comparison mercurial/filelog.py @ 50316:87f0155d68aa stable

revlog: improve the robustness of the splitting process The previous "in-place" splitting, preserving the splitting on transaction failure had a couple of issue in case of transaction rollback: - a race windows that could still lead to a crash and data loss - it corrupted the `fncache`. So instead, we use a new approach that we summarized as "we do a backup of the inline revlog pre-split, and we restore this in case of failure". To make readers live easier, we don't overwrite the inline index file until transaction finalization. (once the transaction get into its finalization phase, it is not expected to rollback, unless some crash happens). To do so, we write the index of the split index in a temporary file that we use until transaction finalization. We also keep a backup of the initial inline file to be able to rollback the split if needed. As a result, transaction rollback cancel the split and no longer corrupt fncache. We also no longer have a small inconsistency windows where the transaction could be unrecoverable.
author Pierre-Yves David <pierre-yves.david@octobus.net>
date Mon, 20 Mar 2023 11:52:17 +0100
parents 152d9c011bcd
children 32837c7e2e4b
comparison
equal deleted inserted replaced
50315:cf6e1d535602 50316:87f0155d68aa
23 ) 23 )
24 24
25 25
26 @interfaceutil.implementer(repository.ifilestorage) 26 @interfaceutil.implementer(repository.ifilestorage)
27 class filelog: 27 class filelog:
28 def __init__(self, opener, path): 28 def __init__(self, opener, path, try_split=False):
29 self._revlog = revlog.revlog( 29 self._revlog = revlog.revlog(
30 opener, 30 opener,
31 # XXX should use the unencoded path 31 # XXX should use the unencoded path
32 target=(revlog_constants.KIND_FILELOG, path), 32 target=(revlog_constants.KIND_FILELOG, path),
33 radix=b'/'.join((b'data', path)), 33 radix=b'/'.join((b'data', path)),
34 censorable=True, 34 censorable=True,
35 canonical_parent_order=False, # see comment in revlog.py 35 canonical_parent_order=False, # see comment in revlog.py
36 try_split=try_split,
36 ) 37 )
37 # Full name of the user visible file, relative to the repository root. 38 # Full name of the user visible file, relative to the repository root.
38 # Used by LFS. 39 # Used by LFS.
39 self._revlog.filename = path 40 self._revlog.filename = path
40 self.nullid = self._revlog.nullid 41 self.nullid = self._revlog.nullid
254 255
255 256
256 class narrowfilelog(filelog): 257 class narrowfilelog(filelog):
257 """Filelog variation to be used with narrow stores.""" 258 """Filelog variation to be used with narrow stores."""
258 259
259 def __init__(self, opener, path, narrowmatch): 260 def __init__(self, opener, path, narrowmatch, try_split=False):
260 super(narrowfilelog, self).__init__(opener, path) 261 super(narrowfilelog, self).__init__(opener, path, try_split=try_split)
261 self._narrowmatch = narrowmatch 262 self._narrowmatch = narrowmatch
262 263
263 def renamed(self, node): 264 def renamed(self, node):
264 res = super(narrowfilelog, self).renamed(node) 265 res = super(narrowfilelog, self).renamed(node)
265 266