void MongoCluster::loadIndex(NsTagUBIndex* index, const std::string& queryNs, MappingType* linkmap,
const std::string& mappingName, const std::string group,
const mongo::BSONObj& key) {
//The indexes held type
using index_mapped_type = typename NsTagUBIndex::mapped_type;
index_mapped_type* idx = nullptr;
mongo::Cursor cur = _dbConn->query(queryNs, mongo::Query().sort(BSON(group << 1)));
std::string prevNs;
while (cur->more()) {
mongo::BSONObj obj = cur->next();
std::string mappingValue = obj.getStringField(mappingName);
assert(!mappingValue.empty());
std::string ns = obj.getStringField(group);
assert(!ns.empty());
if (ns != prevNs) {
if (idx) idx->finalize();
prevNs = ns;
idx = &index->emplace(ns, index_mapped_type(tools::BSONObjCmp(key))).first->second;
}
idx->insertUnordered(obj.getField("max").Obj().getOwned(),
TagRange(linkmap->find(mappingValue),
obj.getField("max").Obj().getOwned(),
obj.getField("min").Obj().getOwned()));
}
//Sort chunks here.
if (idx) idx->finalize();
}
void Balancer::_doBalanceRound( DBClientBase& conn, vector<CandidateChunkPtr>* candidateChunks ) {
verify( candidateChunks );
//
// 1. Check whether there is any sharded collection to be balanced by querying
// the ShardsNS::collections collection
//
auto_ptr<DBClientCursor> cursor = conn.query( ShardNS::collection , BSONObj() );
vector< string > collections;
while ( cursor->more() ) {
BSONObj col = cursor->nextSafe();
// sharded collections will have a shard "key".
if ( ! col["key"].eoo() && ! col["noBalance"].trueValue() ){
collections.push_back( col["_id"].String() );
}
else if( col["noBalance"].trueValue() ){
LOG(1) << "not balancing collection " << col["_id"].String() << ", explicitly disabled" << endl;
}
}
cursor.reset();
if ( collections.empty() ) {
LOG(1) << "no collections to balance" << endl;
return;
}
//
// 2. Get a list of all the shards that are participating in this balance round
// along with any maximum allowed quotas and current utilization. We get the
// latter by issuing db.serverStatus() (mem.mapped) to all shards.
//
// TODO: skip unresponsive shards and mark information as stale.
//
vector<Shard> allShards;
Shard::getAllShards( allShards );
if ( allShards.size() < 2) {
LOG(1) << "can't balance without more active shards" << endl;
return;
}
ShardInfoMap shardInfo;
for ( vector<Shard>::const_iterator it = allShards.begin(); it != allShards.end(); ++it ) {
const Shard& s = *it;
ShardStatus status = s.getStatus();
shardInfo[ s.getName() ] = ShardInfo( s.getMaxSize(),
status.mapped(),
s.isDraining(),
status.hasOpsQueued(),
s.tags()
);
}
//
// 3. For each collection, check if the balancing policy recommends moving anything around.
//
for (vector<string>::const_iterator it = collections.begin(); it != collections.end(); ++it ) {
const string& ns = *it;
map< string,vector<BSONObj> > shardToChunksMap;
cursor = conn.query( ShardNS::chunk , QUERY( "ns" << ns ).sort( "min" ) );
while ( cursor->more() ) {
BSONObj chunk = cursor->nextSafe();
if ( chunk["jumbo"].trueValue() )
continue;
vector<BSONObj>& chunks = shardToChunksMap[chunk["shard"].String()];
chunks.push_back( chunk.getOwned() );
}
cursor.reset();
if (shardToChunksMap.empty()) {
LOG(1) << "skipping empty collection (" << ns << ")";
continue;
}
for ( vector<Shard>::iterator i=allShards.begin(); i!=allShards.end(); ++i ) {
// this just makes sure there is an entry in shardToChunksMap for every shard
Shard s = *i;
shardToChunksMap[s.getName()].size();
}
DistributionStatus status( shardInfo, shardToChunksMap );
// load tags
conn.ensureIndex( ShardNS::tags, BSON( "ns" << 1 << "min" << 1 ), true );
cursor = conn.query( ShardNS::tags , QUERY( "ns" << ns ).sort( "min" ) );
while ( cursor->more() ) {
BSONObj tag = cursor->nextSafe();
uassert( 16356 , str::stream() << "tag ranges not valid for: " << ns ,
status.addTagRange( TagRange( tag["min"].Obj().getOwned(),
tag["max"].Obj().getOwned(),
tag["tag"].String() ) ) );
}
cursor.reset();
CandidateChunk* p = _policy->balance( ns, status, _balancedLastTime );
if ( p ) candidateChunks->push_back( CandidateChunkPtr( p ) );
}
}
StatusWith<vector<MigrateInfo>> Balancer::_getCandidateChunks(OperationContext* txn) {
vector<CollectionType> collections;
Status collsStatus =
grid.catalogManager(txn)->getCollections(txn, nullptr, &collections, nullptr);
if (!collsStatus.isOK()) {
return collsStatus;
}
if (collections.empty()) {
return vector<MigrateInfo>();
}
// Get a list of all the shards that are participating in this balance round along with any
// maximum allowed quotas and current utilization. We get the latter by issuing
// db.serverStatus() (mem.mapped) to all shards.
//
// TODO: skip unresponsive shards and mark information as stale.
auto shardInfoStatus = DistributionStatus::populateShardInfoMap(txn);
if (!shardInfoStatus.isOK()) {
return shardInfoStatus.getStatus();
}
const ShardInfoMap shardInfo(std::move(shardInfoStatus.getValue()));
if (shardInfo.size() < 2) {
return vector<MigrateInfo>();
}
OCCASIONALLY warnOnMultiVersion(shardInfo);
std::vector<MigrateInfo> candidateChunks;
// For each collection, check if the balancing policy recommends moving anything around.
for (const auto& coll : collections) {
// Skip collections for which balancing is disabled
const NamespaceString& nss = coll.getNs();
if (!coll.getAllowBalance()) {
LOG(1) << "Not balancing collection " << nss << "; explicitly disabled.";
continue;
}
std::vector<ChunkType> allNsChunks;
Status status = grid.catalogManager(txn)->getChunks(txn,
BSON(ChunkType::ns(nss.ns())),
BSON(ChunkType::min() << 1),
boost::none, // all chunks
&allNsChunks,
nullptr);
if (!status.isOK()) {
warning() << "failed to load chunks for ns " << nss.ns() << causedBy(status);
continue;
}
set<BSONObj> allChunkMinimums;
map<string, vector<ChunkType>> shardToChunksMap;
for (const ChunkType& chunk : allNsChunks) {
allChunkMinimums.insert(chunk.getMin().getOwned());
vector<ChunkType>& chunksList = shardToChunksMap[chunk.getShard()];
chunksList.push_back(chunk);
}
if (shardToChunksMap.empty()) {
LOG(1) << "skipping empty collection (" << nss.ns() << ")";
continue;
}
for (ShardInfoMap::const_iterator i = shardInfo.begin(); i != shardInfo.end(); ++i) {
// This loop just makes sure there is an entry in shardToChunksMap for every shard
shardToChunksMap[i->first];
}
DistributionStatus distStatus(shardInfo, shardToChunksMap);
// TODO: TagRange contains all the information from TagsType except for the namespace,
// so maybe the two can be merged at some point in order to avoid the
// transformation below.
vector<TagRange> ranges;
{
vector<TagsType> collectionTags;
uassertStatusOK(
grid.catalogManager(txn)->getTagsForCollection(txn, nss.ns(), &collectionTags));
for (const auto& tt : collectionTags) {
ranges.push_back(
TagRange(tt.getMinKey().getOwned(), tt.getMaxKey().getOwned(), tt.getTag()));
uassert(16356,
str::stream() << "tag ranges not valid for: " << nss.ns(),
distStatus.addTagRange(ranges.back()));
}
}
auto statusGetDb = grid.catalogCache()->getDatabase(txn, nss.db().toString());
if (!statusGetDb.isOK()) {
warning() << "could not load db config to balance collection [" << nss.ns()
<< "]: " << statusGetDb.getStatus();
continue;
}
shared_ptr<DBConfig> cfg = statusGetDb.getValue();
// This line reloads the chunk manager once if this process doesn't know the collection
// is sharded yet.
shared_ptr<ChunkManager> cm = cfg->getChunkManagerIfExists(txn, nss.ns(), true);
if (!cm) {
warning() << "could not load chunks to balance " << nss.ns() << " collection";
continue;
}
// Loop through tags to make sure no chunk spans tags. Split on tag min for all chunks.
bool didAnySplits = false;
for (const TagRange& range : ranges) {
BSONObj min =
cm->getShardKeyPattern().getKeyPattern().extendRangeBound(range.min, false);
if (allChunkMinimums.count(min) > 0) {
continue;
}
didAnySplits = true;
log() << "nss: " << nss.ns() << " need to split on " << min
<< " because there is a range there";
vector<BSONObj> splitPoints;
splitPoints.push_back(min);
shared_ptr<Chunk> c = cm->findIntersectingChunk(txn, min);
Status status = c->multiSplit(txn, splitPoints, NULL);
if (!status.isOK()) {
error() << "split failed: " << status;
} else {
LOG(1) << "split worked";
}
break;
}
if (didAnySplits) {
// State change, just wait till next round
continue;
}
shared_ptr<MigrateInfo> migrateInfo(
BalancerPolicy::balance(nss.ns(), distStatus, _balancedLastTime));
if (migrateInfo) {
candidateChunks.emplace_back(*migrateInfo);
}
}
return candidateChunks;
}
