Files
duplicati/Duplicati/Library/Main/Operation/RecreateDatabaseHandler.cs
T
2013-07-22 16:54:19 +02:00

269 lines
15 KiB
C#

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using Duplicati.Library.Main.Database;
using Duplicati.Library.Main.Volumes;
namespace Duplicati.Library.Main.Operation
{
internal class RecreateDatabaseHandler : IDisposable
{
private string m_backendurl;
private Options m_options;
private RecreateDatabaseResults m_result;
public delegate IEnumerable<KeyValuePair<long, IParsedVolume>> NumberedFilterFilelistDelegate(IEnumerable<IParsedVolume> filelist);
public delegate void BlockVolumePostProcessor(string volumename,BlockVolumeReader reader);
public RecreateDatabaseHandler(string backendurl, Options options, RecreateDatabaseResults result)
{
m_options = options;
m_backendurl = backendurl;
m_result = result;
}
/// <summary>
/// Run the recreate procedure
/// </summary>
/// <param name="path">Path to the database that will be created</param>
/// <param name="filelistfilter">A filter that can be used to disregard certain remote files, intended to be used to select a certain filelist</param>
/// <param name="filenamefilter">Filters the files in a filelist to prevent downloading unwanted data</param>
/// <param name="blockprocessor">A callback hook that can be used to work with downloaded block volumes, intended to be use to recover data blocks while processing blocklists</param>
public void Run(string path, Library.Utility.IFilter filter = null, NumberedFilterFilelistDelegate filelistfilter = null, BlockVolumePostProcessor blockprocessor = null)
{
if (System.IO.File.Exists(path))
throw new Exception(string.Format("Cannot recreate database because file already exists: {0}", path));
using(var db = new LocalDatabase(path, "Recreate"))
{
m_result.SetDatabase(db);
DoRun(db, filter, filelistfilter, blockprocessor);
}
}
/// <summary>
/// Run the recreate procedure
/// </summary>
/// <param name="path">Path to the database that will be created</param>
/// <param name="filelistfilter">A filter that can be used to disregard certain remote files, intended to be used to select a certain filelist</param>
/// <param name="filenamefilter">Filters the files in a filelist to prevent downloading unwanted data</param>
/// <param name="blockprocessor">A callback hook that can be used to work with downloaded block volumes, intended to be use to recover data blocks while processing blocklists</param>
internal void DoRun(LocalDatabase dbparent, Library.Utility.IFilter filter = null, NumberedFilterFilelistDelegate filelistfilter = null, BlockVolumePostProcessor blockprocessor = null)
{
var hashalg = System.Security.Cryptography.HashAlgorithm.Create(m_options.BlockHashAlgorithm);
if (hashalg == null)
throw new Exception(string.Format(Strings.Foresthash.InvalidHashAlgorithm, m_options.BlockHashAlgorithm));
var hashsize = hashalg.HashSize / 8;
//We build a local database in steps.
using(var restoredb = new LocalRecreateDatabase(dbparent, m_options))
using(var backend = new BackendManager(m_backendurl, m_options, m_result.BackendWriter, restoredb))
{
var volumeIds = new Dictionary<string, long>();
var rawlist = backend.List();
//First step is to examine the remote storage to see what
// kind of data we can find
var remotefiles =
(from x in rawlist
let n = VolumeBase.ParseFilename(x)
where
n != null
&&
n.Prefix == m_options.Prefix
select n).ToArray(); //ToArray() ensures that we do not remote-request it multiple times
if (remotefiles.Length == 0)
{
if (rawlist.Count == 0)
throw new Exception("No files were found at the remote location, perhaps the target url is incorrect?");
else
{
var tmp =
(from x in rawlist
let n = VolumeBase.ParseFilename(x)
where
n != null
select n.Prefix).ToArray();
var types = tmp.Distinct().ToArray();
if (tmp.Length == 0)
throw new Exception(string.Format("Found {0} files at the remote storage, but none that could be parsed", rawlist.Count));
else if (types.Length == 1)
throw new Exception(string.Format("Found {0} parse-able files with the prefix {1}, did you forget to set the backup-prefix?", tmp.Length, types[0]));
else
throw new Exception(string.Format("Found {0} parse-able files (of {1} files) with different prefixes: {2}, did you forget to set the backup-prefix?", tmp.Length, rawlist.Count, string.Join(", ", types)));
}
}
//Then we select the filelist we should work with,
// and create the filelist table to fit
IEnumerable<IParsedVolume> filelists =
from n in remotefiles
where n.FileType == RemoteVolumeType.Files
orderby n.Time descending
select n;
if (filelistfilter != null)
filelists = filelistfilter(filelists).Select(x => x.Value).ToArray();
foreach(var fl in remotefiles)
volumeIds[fl.File.Name] = restoredb.RegisterRemoteVolume(fl.File.Name, fl.FileType, RemoteVolumeState.Uploaded);
//Record all blocksets and files needed
using(var tr = restoredb.BeginTransaction())
{
var filelistWork = (from n in filelists orderby n.Time select new RemoteVolume(n.File) as IRemoteVolume).ToList();
foreach(var entry in new AsyncDownloader(filelistWork, backend))
try
{
using(var tmpfile = entry.TempFile)
{
if (entry.Hash != null && entry.Size > 0)
restoredb.UpdateRemoteVolume(entry.Name, RemoteVolumeState.Verified, entry.Size, entry.Hash, tr);
var parsed = VolumeBase.ParseFilename(entry.Name);
// Create timestamped operations based on the file timestamp
var filesetid = restoredb.CreateFileset(volumeIds[entry.Name], parsed.Time, tr);
using(var filelistreader = new FilesetVolumeReader(parsed.CompressionModule, tmpfile, m_options))
foreach(var fe in filelistreader.Files.Where(x => Library.Utility.FilterExpression.Matches(filter, x.Path)))
{
try
{
if (fe.Type == FilelistEntryType.Folder)
{
restoredb.AddDirectoryEntry(filesetid, fe.Path, fe.Time, fe.Metahash, fe.Metahash == null ? -1 : fe.Metasize, tr);
}
else if (fe.Type == FilelistEntryType.File)
{
var blocksetid = restoredb.AddBlockset(fe.Hash, fe.Size, fe.BlocklistHashes, tr);
restoredb.AddFileEntry(filesetid, fe.Path, fe.Time, blocksetid, fe.Metahash, fe.Metahash == null ? -1 : fe.Metasize, tr);
}
else if (fe.Type == FilelistEntryType.Symlink)
{
restoredb.AddSymlinkEntry(filesetid, fe.Path, fe.Time, fe.Metahash, fe.Metahash == null ? -1 : fe.Metasize, tr);
}
else
{
m_result.AddWarning(string.Format("Skipping file-entry with unknown type {0}: {1} ", fe.Type, fe.Path), null);
}
}
catch (Exception ex)
{
m_result.AddWarning(string.Format("Failed to process file-entry: {0}", fe.Path), ex);
}
}
}
}
catch (Exception ex)
{
m_result.AddWarning(string.Format("Failed to process file: {0}", entry.Name), ex);
}
using(new Logging.Timer("CommitUpdateFilesetFromRemote"))
tr.Commit();
}
//Grab all index files, and update the block table
using(var tr = restoredb.BeginTransaction())
{
var indexfiles =
from n in remotefiles
where n.FileType == RemoteVolumeType.Index
select new RemoteVolume(n.File) as IRemoteVolume;
foreach(var sf in new AsyncDownloader(indexfiles.ToList(), backend))
try
{
using(var tmpfile = sf.TempFile)
{
if (sf.Hash != null && sf.Size > 0)
restoredb.UpdateRemoteVolume(sf.Name, RemoteVolumeState.Verified, sf.Size, sf.Hash, tr);
using(var svr = new IndexVolumeReader(RestoreHandler.GetCompressionModule(sf.Name), tmpfile, m_options, hashsize))
{
Utility.VerifyParameters(restoredb, m_options);
foreach(var a in svr.Volumes)
{
var volumeID = restoredb.GetRemoteVolumeID(a.Filename);
//Add all block/volume mappings
foreach(var b in a.Blocks)
restoredb.UpdateBlock(b.Key, b.Value, volumeID, tr);
restoredb.UpdateRemoteVolume(a.Filename, RemoteVolumeState.Verified, a.Length, a.Hash, tr);
restoredb.AddIndexBlockLink(restoredb.GetRemoteVolumeID(sf.Name), volumeID, tr);
}
//If there are blocklists in the index file, update the blocklists
foreach(var b in svr.BlockLists)
restoredb.UpdateBlockset(b.Hash, b.Blocklist, hashsize, tr);
}
}
}
catch (Exception ex)
{
//Not fatal
m_result.AddWarning(string.Format("Failed to process index file: {0}", sf.Name), ex);
}
using(new Logging.Timer("CommitRecreatedDb"))
tr.Commit();
// TODO: In some cases, we can avoid downloading all index files,
// if we are lucky and pick the right ones
}
// We have now grabbed as much information as possible,
// if we are still missing data, we must now fetch block files
restoredb.FindMissingBlocklistHashes(hashsize, null);
//We do this in three passes
for(var i = 0; i < 3; i++)
{
// Grab the list matching the pass type
var lst = restoredb.GetMissingBlockListVolumes(i).ToList();
foreach (var sf in new AsyncDownloader(lst, backend))
using (var tmpfile = sf.TempFile)
using (var rd = new BlockVolumeReader(RestoreHandler.GetCompressionModule(sf.Name), tmpfile, m_options))
using (var tr = restoredb.BeginTransaction())
{
var volumeid = restoredb.GetRemoteVolumeID(sf.Name);
// Update the block table so we know about the block/volume map
foreach(var h in rd.Blocks)
restoredb.UpdateBlock(h.Key, h.Value, volumeid, tr);
// Grab all known blocklists from the volume
foreach (var blocklisthash in restoredb.GetBlockLists(volumeid))
restoredb.UpdateBlockset(blocklisthash, rd.ReadBlocklist(blocklisthash, hashsize), hashsize, tr);
// Update tables so we know if we are done
restoredb.FindMissingBlocklistHashes(hashsize, tr);
using(new Logging.Timer("CommitRestoredBlocklist"))
tr.Commit();
//At this point we can patch files with data from the block volume
if (blockprocessor != null)
blockprocessor(sf.Name, rd);
}
}
backend.WaitForComplete(restoredb, null);
//All done, we must verify that we have all blocklist fully intact
// if this fails, the db will not be deleted, so it can be used,
// except to continue a backup
restoredb.VerifyConsistency(null);
}
}
public void Dispose()
{
}
}
}