void ReplSet::haveNewConfig(ReplSetConfig& newConfig, bool addComment) {
bo comment;
if( addComment )
comment = BSON( "msg" << "Reconfig set" << "version" << newConfig.version );
newConfig.saveConfigLocally(comment);
try {
BSONObj oldConfForAudit = config().asBson();
BSONObj newConfForAudit = newConfig.asBson();
audit::logReplSetReconfig(ClientBasic::getCurrent(),
&oldConfForAudit,
&newConfForAudit);
if (initFromConfig(newConfig, true)) {
log() << "replSet replSetReconfig new config saved locally" << rsLog;
}
}
catch(DBException& e) {
log() << "replSet error unexpected exception in haveNewConfig() : " << e.toString() << rsLog;
_fatal();
}
catch(...) {
log() << "replSet error unexpected exception in haveNewConfig()" << rsLog;
_fatal();
}
}
StatusWith<int> validateConfigForInitiate(ReplicationCoordinatorExternalState* externalState,
const ReplSetConfig& newConfig,
ServiceContext* ctx) {
Status status = newConfig.validate();
if (!status.isOK()) {
return StatusWith<int>(status);
}
status = newConfig.checkIfWriteConcernCanBeSatisfied(newConfig.getDefaultWriteConcern());
if (!status.isOK()) {
return status.withContext(
"Found invalid default write concern in 'getLastErrorDefaults' field");
}
status = validateArbiterPriorities(newConfig);
if (!status.isOK()) {
return StatusWith<int>(status);
}
if (newConfig.getConfigVersion() != 1) {
return StatusWith<int>(ErrorCodes::NewReplicaSetConfigurationIncompatible,
str::stream() << "Configuration used to initiate a replica set must "
<< " have version 1, but found "
<< newConfig.getConfigVersion());
}
return findSelfInConfigIfElectable(externalState, newConfig, ctx);
}
// Our own config must be the first one.
bool ReplSetImpl::_loadConfigFinish(OperationContext* txn, vector<ReplSetConfig*>& cfgs) {
int v = -1;
ReplSetConfig *highest = 0;
int myVersion = -2000;
int n = 0;
for (vector<ReplSetConfig*>::iterator i = cfgs.begin(); i != cfgs.end(); i++) {
ReplSetConfig* cfg = *i;
DEV { LOG(1) << n+1 << " config shows version " << cfg->version << rsLog; }
if (++n == 1) myVersion = cfg->version;
if (cfg->ok() && cfg->version > v) {
highest = cfg;
v = cfg->version;
}
}
verify(highest);
if (!initFromConfig(txn, *highest))
return false;
if (highest->version > myVersion && highest->version >= 0) {
log() << "replSet got config version " << highest->version
<< " from a remote, saving locally" << rsLog;
highest->saveConfigLocally(txn, BSONObj());
}
return true;
}
StatusWith<int> validateConfigForReconfig(ReplicationCoordinatorExternalState* externalState,
const ReplSetConfig& oldConfig,
const ReplSetConfig& newConfig,
ServiceContext* ctx,
bool force) {
Status status = newConfig.validate();
if (!status.isOK()) {
return StatusWith<int>(status);
}
status = newConfig.checkIfWriteConcernCanBeSatisfied(newConfig.getDefaultWriteConcern());
if (!status.isOK()) {
return status.withContext(
"Found invalid default write concern in 'getLastErrorDefaults' field");
}
status = validateOldAndNewConfigsCompatible(oldConfig, newConfig);
if (!status.isOK()) {
return StatusWith<int>(status);
}
status = validateArbiterPriorities(newConfig);
if (!status.isOK()) {
return StatusWith<int>(status);
}
if (force) {
return findSelfInConfig(externalState, newConfig, ctx);
}
return findSelfInConfigIfElectable(externalState, newConfig, ctx);
}
void ReplSet::haveNewConfig(OperationContext* txn, ReplSetConfig& newConfig, bool addComment) {
bo comment;
if( addComment )
comment = BSON( "msg" << "Reconfig set" << "version" << newConfig.version );
newConfig.saveConfigLocally(txn, comment);
try {
BSONObj oldConfForAudit = config().asBson();
BSONObj newConfForAudit = newConfig.asBson();
audit::logReplSetReconfig(ClientBasic::getCurrent(),
&oldConfForAudit,
&newConfForAudit);
if (initFromConfig(txn, newConfig, true)) {
log() << "replSet replSetReconfig new config saved locally" << rsLog;
}
}
catch (const DBException& e) {
log() << "replSet error unexpected exception in haveNewConfig() : " << e.toString() << rsLog;
fassertFailedNoTrace(18755);
}
catch (...) {
std::terminate();
}
}
OplogFetcher::OplogFetcher(executor::TaskExecutor* executor,
OpTimeWithHash lastFetched,
HostAndPort source,
NamespaceString nss,
ReplSetConfig config,
std::size_t maxFetcherRestarts,
int requiredRBID,
bool requireFresherSyncSource,
DataReplicatorExternalState* dataReplicatorExternalState,
EnqueueDocumentsFn enqueueDocumentsFn,
OnShutdownCallbackFn onShutdownCallbackFn,
const int batchSize)
: AbstractOplogFetcher(executor,
lastFetched,
source,
nss,
maxFetcherRestarts,
onShutdownCallbackFn,
"oplog fetcher"),
_metadataObject(makeMetadataObject(config.getProtocolVersion() == 1LL)),
_requiredRBID(requiredRBID),
_requireFresherSyncSource(requireFresherSyncSource),
_dataReplicatorExternalState(dataReplicatorExternalState),
_enqueueDocumentsFn(enqueueDocumentsFn),
_awaitDataTimeout(calculateAwaitDataTimeout(config)),
_batchSize(batchSize) {
invariant(config.isInitialized());
invariant(enqueueDocumentsFn);
}
// Our own config must be the first one.
bool ReplSetImpl::_loadConfigFinish(vector<ReplSetConfig>& cfgs) {
int v = -1;
ReplSetConfig *highest = 0;
int myVersion = -2000;
int n = 0;
for( vector<ReplSetConfig>::iterator i = cfgs.begin(); i != cfgs.end(); i++ ) {
ReplSetConfig& cfg = *i;
if( ++n == 1 ) myVersion = cfg.version;
if( cfg.ok() && cfg.version > v ) {
highest = &cfg;
v = cfg.version;
}
}
assert( highest );
if( !initFromConfig(*highest) )
return false;
if( highest->version > myVersion && highest->version >= 0 ) {
log() << "replSet got config version " << highest->version << " from a remote, saving locally" << rsLog;
writelock lk("admin.");
highest->saveConfigLocally(BSONObj());
}
return true;
}
Status checkQuorumForReconfig(executor::TaskExecutor* executor,
const ReplSetConfig& rsConfig,
const int myIndex,
long long term) {
invariant(rsConfig.getConfigVersion() > 1);
return checkQuorumGeneral(executor, rsConfig, myIndex, term, "reconfig quorum check");
}
void ReplSet::haveNewConfig(ReplSetConfig& newConfig, bool addComment) {
bo comment;
if( addComment )
comment = BSON( "msg" << "Reconfig set" << "version" << newConfig.version );
newConfig.saveConfigLocally(comment);
try {
if (initFromConfig(newConfig, true)) {
log() << "replSet replSetReconfig new config saved locally" << rsLog;
}
}
catch(DBException& e) {
if( e.getCode() == 13497 /* removed from set */ ) {
cc().shutdown();
dbexit( EXIT_CLEAN , "removed from replica set" ); // never returns
assert(0);
}
log() << "replSet error unexpected exception in haveNewConfig() : " << e.toString() << rsLog;
_fatal();
}
catch(...) {
log() << "replSet error unexpected exception in haveNewConfig()" << rsLog;
_fatal();
}
}
Status checkQuorumForInitiate(executor::TaskExecutor* executor,
const ReplSetConfig& rsConfig,
const int myIndex,
long long term) {
invariant(rsConfig.getConfigVersion() == 1);
return checkQuorumGeneral(executor, rsConfig, myIndex, term);
}
StatusWith<int> validateConfigForStartUp(ReplicationCoordinatorExternalState* externalState,
const ReplSetConfig& newConfig,
ServiceContext* ctx) {
Status status = newConfig.validate();
if (!status.isOK()) {
return StatusWith<int>(status);
}
return findSelfInConfig(externalState, newConfig, ctx);
}
void ReplSet::haveNewConfig(ReplSetConfig& newConfig, bool addComment) {
lock l(this); // convention is to lock replset before taking the db rwlock
writelock lk("");
bo comment;
if( addComment )
comment = BSON( "msg" << "Reconfig set" << "version" << newConfig.version );
newConfig.saveConfigLocally(comment);
try {
initFromConfig(newConfig, true);
log() << "replSet replSetReconfig new config saved locally" << rsLog;
}
catch(DBException& e) {
log() << "replSet error unexpected exception in haveNewConfig() : " << e.toString() << rsLog;
_fatal();
}
catch(...) {
log() << "replSet error unexpected exception in haveNewConfig()" << rsLog;
_fatal();
}
}
void ReplSet::haveNewConfig(ReplSetConfig& newConfig, bool addComment) {
bo comment;
if( addComment )
comment = BSON( "msg" << "Reconfig set" << "version" << newConfig.version );
newConfig.saveConfigLocally(comment);
try {
if (initFromConfig(newConfig, true)) {
log() << "replSet replSetReconfig new config saved locally" << rsLog;
}
}
catch(DBException& e) {
log() << "replSet error unexpected exception in haveNewConfig() : " << e.toString() << rsLog;
_fatal();
}
catch(...) {
log() << "replSet error unexpected exception in haveNewConfig()" << rsLog;
_fatal();
}
}
// @param reconf true if this is a reconfiguration and not an initial load of the configuration.
// @return true if ok; throws if config really bad; false if config doesn't include self
bool ReplSetImpl::initFromConfig(OperationContext* txn, ReplSetConfig& c, bool reconf) {
// NOTE: haveNewConfig() writes the new config to disk before we get here. So
// we cannot error out at this point, except fatally. Check errors earlier.
lock lk(this);
if (!getLastErrorDefault.isEmpty() || !c.getLastErrorDefaults.isEmpty()) {
getLastErrorDefault = c.getLastErrorDefaults;
}
list<ReplSetConfig::MemberCfg*> newOnes;
// additive short-cuts the new config setup. If we are just adding a
// node/nodes and nothing else is changing, this is additive. If it's
// not a reconfig, we're not adding anything
bool additive = reconf;
bool updateConfigs = false;
{
unsigned nfound = 0;
int me = 0;
for (vector<ReplSetConfig::MemberCfg>::iterator i = c.members.begin();
i != c.members.end();
i++) {
ReplSetConfig::MemberCfg& m = *i;
if (isSelf(m.h)) {
me++;
}
if (reconf) {
const Member *old = findById(m._id);
if (old) {
nfound++;
verify((int) old->id() == m._id);
if (!old->config().isSameIgnoringTags(m)) {
additive = false;
}
if (!updateConfigs && old->config() != m) {
updateConfigs = true;
}
}
else {
newOnes.push_back(&m);
}
}
}
if (me == 0) { // we're not in the config -- we must have been removed
if (state().removed()) {
// already took note of our ejection from the set
// so just sit tight and poll again
return false;
}
_members.orphanAll();
// kill off rsHealthPoll threads (because they Know Too Much about our past)
endOldHealthTasks();
// clear sync target to avoid faulty sync attempts; we must do this before we
// close sockets, since that will trigger the bgsync thread to reconnect.
BackgroundSync::get()->clearSyncTarget();
// close sockets to force clients to re-evaluate this member
MessagingPort::closeAllSockets(0);
// take note of our ejection
changeState(MemberState::RS_REMOVED);
// go into holding pattern
log() << "replSet info self not present in the repl set configuration:" << rsLog;
log() << c.toString() << rsLog;
loadConfig(txn); // redo config from scratch
return false;
}
uassert(13302, "replSet error self appears twice in the repl set configuration", me<=1);
if (state().removed()) {
// If we were removed and have now been added back in, switch state.
changeState(MemberState::RS_RECOVERING);
}
// if we found different members that the original config, reload everything
if (reconf && config().members.size() != nfound)
additive = false;
}
// If we are changing chaining rules, we don't want this to be an additive reconfig so that
// the primary can step down and the sync targets change.
// TODO: This can be removed once SERVER-5208 is fixed.
if (reconf && config().chainingAllowed() != c.chainingAllowed()) {
additive = false;
}
_cfg = new ReplSetConfig(c);
// config() is same thing but const, so we use that when we can for clarity below
dassert(&config() == _cfg);
verify(config().ok());
verify(_name.empty() || _name == config()._id);
_name = config()._id;
verify(!_name.empty());
// this is a shortcut for simple changes
if (additive) {
log() << "replSet info : additive change to configuration" << rsLog;
if (updateConfigs) {
// we have new configs for existing members, so we need to repopulate _members
// with the most recent configs
_members.orphanAll();
// for logging
string members = "";
// not setting _self to 0 as other threads use _self w/o locking
int me = 0;
for(vector<ReplSetConfig::MemberCfg>::const_iterator i = config().members.begin();
i != config().members.end(); i++) {
const ReplSetConfig::MemberCfg& m = *i;
Member *mi;
members += (members == "" ? "" : ", ") + m.h.toString();
if (isSelf(m.h)) {
verify(me++ == 0);
mi = new Member(m.h, m._id, &m, true);
setSelfTo(mi);
}
else {
mi = new Member(m.h, m._id, &m, false);
_members.push(mi);
}
}
// trigger a handshake to update the syncSource of our writeconcern information
syncSourceFeedback.forwardSlaveHandshake();
}
// add any new members
for (list<ReplSetConfig::MemberCfg*>::const_iterator i = newOnes.begin();
i != newOnes.end();
i++) {
ReplSetConfig::MemberCfg *m = *i;
Member *mi = new Member(m->h, m->_id, m, false);
// we will indicate that new members are up() initially so that we don't relinquish
// our primary state because we can't (transiently) see a majority. they should be
// up as we check that new members are up before getting here on reconfig anyway.
mi->get_hbinfo().health = 0.1;
_members.push(mi);
startHealthTaskFor(mi);
}
// if we aren't creating new members, we may have to update the
// groups for the current ones
_cfg->updateMembers(_members);
return true;
}
// start with no members. if this is a reconfig, drop the old ones.
_members.orphanAll();
endOldHealthTasks();
int oldPrimaryId = -1;
{
const Member *p = box.getPrimary();
if (p)
oldPrimaryId = p->id();
}
forgetPrimary(txn);
// not setting _self to 0 as other threads use _self w/o locking
int me = 0;
// For logging
string members = "";
for (vector<ReplSetConfig::MemberCfg>::const_iterator i = config().members.begin();
i != config().members.end();
i++) {
const ReplSetConfig::MemberCfg& m = *i;
Member *mi;
members += (members == "" ? "" : ", ") + m.h.toString();
if (isSelf(m.h)) {
verify(me++ == 0);
mi = new Member(m.h, m._id, &m, true);
if (!reconf) {
log() << "replSet I am " << m.h.toString() << rsLog;
}
setSelfTo(mi);
if ((int)mi->id() == oldPrimaryId)
box.setSelfPrimary(mi);
}
else {
mi = new Member(m.h, m._id, &m, false);
_members.push(mi);
if ((int)mi->id() == oldPrimaryId)
box.setOtherPrimary(mi);
}
}
if (me == 0){
log() << "replSet warning did not detect own host in full reconfig, members "
<< members << " config: " << c << rsLog;
}
else {
// Do this after we've found ourselves, since _self needs
// to be set before we can start the heartbeat tasks
for (Member *mb = _members.head(); mb; mb=mb->next()) {
startHealthTaskFor(mb);
}
}
return true;
}
/** @param reconf true if this is a reconfiguration and not an initial load of the configuration.
@return true if ok; throws if config really bad; false if config doesn't include self
*/
bool ReplSetImpl::initFromConfig(ReplSetConfig& c, bool reconf) {
/* NOTE: haveNewConfig() writes the new config to disk before we get here. So
we cannot error out at this point, except fatally. Check errors earlier.
*/
lock lk(this);
if( getLastErrorDefault || !c.getLastErrorDefaults.isEmpty() ) {
// see comment in dbcommands.cpp for getlasterrordefault
getLastErrorDefault = new BSONObj( c.getLastErrorDefaults );
}
list<const ReplSetConfig::MemberCfg*> newOnes;
bool additive = reconf;
{
unsigned nfound = 0;
int me = 0;
for( vector<ReplSetConfig::MemberCfg>::iterator i = c.members.begin(); i != c.members.end(); i++ ) {
const ReplSetConfig::MemberCfg& m = *i;
if( m.h.isSelf() ) {
nfound++;
me++;
if( !reconf || (_self && _self->id() == (unsigned) m._id) )
;
else {
log() << "replSet " << _self->id() << ' ' << m._id << rsLog;
assert(false);
}
}
else if( reconf ) {
const Member *old = findById(m._id);
if( old ) {
nfound++;
assert( (int) old->id() == m._id );
if( old->config() == m ) {
additive = false;
}
}
else {
newOnes.push_back(&m);
}
}
}
if( me == 0 ) {
// log() << "replSet config : " << _cfg->toString() << rsLog;
log() << "replSet error self not present in the repl set configuration:" << rsLog;
log() << c.toString() << rsLog;
uasserted(13497, "replSet error self not present in the configuration");
}
uassert( 13302, "replSet error self appears twice in the repl set configuration", me<=1 );
if( reconf && config().members.size() != nfound )
additive = false;
}
_cfg = new ReplSetConfig(c);
assert( _cfg->ok() );
assert( _name.empty() || _name == _cfg->_id );
_name = _cfg->_id;
assert( !_name.empty() );
if( additive ) {
log() << "replSet info : additive change to configuration" << rsLog;
for( list<const ReplSetConfig::MemberCfg*>::const_iterator i = newOnes.begin(); i != newOnes.end(); i++ ) {
const ReplSetConfig::MemberCfg* m = *i;
Member *mi = new Member(m->h, m->_id, m, false);
/** we will indicate that new members are up() initially so that we don't relinquish our
primary state because we can't (transiently) see a majority. they should be up as we
check that new members are up before getting here on reconfig anyway.
*/
mi->get_hbinfo().health = 0.1;
_members.push(mi);
startHealthTaskFor(mi);
}
return true;
}
// start with no members. if this is a reconfig, drop the old ones.
_members.orphanAll();
endOldHealthTasks();
int oldPrimaryId = -1;
{
const Member *p = box.getPrimary();
if( p )
oldPrimaryId = p->id();
}
forgetPrimary();
setSelfTo(0);
for( vector<ReplSetConfig::MemberCfg>::iterator i = _cfg->members.begin(); i != _cfg->members.end(); i++ ) {
const ReplSetConfig::MemberCfg& m = *i;
Member *mi;
if( m.h.isSelf() ) {
assert( _self == 0 );
mi = new Member(m.h, m._id, &m, true);
setSelfTo(mi);
if( (int)mi->id() == oldPrimaryId )
box.setSelfPrimary(mi);
} else {
mi = new Member(m.h, m._id, &m, false);
_members.push(mi);
startHealthTaskFor(mi);
if( (int)mi->id() == oldPrimaryId )
box.setOtherPrimary(mi);
}
}
return true;
}
/** @param reconf true if this is a reconfiguration and not an initial load of the configuration.
@return true if ok; throws if config really bad; false if config doesn't include self
*/
bool ReplSetImpl::initFromConfig(ReplSetConfig& c, bool reconf) {
/* NOTE: haveNewConfig() writes the new config to disk before we get here. So
we cannot error out at this point, except fatally. Check errors earlier.
*/
lock lk(this);
if( getLastErrorDefault || !c.getLastErrorDefaults.isEmpty() ) {
// see comment in dbcommands.cpp for getlasterrordefault
getLastErrorDefault = new BSONObj( c.getLastErrorDefaults );
}
list<const ReplSetConfig::MemberCfg*> newOnes;
// additive short-cuts the new config setup. If we are just adding a
// node/nodes and nothing else is changing, this is additive. If it's
// not a reconfig, we're not adding anything
bool additive = reconf;
{
unsigned nfound = 0;
int me = 0;
for( vector<ReplSetConfig::MemberCfg>::iterator i = c.members.begin(); i != c.members.end(); i++ ) {
const ReplSetConfig::MemberCfg& m = *i;
if( m.h.isSelf() ) {
me++;
}
if( reconf ) {
if (m.h.isSelf() && (!_self || (int)_self->id() != m._id)) {
log() << "self doesn't match: " << m._id << rsLog;
assert(false);
}
const Member *old = findById(m._id);
if( old ) {
nfound++;
assert( (int) old->id() == m._id );
if( old->config() != m ) {
additive = false;
}
}
else {
newOnes.push_back(&m);
}
}
}
if( me == 0 ) {
_members.orphanAll();
// hbs must continue to pick up new config
// stop sync thread
box.set(MemberState::RS_STARTUP, 0);
// go into holding pattern
log() << "replSet error self not present in the repl set configuration:" << rsLog;
log() << c.toString() << rsLog;
return false;
}
uassert( 13302, "replSet error self appears twice in the repl set configuration", me<=1 );
// if we found different members that the original config, reload everything
if( reconf && config().members.size() != nfound )
additive = false;
}
_cfg = new ReplSetConfig(c);
assert( _cfg->ok() );
assert( _name.empty() || _name == _cfg->_id );
_name = _cfg->_id;
assert( !_name.empty() );
// this is a shortcut for simple changes
if( additive ) {
log() << "replSet info : additive change to configuration" << rsLog;
for( list<const ReplSetConfig::MemberCfg*>::const_iterator i = newOnes.begin(); i != newOnes.end(); i++ ) {
const ReplSetConfig::MemberCfg* m = *i;
Member *mi = new Member(m->h, m->_id, m, false);
/** we will indicate that new members are up() initially so that we don't relinquish our
primary state because we can't (transiently) see a majority. they should be up as we
check that new members are up before getting here on reconfig anyway.
*/
mi->get_hbinfo().health = 0.1;
_members.push(mi);
startHealthTaskFor(mi);
}
return true;
}
// start with no members. if this is a reconfig, drop the old ones.
_members.orphanAll();
endOldHealthTasks();
int oldPrimaryId = -1;
{
const Member *p = box.getPrimary();
if( p )
oldPrimaryId = p->id();
}
forgetPrimary();
// not setting _self to 0 as other threads use _self w/o locking
int me = 0;
// For logging
string members = "";
for( vector<ReplSetConfig::MemberCfg>::iterator i = _cfg->members.begin(); i != _cfg->members.end(); i++ ) {
const ReplSetConfig::MemberCfg& m = *i;
Member *mi;
members += ( members == "" ? "" : ", " ) + m.h.toString();
if( m.h.isSelf() ) {
assert( me++ == 0 );
mi = new Member(m.h, m._id, &m, true);
setSelfTo(mi);
if( (int)mi->id() == oldPrimaryId )
box.setSelfPrimary(mi);
}
else {
mi = new Member(m.h, m._id, &m, false);
_members.push(mi);
startHealthTaskFor(mi);
if( (int)mi->id() == oldPrimaryId )
box.setOtherPrimary(mi);
}
}
if( me == 0 ) {
log() << "replSet warning did not detect own host in full reconfig, members " << members << " config: " << c << rsLog;
}
return true;
}