services/datamanager/collection/collection.go

   1 // Deals with parsing Collection responses from API Server.
   2
   3 package collection
   4
   5 import (
   6         "flag"
   7         "fmt"
   8         "git.curoverse.com/arvados.git/sdk/go/arvadosclient"
   9         "git.curoverse.com/arvados.git/sdk/go/blockdigest"
  10         "git.curoverse.com/arvados.git/sdk/go/logger"
  11         "git.curoverse.com/arvados.git/sdk/go/manifest"
  12         "git.curoverse.com/arvados.git/sdk/go/util"
  13         "log"
  14         "os"
  15         "runtime/pprof"
  16         "time"
  17 )
  18
  19 var (
  20         HeapProfileFilename string
  21 )
  22
  23 // Collection representation
  24 type Collection struct {
  25         UUID              string
  26         OwnerUUID         string
  27         ReplicationLevel  int
  28         BlockDigestToSize map[blockdigest.BlockDigest]int
  29         TotalSize         int
  30 }
  31
  32 // ReadCollections holds information about collections from API server
  33 type ReadCollections struct {
  34         ReadAllCollections        bool
  35         UUIDToCollection          map[string]Collection
  36         OwnerToCollectionSize     map[string]int
  37         BlockToDesiredReplication map[blockdigest.DigestWithSize]int
  38         CollectionUUIDToIndex     map[string]int
  39         CollectionIndexToUUID     []string
  40         BlockToCollectionIndices  map[blockdigest.DigestWithSize][]int
  41 }
  42
  43 // GetCollectionsParams params
  44 type GetCollectionsParams struct {
  45         Client    arvadosclient.ArvadosClient
  46         Logger    *logger.Logger
  47         BatchSize int
  48 }
  49
  50 // SdkCollectionInfo holds collection info from api
  51 type SdkCollectionInfo struct {
  52         UUID         string    `json:"uuid"`
  53         OwnerUUID    string    `json:"owner_uuid"`
  54         Redundancy   int       `json:"redundancy"`
  55         ModifiedAt   time.Time `json:"modified_at"`
  56         ManifestText string    `json:"manifest_text"`
  57 }
  58
  59 // SdkCollectionList lists collections from api
  60 type SdkCollectionList struct {
  61         ItemsAvailable int                 `json:"items_available"`
  62         Items          []SdkCollectionInfo `json:"items"`
  63 }
  64
  65 func init() {
  66         flag.StringVar(&HeapProfileFilename,
  67                 "heap-profile",
  68                 "",
  69                 "File to write the heap profiles to. Leave blank to skip profiling.")
  70 }
  71
  72 // WriteHeapProfile writes the heap profile to a file for later review.
  73 // Since a file is expected to only contain a single heap profile this
  74 // function overwrites the previously written profile, so it is safe
  75 // to call multiple times in a single run.
  76 // Otherwise we would see cumulative numbers as explained here:
  77 // https://groups.google.com/d/msg/golang-nuts/ZyHciRglQYc/2nh4Ndu2fZcJ
  78 func WriteHeapProfile() error {
  79         if HeapProfileFilename != "" {
  80                 heapProfile, err := os.Create(HeapProfileFilename)
  81                 if err != nil {
  82                         return err
  83                 }
  84
  85                 defer heapProfile.Close()
  86
  87                 err = pprof.WriteHeapProfile(heapProfile)
  88                 return err
  89         }
  90
  91         return nil
  92 }
  93
  94 // GetCollectionsAndSummarize gets collections from api and summarizes
  95 func GetCollectionsAndSummarize(params GetCollectionsParams) (results ReadCollections, err error) {
  96         results, err = GetCollections(params)
  97         if err != nil {
  98                 return
  99         }
 100
 101         results.Summarize(params.Logger)
 102
 103         log.Printf("Uuid to Size used: %v", results.OwnerToCollectionSize)
 104         log.Printf("Read and processed %d collections",
 105                 len(results.UUIDToCollection))
 106
 107         // TODO(misha): Add a "readonly" flag. If we're in readonly mode,
 108         // lots of behaviors can become warnings (and obviously we can't
 109         // write anything).
 110         // if !readCollections.ReadAllCollections {
 111         //      log.Fatalf("Did not read all collections")
 112         // }
 113
 114         return
 115 }
 116
 117 // GetCollections gets collections from api
 118 func GetCollections(params GetCollectionsParams) (results ReadCollections, err error) {
 119         if &params.Client == nil {
 120                 err = fmt.Errorf("params.Client passed to GetCollections() should " +
 121                         "contain a valid ArvadosClient, but instead it is nil.")
 122                 return
 123         }
 124
 125         fieldsWanted := []string{"manifest_text",
 126                 "owner_uuid",
 127                 "uuid",
 128                 "redundancy",
 129                 "modified_at"}
 130
 131         sdkParams := arvadosclient.Dict{
 132                 "select":  fieldsWanted,
 133                 "order":   []string{"modified_at ASC"},
 134                 "filters": [][]string{[]string{"modified_at", ">=", "1900-01-01T00:00:00Z"}}}
 135
 136         if params.BatchSize > 0 {
 137                 sdkParams["limit"] = params.BatchSize
 138         }
 139
 140         var defaultReplicationLevel int
 141         {
 142                 var value interface{}
 143                 value, err = params.Client.Discovery("defaultCollectionReplication")
 144                 if err != nil {
 145                         return
 146                 }
 147
 148                 defaultReplicationLevel = int(value.(float64))
 149                 if defaultReplicationLevel <= 0 {
 150                         err = fmt.Errorf("Default collection replication returned by arvados SDK "+
 151                                 "should be a positive integer but instead it was %d.",
 152                                 defaultReplicationLevel)
 153                         return
 154                 }
 155         }
 156
 157         initialNumberOfCollectionsAvailable, err :=
 158                 util.NumberItemsAvailable(params.Client, "collections")
 159         if err != nil {
 160                 return
 161         }
 162         // Include a 1% margin for collections added while we're reading so
 163         // that we don't have to grow the map in most cases.
 164         maxExpectedCollections := int(
 165                 float64(initialNumberOfCollectionsAvailable) * 1.01)
 166         results.UUIDToCollection = make(map[string]Collection, maxExpectedCollections)
 167
 168         if params.Logger != nil {
 169                 params.Logger.Update(func(p map[string]interface{}, e map[string]interface{}) {
 170                         collectionInfo := logger.GetOrCreateMap(p, "collection_info")
 171                         collectionInfo["num_collections_at_start"] = initialNumberOfCollectionsAvailable
 172                         collectionInfo["batch_size"] = params.BatchSize
 173                         collectionInfo["default_replication_level"] = defaultReplicationLevel
 174                 })
 175         }
 176
 177         // These values are just for getting the loop to run the first time,
 178         // afterwards they'll be set to real values.
 179         previousTotalCollections := -1
 180         totalCollections := 0
 181         for totalCollections > previousTotalCollections {
 182                 // We're still finding new collections
 183
 184                 // Write the heap profile for examining memory usage
 185                 err = WriteHeapProfile()
 186                 if err != nil {
 187                         return
 188                 }
 189
 190                 // Get next batch of collections.
 191                 var collections SdkCollectionList
 192                 err = params.Client.List("collections", sdkParams, &collections)
 193                 if err != nil {
 194                         return
 195                 }
 196
 197                 // Process collection and update our date filter.
 198                 latestModificationDate, maxManifestSize, totalManifestSize, err := ProcessCollections(params.Logger,
 199                         collections.Items,
 200                         defaultReplicationLevel,
 201                         results.UUIDToCollection)
 202                 if err != nil {
 203                         return results, err
 204                 }
 205                 sdkParams["filters"].([][]string)[0][2] = latestModificationDate.Format(time.RFC3339)
 206
 207                 // update counts
 208                 previousTotalCollections = totalCollections
 209                 totalCollections = len(results.UUIDToCollection)
 210
 211                 log.Printf("%d collections read, %d new in last batch, "+
 212                         "%s latest modified date, %.0f %d %d avg,max,total manifest size",
 213                         totalCollections,
 214                         totalCollections-previousTotalCollections,
 215                         sdkParams["filters"].([][]string)[0][2],
 216                         float32(totalManifestSize)/float32(totalCollections),
 217                         maxManifestSize, totalManifestSize)
 218
 219                 if params.Logger != nil {
 220                         params.Logger.Update(func(p map[string]interface{}, e map[string]interface{}) {
 221                                 collectionInfo := logger.GetOrCreateMap(p, "collection_info")
 222                                 collectionInfo["collections_read"] = totalCollections
 223                                 collectionInfo["latest_modified_date_seen"] = sdkParams["filters"].([][]string)[0][2]
 224                                 collectionInfo["total_manifest_size"] = totalManifestSize
 225                                 collectionInfo["max_manifest_size"] = maxManifestSize
 226                         })
 227                 }
 228         }
 229
 230         // Write the heap profile for examining memory usage
 231         err = WriteHeapProfile()
 232
 233         return
 234 }
 235
 236 // StrCopy returns a newly allocated string.
 237 // It is useful to copy slices so that the garbage collector can reuse
 238 // the memory of the longer strings they came from.
 239 func StrCopy(s string) string {
 240         return string([]byte(s))
 241 }
 242
 243 // ProcessCollections read from api server
 244 func ProcessCollections(arvLogger *logger.Logger,
 245         receivedCollections []SdkCollectionInfo,
 246         defaultReplicationLevel int,
 247         UUIDToCollection map[string]Collection,
 248 ) (
 249         latestModificationDate time.Time,
 250         maxManifestSize, totalManifestSize uint64,
 251         err error,
 252 ) {
 253         for _, sdkCollection := range receivedCollections {
 254                 collection := Collection{UUID: StrCopy(sdkCollection.UUID),
 255                         OwnerUUID:         StrCopy(sdkCollection.OwnerUUID),
 256                         ReplicationLevel:  sdkCollection.Redundancy,
 257                         BlockDigestToSize: make(map[blockdigest.BlockDigest]int)}
 258
 259                 if sdkCollection.ModifiedAt.IsZero() {
 260                         err = fmt.Errorf(
 261                                 "Arvados SDK collection returned with unexpected zero "+
 262                                         "modification date. This probably means that either we failed to "+
 263                                         "parse the modification date or the API server has changed how "+
 264                                         "it returns modification dates: %+v",
 265                                 collection)
 266                         return
 267                 }
 268
 269                 if sdkCollection.ModifiedAt.After(latestModificationDate) {
 270                         latestModificationDate = sdkCollection.ModifiedAt
 271                 }
 272
 273                 if collection.ReplicationLevel == 0 {
 274                         collection.ReplicationLevel = defaultReplicationLevel
 275                 }
 276
 277                 manifest := manifest.Manifest{Text: sdkCollection.ManifestText}
 278                 manifestSize := uint64(len(sdkCollection.ManifestText))
 279
 280                 if _, alreadySeen := UUIDToCollection[collection.UUID]; !alreadySeen {
 281                         totalManifestSize += manifestSize
 282                 }
 283                 if manifestSize > maxManifestSize {
 284                         maxManifestSize = manifestSize
 285                 }
 286
 287                 blockChannel := manifest.BlockIterWithDuplicates()
 288                 for block := range blockChannel {
 289                         if storedSize, stored := collection.BlockDigestToSize[block.Digest]; stored && storedSize != block.Size {
 290                                 log.Printf(
 291                                         "Collection %s contains multiple sizes (%d and %d) for block %s",
 292                                         collection.UUID,
 293                                         storedSize,
 294                                         block.Size,
 295                                         block.Digest)
 296                         }
 297                         collection.BlockDigestToSize[block.Digest] = block.Size
 298                 }
 299                 if manifest.Err != nil {
 300                         err = manifest.Err
 301                         return
 302                 }
 303
 304                 collection.TotalSize = 0
 305                 for _, size := range collection.BlockDigestToSize {
 306                         collection.TotalSize += size
 307                 }
 308                 UUIDToCollection[collection.UUID] = collection
 309
 310                 // Clear out all the manifest strings that we don't need anymore.
 311                 // These hopefully form the bulk of our memory usage.
 312                 manifest.Text = ""
 313                 sdkCollection.ManifestText = ""
 314         }
 315
 316         return
 317 }
 318
 319 // Summarize the collections read
 320 func (readCollections *ReadCollections) Summarize(arvLogger *logger.Logger) {
 321         readCollections.OwnerToCollectionSize = make(map[string]int)
 322         readCollections.BlockToDesiredReplication = make(map[blockdigest.DigestWithSize]int)
 323         numCollections := len(readCollections.UUIDToCollection)
 324         readCollections.CollectionUUIDToIndex = make(map[string]int, numCollections)
 325         readCollections.CollectionIndexToUUID = make([]string, 0, numCollections)
 326         readCollections.BlockToCollectionIndices = make(map[blockdigest.DigestWithSize][]int)
 327
 328         for _, coll := range readCollections.UUIDToCollection {
 329                 collectionIndex := len(readCollections.CollectionIndexToUUID)
 330                 readCollections.CollectionIndexToUUID =
 331                         append(readCollections.CollectionIndexToUUID, coll.UUID)
 332                 readCollections.CollectionUUIDToIndex[coll.UUID] = collectionIndex
 333
 334                 readCollections.OwnerToCollectionSize[coll.OwnerUUID] =
 335                         readCollections.OwnerToCollectionSize[coll.OwnerUUID] + coll.TotalSize
 336
 337                 for block, size := range coll.BlockDigestToSize {
 338                         locator := blockdigest.DigestWithSize{Digest: block, Size: uint32(size)}
 339                         readCollections.BlockToCollectionIndices[locator] =
 340                                 append(readCollections.BlockToCollectionIndices[locator],
 341                                         collectionIndex)
 342                         storedReplication := readCollections.BlockToDesiredReplication[locator]
 343                         if coll.ReplicationLevel > storedReplication {
 344                                 readCollections.BlockToDesiredReplication[locator] =
 345                                         coll.ReplicationLevel
 346                         }
 347                 }
 348         }
 349
 350         if arvLogger != nil {
 351                 arvLogger.Update(func(p map[string]interface{}, e map[string]interface{}) {
 352                         collectionInfo := logger.GetOrCreateMap(p, "collection_info")
 353                         // Since maps are shallow copied, we run a risk of concurrent
 354                         // updates here. By copying results.OwnerToCollectionSize into
 355                         // the log, we're assuming that it won't be updated.
 356                         collectionInfo["owner_to_collection_size"] =
 357                                 readCollections.OwnerToCollectionSize
 358                         collectionInfo["distinct_blocks_named"] =
 359                                 len(readCollections.BlockToDesiredReplication)
 360                 })
 361         }
 362
 363         return
 364 }