddr(), last_myvs, vid_insync);
214 rsm_mutex_lock.lock();
216 if (cl == 0 || ret != rsm_protocol::OK) {
217 LOG << "couldn't reach " << m << " " << std::hex << cl << " " << std::dec << ret;
220 if (stf && last_myvs != r.last) {
221 stf->unmarshal_state(r.state);
224 LOG << "transfer from " << m << " success, vs(" << last_myvs.vid << "," << last_myvs.seqno << ")";
228 bool rsm::statetransferdone(const string & m, lock & rsm_mutex_lock) {
229 rsm_mutex_lock.unlock();
231 if (auto cl = rpcc::bind_cached(m)) {
233 auto ret = (rsm_protocol::status)cl->call(rsm_protocol::transferdonereq, r, cfg->myaddr(), vid_insync);
234 done = (ret == rsm_protocol::OK);
236 rsm_mutex_lock.lock();
241 bool rsm::join(const string & m, lock & rsm_mutex_lock) {
245 LOG << "contacting " << m << " mylast (" << last_myvs.vid << "," << last_myvs.seqno << ")";
247 rsm_mutex_lock.unlock();
248 auto cl = rpcc::bind_cached(m);
250 ret = cl->call_timeout(rsm_protocol::joinreq, milliseconds(12000), log, cfg->myaddr(), last_myvs);
251 rsm_mutex_lock.lock();
253 if (cl == 0 || ret != rsm_protocol::OK) {
254 LOG << "couldn't reach " << m << " " << std::hex << cl << " " << std::dec << ret;
257 LOG << "succeeded " << log;
263 // Config informs rsm whenever it has successfully
264 // completed a view change
266 void rsm::commit_change(unsigned vid) {
268 commit_change(vid, ml);
269 if (cfg->ismember(cfg->myaddr(), vid_commit))
273 void rsm::commit_change(unsigned vid, lock &) {
274 if (vid <= vid_commit)
276 LOG << "new view (" << vid << ") last vs (" << last_myvs.vid << ","
277 << last_myvs.seqno << ") " << primary << " insync " << insync;
281 recovery_cond.notify_one();
282 sync_cond.notify_one();
283 if (cfg->ismember(cfg->myaddr(), vid_commit))
288 void rsm::execute(rpc_protocol::proc_id_t procno, const string & req, string & r) {
290 handler *h = procs[procno];
293 auto ret = (rsm_protocol::status)(*h)(unmarshall(req, false), rep);
294 r = marshall(ret, rep.content()).content();
297 static void logHexString(locked_ostream && log, const string & s) {
298 log << std::setfill('0') << std::setw(2) << std::hex;
299 for (size_t i=0; i<s.size(); i++)
300 log << (unsigned int)(unsigned char)s[i];
304 // Clients call client_invoke to invoke a procedure on the replicated state
305 // machine: the primary receives the request, assigns it a sequence
306 // number, and invokes it on all members of the replicated state
309 rsm_client_protocol::status rsm::client_invoke(string & r, rpc_protocol::proc_id_t procno, const string & req) {
310 LOG << "invoke procno 0x" << std::hex << procno;
311 lock ml(invoke_mutex);
317 LOG << "Checking for inviewchange";
319 return rsm_client_protocol::BUSY;
320 LOG << "Checking for primacy";
321 myaddr = cfg->myaddr();
322 if (primary != myaddr)
323 return rsm_client_protocol::NOTPRIMARY;
324 LOG << "Assigning a viewstamp";
325 cfg->get_view(vid_commit, m);
326 // assign the RPC the next viewstamp number
331 // send an invoke RPC to all slaves in the current view with a timeout of 1 second
332 LOG << "Invoking slaves";
333 for (unsigned i = 0; i < m.size(); i++) {
334 if (m[i] != myaddr) {
335 // if invoke on slave fails, return rsm_client_protocol::BUSY
336 LOG << "Sending invoke to " << m[i];
337 auto cl = rpcc::bind_cached(m[i]);
339 return rsm_client_protocol::BUSY;
341 auto ret = (rsm_protocol::status)cl->call_timeout(rsm_protocol::invoke, milliseconds(100), ignored_rval, procno, vs, req);
342 LOG << "Invoke returned " << ret;
343 if (ret != rsm_protocol::OK)
344 return rsm_client_protocol::BUSY;
346 lock rsm_mutex_lock(rsm_mutex);
347 partition1(rsm_mutex_lock);
350 logHexString(LOG, req);
351 execute(procno, req, r);
352 logHexString(LOG, r);
354 return rsm_client_protocol::OK;
358 // The primary calls the internal invoke at each member of the
359 // replicated state machine
361 // the replica must execute requests in order (with no gaps)
362 // according to requests' seqno
364 rsm_protocol::status rsm::invoke(int &, rpc_protocol::proc_id_t proc, viewstamp vs, const string & req) {
365 LOG << "invoke procno 0x" << std::hex << proc;
366 lock ml(invoke_mutex);
371 // check if !inviewchange
372 LOG << "Checking for view change";
374 return rsm_protocol::ERR;
376 LOG << "Checking for slave status";
377 myaddr = cfg->myaddr();
378 if (primary == myaddr)
379 return rsm_protocol::ERR;
380 cfg->get_view(vid_commit, m);
381 if (std::find(m.begin(), m.end(), myaddr) == m.end())
382 return rsm_protocol::ERR;
383 // check sequence number
384 LOG << "Checking sequence number";
386 return rsm_protocol::ERR;
390 execute(proc, req, r);
393 return rsm_protocol::OK;
397 // RPC handler: Send back the local node's state to the caller
399 rsm_protocol::status rsm::transferreq(rsm_protocol::transferres & r, const string & src,
400 viewstamp last, unsigned vid) {
402 LOG << "transferreq from " << src << " (" << last.vid << "," << last.seqno << ") vs ("
403 << last_myvs.vid << "," << last_myvs.seqno << ")";
404 if (!insync || vid != vid_insync)
405 return rsm_protocol::BUSY;
406 if (stf && last != last_myvs)
407 r.state = stf->marshal_state();
409 return rsm_protocol::OK;
413 // RPC handler: Inform the local node (the primary) that node m has synchronized
416 rsm_protocol::status rsm::transferdonereq(int &, const string & m, unsigned vid) {
418 if (!insync || vid != vid_insync)
419 return rsm_protocol::BUSY;
420 backups.erase(std::find(backups.begin(), backups.end(), m));
422 sync_cond.notify_one();
423 return rsm_protocol::OK;
426 // a node that wants to join an RSM as a server sends a
427 // joinreq to the RSM's current primary; this is the
428 // handler for that RPC.
429 rsm_protocol::status rsm::joinreq(string & log, const string & m, viewstamp last) {
430 auto ret = rsm_protocol::OK;
433 LOG << "join request from " << m << "; last=(" << last.vid << "," << last.seqno << "), mylast=("
434 << last_myvs.vid << "," << last_myvs.seqno << ")";
435 if (cfg->ismember(m, vid_commit)) {
436 LOG << m << " is still a member -- nothing to do";
438 } else if (cfg->myaddr() != primary) {
439 LOG << "but I, " << cfg->myaddr() << ", am not the primary, " << primary << "!";
440 ret = rsm_protocol::BUSY;
442 // We cache vid_commit to avoid adding m to a view which already contains
443 // m due to race condition
444 LOG << "calling down to config layer";
445 unsigned vid_cache = vid_commit;
449 succ = cfg->add(m, vid_cache);
452 if (cfg->ismember(m, cfg->view_id())) {
454 LOG << "ret " << ret << " log " << log;
456 LOG << "failed; proposer couldn't add " << succ;
457 ret = rsm_protocol::BUSY;
464 // RPC handler: Responds with the list of known nodes for fall-back on a
467 rsm_client_protocol::status rsm::client_members(vector<string> & r, int) {
470 cfg->get_view(vid_commit, m);
471 m.push_back(primary);
473 LOG << "return " << m << " m " << primary;
474 return rsm_client_protocol::OK;
477 // if primary is member of new view, that node is primary
478 // otherwise, the lowest number node of the previous view.
479 // caller should hold rsm_mutex
480 void rsm::set_primary(unsigned vid) {
482 cfg->get_view(vid, c);
483 cfg->get_view(vid - 1, p);
484 VERIFY (c.size() > 0);
486 if (isamember(primary,c)) {
487 LOG << "primary stays " << primary;
491 VERIFY(p.size() > 0);
492 for (unsigned i = 0; i < p.size(); i++) {
493 if (isamember(p[i], c)) {
495 LOG << "primary is " << primary;
502 bool rsm::amiprimary() {
504 return primary == cfg->myaddr() && !inviewchange;
508 // Test RPCs -- simulate partitions and failures
510 void rsm::net_repair(bool heal, lock & rsm_mutex_lock) {
511 VERIFY(rsm_mutex_lock);
513 cfg->get_view(vid_commit, m);
514 for (unsigned i = 0; i < m.size(); i++) {
515 if (m[i] != cfg->myaddr()) {
516 LOG << "member " << m[i] << " " << heal;
517 if (auto cl = rpcc::bind_cached(m[i]))
518 cl->set_reachable(heal);
521 rsmrpc->set_reachable(heal);
524 rsm_test_protocol::status rsm::test_net_repairreq(rsm_test_protocol::status & r, int heal) {
526 LOG << "heal " << heal << " (dopartition "
527 << dopartition << ", partitioned " << partitioned << ")";
529 net_repair(heal, ml);
533 return r = rsm_test_protocol::OK;
536 // simulate failure at breakpoint 1 and 2
538 void rsm::breakpoint(int b) {
539 if (breakpoints[b-1]) {
540 LOG << "Dying at breakpoint " << b << " in rsm!";
545 void rsm::partition1(lock & rsm_mutex_lock) {
547 net_repair(false, rsm_mutex_lock);
553 rsm_test_protocol::status rsm::breakpointreq(rsm_test_protocol::status & r, int b) {
554 r = rsm_test_protocol::OK;
556 LOG << "breakpoint " << b;
557 if (b == 1) breakpoints[1-1] = true;
558 else if (b == 2) breakpoints[2-1] = true;
559 else if (b == 3 || b == 4) cfg->breakpoint(b);
560 else r = rsm_test_protocol::ERR;