ReplSetConfig::ReplSetConfig(const HostAndPort& h) {
clear();
int level = 2;
DEV level = 0;
BSONObj cfg;
int v = -5;
try {
if( h.isSelf() ) {
;
}
else {
/* first, make sure other node is configured to be a replset. just to be safe. */
string setname = cmdLine.ourSetName();
BSONObj cmd = BSON( "replSetHeartbeat" << setname );
int theirVersion;
BSONObj info;
log() << "trying to contact " << h.toString() << rsLog;
bool ok = requestHeartbeat(setname, "", h.toString(), info, -2, theirVersion);
if( info["rs"].trueValue() ) {
// yes, it is a replicate set, although perhaps not yet initialized
}
else {
if( !ok ) {
log() << "replSet TEMP !ok heartbeating " << h.toString() << " on cfg load" << rsLog;
if( !info.isEmpty() )
log() << "replSet info " << h.toString() << " : " << info.toString() << rsLog;
return;
}
{
stringstream ss;
ss << "replSet error: member " << h.toString() << " is not in --replSet mode";
msgassertedNoTrace(13260, ss.str().c_str()); // not caught as not a user exception - we want it not caught
//for python err# checker: uassert(13260, "", false);
}
}
}
v = -4;
unsigned long long count = 0;
try {
ScopedConn conn(h.toString());
v = -3;
cfg = conn.findOne(rsConfigNs, Query()).getOwned();
count = conn.count(rsConfigNs);
}
catch ( DBException& ) {
if ( !h.isSelf() ) {
throw;
}
// on startup, socket is not listening yet
DBDirectClient cli;
cfg = cli.findOne( rsConfigNs, Query() ).getOwned();
count = cli.count(rsConfigNs);
}
if( count > 1 )
uasserted(13109, str::stream() << "multiple rows in " << rsConfigNs << " not supported host: " << h.toString());
if( cfg.isEmpty() ) {
version = EMPTYCONFIG;
return;
}
version = -1;
}
catch( DBException& e) {
version = v;
log(level) << "replSet load config couldn't get from " << h.toString() << ' ' << e.what() << rsLog;
return;
}
from(cfg);
checkRsConfig();
_ok = true;
log(level) << "replSet load config ok from " << (h.isSelf() ? "self" : h.toString()) << rsLog;
}
/* called on a reconfig AND on initiate
throws
@param initial true when initiating
*/
void checkMembersUpForConfigChange(const ReplSetConfig& cfg, bool initial) {
int failures = 0, majority = 0;
int me = 0;
stringstream selfs;
for( vector<ReplSetConfig::MemberCfg>::const_iterator i = cfg.members.begin(); i != cfg.members.end(); i++ ) {
if( i->h.isSelf() ) {
me++;
if( me > 1 )
selfs << ',';
selfs << i->h.toString();
if( !i->potentiallyHot() ) {
uasserted(13420, "initiation and reconfiguration of a replica set must be sent to a node that can become primary");
}
majority += i->votes;
}
}
majority = (majority / 2) + 1;
uassert(13278, "bad config: isSelf is true for multiple hosts: " + selfs.str(), me <= 1); // dups?
if( me != 1 ) {
stringstream ss;
ss << "can't find self in the replset config";
if( !cmdLine.isDefaultPort() ) ss << " my port: " << cmdLine.port;
if( me != 0 ) ss << " found: " << me;
uasserted(13279, ss.str());
}
for( vector<ReplSetConfig::MemberCfg>::const_iterator i = cfg.members.begin(); i != cfg.members.end(); i++ ) {
// we know we're up
if (i->h.isSelf()) {
continue;
}
BSONObj res;
{
bool ok = false;
try {
int theirVersion = -1000;
ok = requestHeartbeat(cfg._id, "", i->h.toString(), res, -1, theirVersion, initial/*check if empty*/);
if( theirVersion >= cfg.version ) {
stringstream ss;
ss << "replSet member " << i->h.toString() << " has too new a config version (" << theirVersion << ") to reconfigure";
uasserted(13259, ss.str());
}
}
catch(DBException& e) {
log() << "replSet cmufcc requestHeartbeat " << i->h.toString() << " : " << e.toString() << rsLog;
}
catch(...) {
log() << "replSet cmufcc error exception in requestHeartbeat?" << rsLog;
}
if( res.getBoolField("mismatch") )
uasserted(13145, "set name does not match the set name host " + i->h.toString() + " expects");
if( *res.getStringField("set") ) {
if( cfg.version <= 1 ) {
// this was to be initiation, no one shoudl be initiated already.
uasserted(13256, "member " + i->h.toString() + " is already initiated");
}
else {
// Assure no one has a newer config.
if( res["v"].Int() >= cfg.version ) {
uasserted(13341, "member " + i->h.toString() + " has a config version >= to the new cfg version; cannot change config");
}
}
}
if( !ok && !res["rs"].trueValue() ) {
if( !res.isEmpty() ) {
/* strange. got a response, but not "ok". log it. */
log() << "replSet warning " << i->h.toString() << " replied: " << res.toString() << rsLog;
}
bool allowFailure = false;
failures += i->votes;
if( res.isEmpty() && !initial && failures >= majority ) {
const Member* m = theReplSet->findById( i->_id );
if( m ) {
assert( m->h().toString() == i->h.toString() );
}
// it's okay if the down member isn't part of the config,
// we might be adding a new member that isn't up yet
allowFailure = true;
}
if( !allowFailure ) {
string msg = string("need all members up to initiate, not ok : ") + i->h.toString();
if( !initial )
msg = string("need most members up to reconfigure, not ok : ") + i->h.toString();
uasserted(13144, msg);
}
}
}
if( initial ) {
bool hasData = res["hasData"].Bool();
uassert(13311, "member " + i->h.toString() + " has data already, cannot initiate set. All members except initiator must be empty.",
!hasData || i->h.isSelf());
}
}
}
/* called on a reconfig AND on initiate
throws
@param initial true when initiating
*/
void checkMembersUpForConfigChange(const ReplSetConfig& cfg, bool initial) {
int failures = 0;
int me = 0;
for( vector<ReplSetConfig::MemberCfg>::const_iterator i = cfg.members.begin(); i != cfg.members.end(); i++ ) {
if( i->h.isSelf() ) {
me++;
if( !i->potentiallyHot() ) {
uasserted(13420, "initiation and reconfiguration of a replica set must be sent to a node that can become primary");
}
}
}
uassert(13278, "bad config - dups?", me <= 1); // dups?
uassert(13279, "can't find self in the replset config", me == 1);
for( vector<ReplSetConfig::MemberCfg>::const_iterator i = cfg.members.begin(); i != cfg.members.end(); i++ ) {
BSONObj res;
{
bool ok = false;
try {
int theirVersion = -1000;
ok = requestHeartbeat(cfg._id, "", i->h.toString(), res, -1, theirVersion, initial/*check if empty*/);
if( theirVersion >= cfg.version ) {
stringstream ss;
ss << "replSet member " << i->h.toString() << " has too new a config version (" << theirVersion << ") to reconfigure";
uasserted(13259, ss.str());
}
}
catch(DBException& e) {
log() << "replSet cmufcc requestHeartbeat " << i->h.toString() << " : " << e.toString() << rsLog;
}
catch(...) {
log() << "replSet cmufcc error exception in requestHeartbeat?" << rsLog;
}
if( res.getBoolField("mismatch") )
uasserted(13145, "set name does not match the set name host " + i->h.toString() + " expects");
if( *res.getStringField("set") ) {
if( cfg.version <= 1 ) {
// this was to be initiation, no one shoudl be initiated already.
uasserted(13256, "member " + i->h.toString() + " is already initiated");
}
else {
// Assure no one has a newer config.
if( res["v"].Int() >= cfg.version ) {
uasserted(13341, "member " + i->h.toString() + " has a config version >= to the new cfg version; cannot change config");
}
}
}
if( !ok && !res["rs"].trueValue() ) {
if( !res.isEmpty() ) {
/* strange. got a response, but not "ok". log it. */
log() << "replSet warning " << i->h.toString() << " replied: " << res.toString() << rsLog;
}
bool allowFailure = false;
failures++;
if( res.isEmpty() && !initial && failures == 1 ) {
/* for now we are only allowing 1 node to be down on a reconfig. this can be made to be a minority
trying to keep change small as release is near.
*/
const Member* m = theReplSet->findById( i->_id );
if( m ) {
// ok, so this was an existing member (wouldn't make sense to add to config a new member that is down)
assert( m->h().toString() == i->h.toString() );
allowFailure = true;
}
}
if( !allowFailure ) {
string msg = string("need members up to initiate, not ok : ") + i->h.toString();
if( !initial )
msg = string("need most members up to reconfigure, not ok : ") + i->h.toString();
uasserted(13144, msg);
}
}
}
if( initial ) {
bool hasData = res["hasData"].Bool();
uassert(13311, "member " + i->h.toString() + " has data already, cannot initiate set. All members except initiator must be empty.",
!hasData || i->h.isSelf());
}
}
}