summaryrefslogtreecommitdiff
AgeCommit message (Collapse)AuthorFilesLines
2010-07-05Prepare for 0.11 release.version_0_11Lars Wirzenius1-1/+1
2010-07-05Prepare for 0.11 release.Lars Wirzenius3-1/+33
2010-07-05Improve progress reporting during backups.Lars Wirzenius2-14/+23
2010-07-05Add --upload-queue-size for tweaking size of upload queue for B-tree nodes.Lars Wirzenius12-39/+76
2010-07-05Change checkpoint interval to 1 GiB.Lars Wirzenius1-1/+1
2010-07-04Merge changes to allow user to choose node, chunk, chunk group sizes.Lars Wirzenius14-44/+82
2010-07-04Add --node-size option and make everywhere use that.Lars Wirzenius12-39/+46
2010-07-04Add --chunk-group-size option.Lars Wirzenius2-1/+8
2010-07-04Add --chunk-size option to set size of chunks.Lars Wirzenius4-5/+13
Previous hard-coded value is now just the default. Some day the default should be upped for more speed.
2010-07-04Add new option type for specifying sizes in bytes.Lars Wirzenius2-0/+16
2010-07-04Add --exclude option.Lars Wirzenius1-1/+26
2010-07-04Merge changes to implement checkpoints during backups.Lars Wirzenius12-7/+261
2010-07-04Mark checkpoint generations as such, and report them in generation listing.Lars Wirzenius4-4/+44
2010-07-04Fix parsing of b-less units.Lars Wirzenius2-8/+20
2010-07-04Reset the right vfs written counter when making a checkpoint.Lars Wirzenius1-1/+1
2010-07-04Remove unused variables.Lars Wirzenius1-2/+0
2010-07-04Make checkpoints at suitable intervals.Lars Wirzenius1-1/+9
2010-07-04Add counter of written bytes to VFS.Lars Wirzenius3-1/+14
2010-07-04Add a --checkpoint option.Lars Wirzenius1-0/+10
2010-07-04Add new option type: processed.Lars Wirzenius2-0/+14
2010-07-04Implement ByteSizeParser.Lars Wirzenius3-4/+76
2010-07-04Add tests for a library to parse file sizes.Lars Wirzenius3-0/+87
2010-07-04Merge refactoring change to how files for backups are found.Lars Wirzenius1-46/+71
2010-07-04Read metadata for a file only once.Lars Wirzenius1-11/+10
2010-07-04Change backup plugin to use a generator instead of recursion.Lars Wirzenius1-46/+72
This will be useful later, when I implement checkpoints.
2010-07-04Fix NodeStoreVfs.mkdir to create parents, too.Lars Wirzenius1-1/+1
2010-07-01Store chunk filenames using base 16.Lars Wirzenius1-3/+3
Just 'cause I like them better that way. Well, actually, these are shorter, and if I do an ls, the listing takes up less space and is easier to scan.
2010-07-01Only store metadata if it has changed.Lars Wirzenius1-1/+7
B-tree lookups are massively faster than inserts, so this saves time on incremental backups, since most of the time, the metadata does not actually change.
2010-07-01Remove ROADMAP. All of it has been moved to the bug tracker.Lars Wirzenius1-121/+0
2010-06-30Speed up adding of newly created file to parent.Lars Wirzenius1-4/+7
No need to do listdir to lookup, we can just lookup using key.
2010-06-29Merge changes to speed up backups by better chunk filenaming.Lars Wirzenius1-4/+12
2010-06-29Simplify chunk directory hierarchy.Lars Wirzenius1-3/+2
Modern filesystems can handle large numbers of files in a directory, as long as things don't get excessive, and having many small dirs is wasteful of disk space.
2010-06-29Make put_chunk jump to random id if there's a collision.Lars Wirzenius1-0/+1
2010-06-29Change Store.put_chunk to choose randomly only the first chunk id.Lars Wirzenius1-1/+9
Subsequent chunks get successive ids.
2010-06-29Update setup.py so it installs the script and manpage.version_0_10Lars Wirzenius1-0/+3
2010-06-29Update debian/control for Python policy conformance.Lars Wirzenius1-0/+2
2010-06-29Update NEWS for version 0.10 release.Lars Wirzenius1-280/+4
2010-06-29Debian packaging fixes.Lars Wirzenius3-5/+7
2010-06-29Add Debian packaging.Lars Wirzenius9-2/+67
2010-06-28Only create the directory for chunks if it does not exist.Lars Wirzenius1-1/+2
2010-06-28Change how directory contents are stored, for higher speed.Lars Wirzenius1-26/+7
Previously, the contents of a directory was stored using key(dirname, DIR_CONTENTS, counter)=namehash, where counter was increased for each new entry, and namehash was the 64-bit hash of the child's pathname. The problem this has is that when listing the contents of a directory, we have to look up the pathname for each child, and that's pretty slow. Now we store the contents of the directory as key(dirname, DIR_CONTENTS, namehash)=basename, so that listdir can be done with a single range query. This does not handle hash collisions, but that'll be doable, we just need to make sure we find and store the right namehash when inserting the value, and possibly when accessing the child as well.
2010-06-26Fix Store test to not make assumption about generation ids.Lars Wirzenius1-1/+1
btree does not really make a guarantee about the order in which root ids are assigned, so don't make the test assume it knows what the id of the first root node will be.
2010-06-20Remove unnecessary variables and imports, suggested by pyflakes.Lars Wirzenius6-12/+1
2010-06-19Do not run seivot report, but do save log file.Lars Wirzenius1-2/+2
The report graphs are not useful at this time, and can be re-generated at any point from the .seivot files.
2010-06-17Remove largest benchmark size.Lars Wirzenius1-1/+1
It takes over 30 hours to run on my setup, and does not finish, so that's unusably slow.
2010-06-17Merge changes to put chunks in a hierarchy of directories, rather than a ↵Lars Wirzenius1-4/+11
flat one.
2010-06-17Merged from trunk.Lars Wirzenius10-181/+258
2010-06-14Fix --quiet option in terminal_status_plugin.py.Lars Wirzenius2-8/+14
2010-06-14Prefix error messages with 'Error: '; blackboxtest assumes it.Lars Wirzenius1-1/+1
2010-06-14Remove spurious call to progress-found-file-hook with None filename.Lars Wirzenius1-2/+0
generated by cgit v1.2.1 (git 2.18.0) at 2024-05-15 09:22:40 +0000