private void ApplyPrecomputedBatchForNewIndex(Index index, AbstractViewGenerator generator)
{
const string DocumentsByEntityNameIndex = "Raven/DocumentsByEntityName";
PrecomputedIndexingBatch result = null;
var docsToIndex = new List<JsonDocument>();
TransactionalStorage.Batch(actions =>
{
var tags = generator.ForEntityNames.Select(entityName => "Tag:[[" + entityName + "]]").ToList();
var query = string.Join(" OR ", tags);
JsonDocument highestByEtag = null;
var cts = new CancellationTokenSource();
using (var linked = CancellationTokenSource.CreateLinkedTokenSource(cts.Token, WorkContext.CancellationToken))
using (var op = new QueryActions.DatabaseQueryOperation(Database, DocumentsByEntityNameIndex, new IndexQuery
{
Query = query,
PageSize = Database.Configuration.MaxNumberOfItemsToProcessInSingleBatch
}, actions, linked)
{
ShouldSkipDuplicateChecking = true
})
{
op.Init();
if (op.Header.TotalResults == 0 ||
(op.Header.TotalResults > Database.Configuration.MaxNumberOfItemsToProcessInSingleBatch))
{
// we don't apply this optimization if the total number of results
// to index is more than the max numbers to index in a single batch.
// The idea here is that we need to keep the amount
// of memory we use to a manageable level even when introducing a new index to a BIG
// database
try
{
cts.Cancel();
// we have to run just a little bit of the query to properly setup the disposal
op.Execute(o => { });
}
catch (OperationCanceledException)
{
}
return;
}
Log.Debug("For new index {0}, using precomputed indexing batch optimization for {1} docs", index,
op.Header.TotalResults);
op.Execute(document =>
{
var metadata = document.Value<RavenJObject>(Constants.Metadata);
var key = metadata.Value<string>("@id");
var etag = Etag.Parse(metadata.Value<string>("@etag"));
var lastModified = DateTime.Parse(metadata.Value<string>(Constants.LastModified));
document.Remove(Constants.Metadata);
var doc = new JsonDocument
{
DataAsJson = document,
Etag = etag,
Key = key,
LastModified = lastModified,
SkipDeleteFromIndex = true,
Metadata = metadata
};
docsToIndex.Add(doc);
if (highestByEtag == null || doc.Etag.CompareTo(highestByEtag.Etag) > 0)
highestByEtag = doc;
});
}
result = new PrecomputedIndexingBatch
{
LastIndexed = highestByEtag.Etag,
LastModified = highestByEtag.LastModified.Value,
Documents = docsToIndex,
Index = index
};
});
if (result != null && result.Documents != null && result.Documents.Count > 0)
Database.IndexingExecuter.IndexPrecomputedBatch(result);
}
private void ApplyPrecomputedBatchForNewIndex(Index index, AbstractViewGenerator generator, int pageSize, CancellationTokenSource cts)
{
PrecomputedIndexingBatch result = null;
var docsToIndex = new List <JsonDocument>();
TransactionalStorage.Batch(actions =>
{
var query = GetQueryForAllMatchingDocumentsForIndex(generator);
JsonDocument highestByEtag = null;
using (var linked = CancellationTokenSource.CreateLinkedTokenSource(cts.Token, WorkContext.CancellationToken))
using (var op = new QueryActions.DatabaseQueryOperation(Database, Constants.DocumentsByEntityNameIndex, new IndexQuery
{
Query = query,
PageSize = pageSize
}, actions, linked)
{
ShouldSkipDuplicateChecking = true
})
{
op.Init();
if (op.Header.TotalResults == 0 ||
(op.Header.TotalResults > Database.Configuration.MaxNumberOfItemsToProcessInSingleBatch))
{
// we don't apply this optimization if the total number of results
// to index is more than the max numbers to index in a single batch.
// The idea here is that we need to keep the amount
// of memory we use to a manageable level even when introducing a new index to a BIG
// database
try
{
cts.Cancel();
// we have to run just a little bit of the query to properly setup the disposal
op.Execute(o => { });
}
catch (OperationCanceledException)
{
}
return;
}
Log.Debug("For new index {0}, using precomputed indexing batch optimization for {1} docs", index,
op.Header.TotalResults);
op.Execute(document =>
{
var metadata = document.Value <RavenJObject>(Constants.Metadata);
var key = metadata.Value <string>("@id");
var etag = Etag.Parse(metadata.Value <string>("@etag"));
var lastModified = DateTime.Parse(metadata.Value <string>(Constants.LastModified));
document.Remove(Constants.Metadata);
var doc = new JsonDocument
{
DataAsJson = document,
Etag = etag,
Key = key,
LastModified = lastModified,
SkipDeleteFromIndex = true,
Metadata = metadata
};
docsToIndex.Add(doc);
if (highestByEtag == null || doc.Etag.CompareTo(highestByEtag.Etag) > 0)
{
highestByEtag = doc;
}
});
}
result = new PrecomputedIndexingBatch
{
LastIndexed = highestByEtag.Etag,
LastModified = highestByEtag.LastModified.Value,
Documents = docsToIndex,
Index = index
};
});
if (result != null && result.Documents != null && result.Documents.Count > 0)
{
using (var linked = CancellationTokenSource.CreateLinkedTokenSource(cts.Token, WorkContext.CancellationToken))
{
Database.IndexingExecuter.IndexPrecomputedBatch(result, linked.Token);
if (index.IsTestIndex)
{
TransactionalStorage.Batch(accessor => accessor.Indexing.TouchIndexEtag(index.IndexId));
}
}
}
}
private void ApplyPrecomputedBatchForNewIndex(Index index, AbstractViewGenerator generator, int pageSize, CancellationTokenSource cts)
{
PrecomputedIndexingBatch result = null;
var docsToIndex = new List <JsonDocument>();
TransactionalStorage.Batch(actions =>
{
var query = QueryBuilder.GetQueryForAllMatchingDocumentsForIndex(Database, generator.ForEntityNames);
using (DocumentCacher.SkipSetDocumentsInDocumentCache())
using (var linked = CancellationTokenSource.CreateLinkedTokenSource(cts.Token, WorkContext.CancellationToken))
using (var op = new QueryActions.DatabaseQueryOperation(Database, Constants.DocumentsByEntityNameIndex, new IndexQuery
{
Query = query,
PageSize = pageSize
}, actions, linked)
{
ShouldSkipDuplicateChecking = true
})
{
op.Init();
//if we are working on a test index, apply the optimization anyway, as the index is capped by small number of results
if (op.Header.TotalResults > pageSize && index.IsTestIndex == false)
{
// we don't apply this optimization if the total number of results
// to index is more than the max numbers to index in a single batch.
// The idea here is that we need to keep the amount
// of memory we use to a manageable level even when introducing a new index to a BIG
// database
try
{
cts.Cancel();
// we have to run just a little bit of the query to properly setup the disposal
op.Execute(o => { });
}
catch (OperationCanceledException)
{
}
return;
}
if (Log.IsDebugEnabled)
{
Log.Debug("For new index {0}, using precomputed indexing batch optimization for {1} docs",
index, op.Header.TotalResults);
}
var totalLoadedDocumentSize = 0;
const int totalSizeToCheck = 16 * 1024 * 1024; //16MB
var localLoadedDocumentSize = 0;
op.Execute(document =>
{
var metadata = document.Value <RavenJObject>(Constants.Metadata);
var key = metadata.Value <string>("@id");
var etag = Etag.Parse(metadata.Value <string>("@etag"));
var lastModified = DateTime.Parse(metadata.Value <string>(Constants.LastModified));
document.Remove(Constants.Metadata);
var serializedSizeOnDisk = metadata.Value <int>(Constants.SerializedSizeOnDisk);
metadata.Remove(Constants.SerializedSizeOnDisk);
var doc = new JsonDocument
{
DataAsJson = document,
Etag = etag,
Key = key,
SerializedSizeOnDisk = serializedSizeOnDisk,
LastModified = lastModified,
SkipDeleteFromIndex = true,
Metadata = metadata
};
docsToIndex.Add(doc);
totalLoadedDocumentSize += serializedSizeOnDisk;
localLoadedDocumentSize += serializedSizeOnDisk;
if (totalLoadedDocumentSize > Database.Configuration.MaxPrecomputedBatchTotalDocumentSizeInBytes)
{
var error = $"Aborting applying precomputed batch for index id: {index.indexId}, " +
$"name: {index.PublicName} because we have {totalLoadedDocumentSize}MB of documents that were fetched" +
$"and the configured max data to fetch is " +
$"{Database.Configuration.MaxPrecomputedBatchTotalDocumentSizeInBytes/1024/1024}MB";
//we are aborting operation, so don't keep the references
docsToIndex.Clear();
throw new TotalDataSizeExceededException(error);
}
if (localLoadedDocumentSize <= totalSizeToCheck)
{
return;
}
localLoadedDocumentSize = 0;
if (Database.Configuration.MemoryLimitForProcessingInMb > MemoryStatistics.AvailableMemoryInMb)
{
var error = $"Aborting applying precomputed batch for index id: {index.indexId}, " +
$"name: {index.PublicName} because we have {MemoryStatistics.AvailableMemoryInMb}MB " +
$"of available memory and the available memory for processing is: " +
$"{Database.Configuration.MemoryLimitForProcessingInMb}MB";
//we are aborting operation, so don't keep the references
docsToIndex.Clear();
throw new TotalDataSizeExceededException(error);
}
});
result = new PrecomputedIndexingBatch
{
LastIndexed = op.Header.IndexEtag,
LastModified = op.Header.IndexTimestamp,
Documents = docsToIndex,
Index = index
};
}
});
if (result != null && result.Documents != null && result.Documents.Count >= 0)
{
using (var linked = CancellationTokenSource.CreateLinkedTokenSource(cts.Token, WorkContext.CancellationToken))
{
Database.IndexingExecuter.IndexPrecomputedBatch(result, linked.Token);
if (index.IsTestIndex)
{
TransactionalStorage.Batch(accessor => accessor.Indexing.TouchIndexEtag(index.IndexId));
}
}
}
}
private void ApplyPrecomputedBatchForNewIndex(Index index, AbstractViewGenerator generator)
{
const string DocumentsByEntityNameIndex = "Raven/DocumentsByEntityName";
PrecomputedIndexingBatch result = null;
var docsToIndex = new List <JsonDocument>();
TransactionalStorage.Batch(actions =>
{
var countOfDocuments = actions.Documents.GetDocumentsCount();
var tags = generator.ForEntityNames.Select(entityName => "Tag:[[" + entityName + "]]").ToList();
var query = string.Join(" OR ", tags);
var stats =
actions.Indexing.GetIndexStats(
IndexDefinitionStorage.GetIndexDefinition(DocumentsByEntityNameIndex).IndexId);
var lastIndexedEtagByRavenDocumentsByEntityName = stats.LastIndexedEtag;
var lastModifiedByRavenDocumentsByEntityName = stats.LastIndexedTimestamp;
var cts = new CancellationTokenSource();
using (var linked = CancellationTokenSource.CreateLinkedTokenSource(cts.Token, WorkContext.CancellationToken))
using (var op = new QueryActions.DatabaseQueryOperation(Database, DocumentsByEntityNameIndex, new IndexQuery
{
Query = query
}, actions, linked)
{
ShouldSkipDuplicateChecking = true
})
{
op.Init();
if (op.Header.TotalResults == 0 ||
op.Header.TotalResults > (countOfDocuments * 0.25) ||
(op.Header.TotalResults > Database.Configuration.MaxNumberOfItemsToIndexInSingleBatch * 4))
{
// we don't apply this optimization if the total number of results is more than
// 25% of the count of documents (would be easier to just run it regardless).
// or if the number of docs to index is significantly more than the max numbers
// to index in a single batch. The idea here is that we need to keep the amount
// of memory we use to a manageable level even when introducing a new index to a BIG
// database
try
{
cts.Cancel();
// we have to run just a little bit of the query to properly setup the disposal
op.Execute(o => { });
}
catch (OperationCanceledException)
{
}
return;
}
Log.Debug("For new index {0}, using precomputed indexing batch optimization for {1} docs", index,
op.Header.TotalResults);
op.Execute(document =>
{
var metadata = document.Value <RavenJObject>(Constants.Metadata);
var key = metadata.Value <string>("@id");
var etag = Etag.Parse(metadata.Value <string>("@etag"));
var lastModified = DateTime.Parse(metadata.Value <string>(Constants.LastModified));
document.Remove(Constants.Metadata);
docsToIndex.Add(new JsonDocument
{
DataAsJson = document,
Etag = etag,
Key = key,
LastModified = lastModified,
SkipDeleteFromIndex = true,
Metadata = metadata
});
});
}
result = new PrecomputedIndexingBatch
{
LastIndexed = lastIndexedEtagByRavenDocumentsByEntityName,
LastModified = lastModifiedByRavenDocumentsByEntityName,
Documents = docsToIndex,
Index = index
};
});
if (result != null && result.Documents != null && result.Documents.Count > 0)
{
Database.IndexingExecuter.IndexPrecomputedBatch(result);
}
}
private void ApplyPrecomputedBatchForNewIndex(Index index, AbstractViewGenerator generator, int pageSize, CancellationTokenSource cts)
{
PrecomputedIndexingBatch result = null;
var docsToIndex = new List<JsonDocument>();
TransactionalStorage.Batch(actions =>
{
var query = GetQueryForAllMatchingDocumentsForIndex(generator);
using (DocumentCacher.SkipSetDocumentsInDocumentCache())
using (var linked = CancellationTokenSource.CreateLinkedTokenSource(cts.Token, WorkContext.CancellationToken))
using (var op = new QueryActions.DatabaseQueryOperation(Database, Constants.DocumentsByEntityNameIndex, new IndexQuery
{
Query = query,
PageSize = pageSize
}, actions, linked)
{
ShouldSkipDuplicateChecking = true
})
{
op.Init();
//if we are working on a test index, apply the optimization anyway, as the index is capped by small number of results
if (index.IsTestIndex == false && op.Header.TotalResults > pageSize)
{
// we don't apply this optimization if the total number of results
// to index is more than the max numbers to index in a single batch.
// The idea here is that we need to keep the amount
// of memory we use to a manageable level even when introducing a new index to a BIG
// database
try
{
cts.Cancel();
// we have to run just a little bit of the query to properly setup the disposal
op.Execute(o => { });
}
catch (OperationCanceledException)
{
}
return;
}
if (Log.IsDebugEnabled)
{
Log.Debug("For new index {0}, using precomputed indexing batch optimization for {1} docs",
index, op.Header.TotalResults);
}
var totalLoadedDocumentSize = 0;
const int totalSizeToCheck = 16 * 1024 * 1024; //16MB
var localLoadedDocumentSize = 0;
op.Execute(document =>
{
var metadata = document.Value<RavenJObject>(Constants.Metadata);
var key = metadata.Value<string>("@id");
var etag = Etag.Parse(metadata.Value<string>("@etag"));
var lastModified = DateTime.Parse(metadata.Value<string>(Constants.LastModified));
document.Remove(Constants.Metadata);
var serializedSizeOnDisk = metadata.Value<int>(Constants.SerializedSizeOnDisk);
metadata.Remove(Constants.SerializedSizeOnDisk);
var doc = new JsonDocument
{
DataAsJson = document,
Etag = etag,
Key = key,
SerializedSizeOnDisk = serializedSizeOnDisk,
LastModified = lastModified,
SkipDeleteFromIndex = true,
Metadata = metadata
};
docsToIndex.Add(doc);
totalLoadedDocumentSize += serializedSizeOnDisk;
localLoadedDocumentSize += serializedSizeOnDisk;
if (totalLoadedDocumentSize > Database.Configuration.MaxPrecomputedBatchTotalDocumentSizeInBytes)
{
var error = string.Format(
@"Aborting applying precomputed batch for index id: {0}, name: {1}
because we have {2}mb of documents that were fetched
and the configured max data to fetch is {3}mb",
index.indexId, index.PublicName, totalLoadedDocumentSize,
Database.Configuration.MaxPrecomputedBatchTotalDocumentSizeInBytes/1024/1024);
//we are aborting operation, so don't keep the references
docsToIndex.Clear();
throw new TotalDataSizeExceededException(error);
}
if (localLoadedDocumentSize <= totalSizeToCheck)
return;
localLoadedDocumentSize = 0;
if (Database.Configuration.MemoryLimitForProcessingInMb > MemoryStatistics.AvailableMemoryInMb)
{
var error = string.Format(
@"Aborting applying precomputed batch for index id: {0}, name: {1}
because we have {2}mb of available memory and the available memory for processing is: {3}mb",
index.indexId, index.PublicName,
MemoryStatistics.AvailableMemoryInMb, Database.Configuration.MemoryLimitForProcessingInMb);
//we are aborting operation, so don't keep the references
docsToIndex.Clear();
throw new TotalDataSizeExceededException(error);
}
});
result = new PrecomputedIndexingBatch
{
LastIndexed = op.Header.IndexEtag,
LastModified = op.Header.IndexTimestamp,
Documents = docsToIndex,
Index = index
};
}
});
if (result != null && result.Documents != null && result.Documents.Count >= 0)
{
using (var linked = CancellationTokenSource.CreateLinkedTokenSource(cts.Token, WorkContext.CancellationToken))
{
Database.IndexingExecuter.IndexPrecomputedBatch(result, linked.Token);
if (index.IsTestIndex)
TransactionalStorage.Batch(accessor => accessor.Indexing.TouchIndexEtag(index.IndexId));
}
}
}