lbcd/database/cmd/dbtool/insecureimport.go

397 lines
11 KiB
Go
Raw Normal View History

database: Major redesign of database package. This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in #255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes #255 - Fixes #199 - Fixes #201 - Implements and closes #256 - Obsoletes and closes #257 - Closes #247 once the required chain and btcd modifications are in place to make use of this new code
2016-02-03 18:42:04 +01:00
// Copyright (c) 2015-2016 The btcsuite developers
// Use of this source code is governed by an ISC
// license that can be found in the LICENSE file.
package main
import (
"encoding/binary"
"fmt"
"io"
"os"
"sync"
"time"
"github.com/btcsuite/btcd/chaincfg/chainhash"
"github.com/btcsuite/btcd/database"
database: Major redesign of database package. This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in #255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes #255 - Fixes #199 - Fixes #201 - Implements and closes #256 - Obsoletes and closes #257 - Closes #247 once the required chain and btcd modifications are in place to make use of this new code
2016-02-03 18:42:04 +01:00
"github.com/btcsuite/btcd/wire"
"github.com/btcsuite/btcutil"
)
// importCmd defines the configuration options for the insecureimport command.
type importCmd struct {
InFile string `short:"i" long:"infile" description:"File containing the block(s)"`
Progress int `short:"p" long:"progress" description:"Show a progress message each time this number of seconds have passed -- Use 0 to disable progress announcements"`
}
var (
// importCfg defines the configuration options for the command.
importCfg = importCmd{
InFile: "bootstrap.dat",
Progress: 10,
}
// zeroHash is a simply a hash with all zeros. It is defined here to
// avoid creating it multiple times.
zeroHash = chainhash.Hash{}
database: Major redesign of database package. This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in #255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes #255 - Fixes #199 - Fixes #201 - Implements and closes #256 - Obsoletes and closes #257 - Closes #247 once the required chain and btcd modifications are in place to make use of this new code
2016-02-03 18:42:04 +01:00
)
// importResults houses the stats and result as an import operation.
type importResults struct {
blocksProcessed int64
blocksImported int64
err error
}
// blockImporter houses information about an ongoing import from a block data
// file to the block database.
type blockImporter struct {
db database.DB
r io.ReadSeeker
processQueue chan []byte
doneChan chan bool
errChan chan error
quit chan struct{}
wg sync.WaitGroup
blocksProcessed int64
blocksImported int64
receivedLogBlocks int64
receivedLogTx int64
lastHeight int64
lastBlockTime time.Time
lastLogTime time.Time
}
// readBlock reads the next block from the input file.
func (bi *blockImporter) readBlock() ([]byte, error) {
// The block file format is:
// <network> <block length> <serialized block>
var net uint32
err := binary.Read(bi.r, binary.LittleEndian, &net)
if err != nil {
if err != io.EOF {
return nil, err
}
// No block and no error means there are no more blocks to read.
return nil, nil
}
if net != uint32(activeNetParams.Net) {
return nil, fmt.Errorf("network mismatch -- got %x, want %x",
net, uint32(activeNetParams.Net))
}
// Read the block length and ensure it is sane.
var blockLen uint32
if err := binary.Read(bi.r, binary.LittleEndian, &blockLen); err != nil {
return nil, err
}
if blockLen > wire.MaxBlockPayload {
return nil, fmt.Errorf("block payload of %d bytes is larger "+
"than the max allowed %d bytes", blockLen,
wire.MaxBlockPayload)
}
serializedBlock := make([]byte, blockLen)
if _, err := io.ReadFull(bi.r, serializedBlock); err != nil {
return nil, err
}
return serializedBlock, nil
}
// processBlock potentially imports the block into the database. It first
// deserializes the raw block while checking for errors. Already known blocks
// are skipped and orphan blocks are considered errors. Returns whether the
// block was imported along with any potential errors.
//
// NOTE: This is not a safe import as it does not verify chain rules.
func (bi *blockImporter) processBlock(serializedBlock []byte) (bool, error) {
// Deserialize the block which includes checks for malformed blocks.
block, err := btcutil.NewBlockFromBytes(serializedBlock)
if err != nil {
return false, err
}
// update progress statistics
bi.lastBlockTime = block.MsgBlock().Header.Timestamp
bi.receivedLogTx += int64(len(block.MsgBlock().Transactions))
// Skip blocks that already exist.
var exists bool
err = bi.db.View(func(tx database.Tx) error {
exists, err = tx.HasBlock(block.Hash())
return err
database: Major redesign of database package. This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in #255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes #255 - Fixes #199 - Fixes #201 - Implements and closes #256 - Obsoletes and closes #257 - Closes #247 once the required chain and btcd modifications are in place to make use of this new code
2016-02-03 18:42:04 +01:00
})
if err != nil {
return false, err
}
if exists {
return false, nil
}
// Don't bother trying to process orphans.
prevHash := &block.MsgBlock().Header.PrevBlock
if !prevHash.IsEqual(&zeroHash) {
var exists bool
err := bi.db.View(func(tx database.Tx) error {
exists, err = tx.HasBlock(prevHash)
return err
database: Major redesign of database package. This commit contains a complete redesign and rewrite of the database package that approaches things in a vastly different manner than the previous version. This is the first part of several stages that will be needed to ultimately make use of this new package. Some of the reason for this were discussed in #255, however a quick summary is as follows: - The previous database could only contain blocks on the main chain and reorgs required deleting the blocks from the database. This made it impossible to store orphans and could make external RPC calls for information about blocks during the middle of a reorg fail. - The previous database interface forced a high level of bitcoin-specific intelligence such as spend tracking into each backend driver. - The aforementioned point led to making it difficult to implement new backend drivers due to the need to repeat a lot of non-trivial logic which is better handled at a higher layer, such as the blockchain package. - The old database stored all blocks in leveldb. This made it extremely inefficient to do things such as lookup headers and individual transactions since the entire block had to be loaded from leveldb (which entails it doing data copies) to get access. In order to address all of these concerns, and others not mentioned, the database interface has been redesigned as follows: - Two main categories of functionality are provided: block storage and metadata storage - All block storage and metadata storage are done via read-only and read-write MVCC transactions with both manual and managed modes - Support for multiple concurrent readers and a single writer - Readers use a snapshot and therefore are not blocked by the writer - Some key properties of the block storage and retrieval API: - It is generic and does NOT contain additional bitcoin logic such spend tracking and block linking - Provides access to the raw serialized bytes so deserialization is not forced for callers that don't need it - Support for fetching headers via independent functions which allows implementations to provide significant optimizations - Ability to efficiently retrieve arbitrary regions of blocks (transactions, scripts, etc) - A rich metadata storage API is provided: - Key/value with arbitrary data - Support for buckets and nested buckets - Bucket iteration through a couple of different mechanisms - Cursors for efficient and direct key seeking - Supports registration of backend database implementations - Comprehensive test coverage - Provides strong documentation with example usage This commit also contains an implementation of the previously discussed interface named ffldb (flat file plus leveldb metadata backend). Here is a quick overview: - Highly optimized for read performance with consistent write performance regardless of database size - All blocks are stored in flat files on the file system - Bulk block region fetching is optimized to perform linear reads which improves performance on spindle disks - Anti-corruption mechanisms: - Flat files contain full block checksums to quickly an easily detect database corruption without needing to do expensive merkle root calculations - Metadata checksums - Open reconciliation - Extensive test coverage: - Comprehensive blackbox interface testing - Whitebox testing which uses intimate knowledge to exercise uncommon failure paths such as deleting files out from under the database - Corruption tests (replacing random data in the files) In addition, this commit also contains a new tool under the new database directory named dbtool which provides a few basic commands for testing the database. It is designed around commands, so it could be useful to expand on in the future. Finally, this commit addresses the following issues: - Adds support for and therefore closes #255 - Fixes #199 - Fixes #201 - Implements and closes #256 - Obsoletes and closes #257 - Closes #247 once the required chain and btcd modifications are in place to make use of this new code
2016-02-03 18:42:04 +01:00
})
if err != nil {
return false, err
}
if !exists {
return false, fmt.Errorf("import file contains block "+
"%v which does not link to the available "+
"block chain", prevHash)
}
}
// Put the blocks into the database with no checking of chain rules.
err = bi.db.Update(func(tx database.Tx) error {
return tx.StoreBlock(block)
})
if err != nil {
return false, err
}
return true, nil
}
// readHandler is the main handler for reading blocks from the import file.
// This allows block processing to take place in parallel with block reads.
// It must be run as a goroutine.
func (bi *blockImporter) readHandler() {
out:
for {
// Read the next block from the file and if anything goes wrong
// notify the status handler with the error and bail.
serializedBlock, err := bi.readBlock()
if err != nil {
bi.errChan <- fmt.Errorf("Error reading from input "+
"file: %v", err.Error())
break out
}
// A nil block with no error means we're done.
if serializedBlock == nil {
break out
}
// Send the block or quit if we've been signalled to exit by
// the status handler due to an error elsewhere.
select {
case bi.processQueue <- serializedBlock:
case <-bi.quit:
break out
}
}
// Close the processing channel to signal no more blocks are coming.
close(bi.processQueue)
bi.wg.Done()
}
// logProgress logs block progress as an information message. In order to
// prevent spam, it limits logging to one message every importCfg.Progress
// seconds with duration and totals included.
func (bi *blockImporter) logProgress() {
bi.receivedLogBlocks++
now := time.Now()
duration := now.Sub(bi.lastLogTime)
if duration < time.Second*time.Duration(importCfg.Progress) {
return
}
// Truncate the duration to 10s of milliseconds.
durationMillis := int64(duration / time.Millisecond)
tDuration := 10 * time.Millisecond * time.Duration(durationMillis/10)
// Log information about new block height.
blockStr := "blocks"
if bi.receivedLogBlocks == 1 {
blockStr = "block"
}
txStr := "transactions"
if bi.receivedLogTx == 1 {
txStr = "transaction"
}
log.Infof("Processed %d %s in the last %s (%d %s, height %d, %s)",
bi.receivedLogBlocks, blockStr, tDuration, bi.receivedLogTx,
txStr, bi.lastHeight, bi.lastBlockTime)
bi.receivedLogBlocks = 0
bi.receivedLogTx = 0
bi.lastLogTime = now
}
// processHandler is the main handler for processing blocks. This allows block
// processing to take place in parallel with block reads from the import file.
// It must be run as a goroutine.
func (bi *blockImporter) processHandler() {
out:
for {
select {
case serializedBlock, ok := <-bi.processQueue:
// We're done when the channel is closed.
if !ok {
break out
}
bi.blocksProcessed++
bi.lastHeight++
imported, err := bi.processBlock(serializedBlock)
if err != nil {
bi.errChan <- err
break out
}
if imported {
bi.blocksImported++
}
bi.logProgress()
case <-bi.quit:
break out
}
}
bi.wg.Done()
}
// statusHandler waits for updates from the import operation and notifies
// the passed doneChan with the results of the import. It also causes all
// goroutines to exit if an error is reported from any of them.
func (bi *blockImporter) statusHandler(resultsChan chan *importResults) {
select {
// An error from either of the goroutines means we're done so signal
// caller with the error and signal all goroutines to quit.
case err := <-bi.errChan:
resultsChan <- &importResults{
blocksProcessed: bi.blocksProcessed,
blocksImported: bi.blocksImported,
err: err,
}
close(bi.quit)
// The import finished normally.
case <-bi.doneChan:
resultsChan <- &importResults{
blocksProcessed: bi.blocksProcessed,
blocksImported: bi.blocksImported,
err: nil,
}
}
}
// Import is the core function which handles importing the blocks from the file
// associated with the block importer to the database. It returns a channel
// on which the results will be returned when the operation has completed.
func (bi *blockImporter) Import() chan *importResults {
// Start up the read and process handling goroutines. This setup allows
// blocks to be read from disk in parallel while being processed.
bi.wg.Add(2)
go bi.readHandler()
go bi.processHandler()
// Wait for the import to finish in a separate goroutine and signal
// the status handler when done.
go func() {
bi.wg.Wait()
bi.doneChan <- true
}()
// Start the status handler and return the result channel that it will
// send the results on when the import is done.
resultChan := make(chan *importResults)
go bi.statusHandler(resultChan)
return resultChan
}
// newBlockImporter returns a new importer for the provided file reader seeker
// and database.
func newBlockImporter(db database.DB, r io.ReadSeeker) *blockImporter {
return &blockImporter{
db: db,
r: r,
processQueue: make(chan []byte, 2),
doneChan: make(chan bool),
errChan: make(chan error),
quit: make(chan struct{}),
lastLogTime: time.Now(),
}
}
// Execute is the main entry point for the command. It's invoked by the parser.
func (cmd *importCmd) Execute(args []string) error {
// Setup the global config options and ensure they are valid.
if err := setupGlobalConfig(); err != nil {
return err
}
// Ensure the specified block file exists.
if !fileExists(cmd.InFile) {
str := "The specified block file [%v] does not exist"
return fmt.Errorf(str, cmd.InFile)
}
// Load the block database.
db, err := loadBlockDB()
if err != nil {
return err
}
defer db.Close()
// Ensure the database is sync'd and closed on Ctrl+C.
addInterruptHandler(func() {
log.Infof("Gracefully shutting down the database...")
db.Close()
})
fi, err := os.Open(importCfg.InFile)
if err != nil {
return err
}
defer fi.Close()
// Create a block importer for the database and input file and start it.
// The results channel returned from start will contain an error if
// anything went wrong.
importer := newBlockImporter(db, fi)
// Perform the import asynchronously and signal the main goroutine when
// done. This allows blocks to be processed and read in parallel. The
// results channel returned from Import contains the statistics about
// the import including an error if something went wrong. This is done
// in a separate goroutine rather than waiting directly so the main
// goroutine can be signaled for shutdown by either completion, error,
// or from the main interrupt handler. This is necessary since the main
// goroutine must be kept running long enough for the interrupt handler
// goroutine to finish.
go func() {
log.Info("Starting import")
resultsChan := importer.Import()
results := <-resultsChan
if results.err != nil {
dbErr, ok := results.err.(database.Error)
if !ok || ok && dbErr.ErrorCode != database.ErrDbNotOpen {
shutdownChannel <- results.err
return
}
}
log.Infof("Processed a total of %d blocks (%d imported, %d "+
"already known)", results.blocksProcessed,
results.blocksImported,
results.blocksProcessed-results.blocksImported)
shutdownChannel <- nil
}()
// Wait for shutdown signal from either a normal completion or from the
// interrupt handler.
err = <-shutdownChannel
return err
}