manta-sync

Rsync style command for Joyent's Manta

Installation

npm install -g manta-sync

Usage

manta-sync ./ ~~/stor/foo

manta-sync requires 2 arguments, the first is a local directory that you would like to sync the contents of into manta. The second is a manta directory that you would like the files to by synced to.

All remote directories will be lazily created for you if they do not exist, relying on the latest manta node module for this behavior.

If you supply -r, manta-sync will work in reverse, pulling files from manta onto your local filesystem.

manta-sync -r ~~/stor/foo ./foo

Usage

$ manta-sync -h
usage: manta-sync [OPTIONS] localdir ~~/remotedir

synchronize all files found inside `localdir` to `~~/remotedir`

examples

  manta-sync ./ ~~/stor/foo
    - sync all files in your cwd to the dir ~~/stor/foo

  manta-sync --dry-run ./ ~~/stor/foo
    - same as above, but just HEAD the data, don't PUT

  manta-sync -r ~~/stor/foo ./bar
    - sync all files from manta in ~~/stor/foo to the local dir ./bar

options:
    -h, --help                          Print this help and exit.
    --version                           Print version and exit.
    -v, --verbose                       Verbose trace logging.

  Manta connection options:
    -a ACCOUNT, --account=ACCOUNT       Manta Account (login name). Environment:
                                        MANTA_USER=ACCOUNT
    --user=USER, --subuser=USER         Manta User (login name). Environment:
                                        MANTA_SUBUSER=USER
    --role=ROLE,ROLE,...                Assume a role. Use multiple times or
                                        once with a list. Environment:
                                        MANTA_ROLE=ROLE,ROLE,...
    -i, --insecure                      Do not validate SSL certificate.
                                        Environment: MANTA_TLS_INSECURE=1
    -k FP, --keyId=FP                   SSH key fingerprint. Environment:
                                        MANTA_KEY_ID=FP
    -u URL, --url=URL                   Manta URL. Environment: MANTA_URL=URL
    -c COPIES, --copies=COPIES          number of copies to make.
    -d, --delete                        delete files on the remote end not found
                                        locally.
    -x ARG, --exclude=ARG               a pattern to ignore when searching the
                                        local filesystem.
    -H HEADER, --header=HEADER          HTTP headers to include.
    -j, --just-delete                   don't send local files, just delete
                                        extra remote files.
    -l, --ignore-links                  ignore symlinks.
    -m, --md5                           use md5 instead of file size (slower,
                                        but more accurate).
    -n, --dry-run                       don't perform any remote PUT or DELETE
                                        operations.
    -p CONCURRENCY, --parallel=CONCURRENCY
                                        limit concurrent operations.
    -q, --quiet                         suppress all output.
    -r, --reverse                       manta to local sync.
    -U, --updates                       check for available updates on npm.

Example

First we'll create a basic directory structure we want to sync to manta

$ mkdir foo
$ touch foo/a foo/b foo/c
$ mkdir foo/d
$ touch foo/d/e
$ ls foo/
a  b  c  d/
$ ls foo/d
e

Now, let's look at the remote end to see what we're dealing with

$ mls ~~/stor
$

Nothing on the remote end yet, let's sync the files up

$ manta-sync foo/ ~~/stor/foo
building local file list...
local file list built, 4 files found

~~/stor/foo/d/e... not found, adding to put list (1/4)
~~/stor/foo/a... not found, adding to put list (2/4)
~~/stor/foo/c... not found, adding to put list (3/4)
~~/stor/foo/b... not found, adding to put list (4/4)

upload list built, 4 files staged for uploading (took 1016ms)

~~/stor/foo/a... uploaded (1/4)
~~/stor/foo/b... uploaded (2/4)
~~/stor/foo/c... uploaded (3/4)
~~/stor/foo/d/e... uploaded (4/4)

4 files (0 bytes) put successfully, 0 files failed to put (took 474ms)

done

All 4 files were uploaded (and their directories created), we can verify this with

$ mls ~~/stor
foo/
$ mls ~~/stor/foo
a
b
c
d/
$ mls ~~/stor/foo/d
e

Now that we are synced up, let's run it again and see what happens

$ manta-sync foo/ ~~/stor/foo
building local file list...
local file list built, 4 files found

~~/stor/foo/a... size same as local file, skipping (1/4)
~~/stor/foo/b... size same as local file, skipping (2/4)
~~/stor/foo/c... size same as local file, skipping (3/4)
~~/stor/foo/d/e... size same as local file, skipping (4/4)

upload list built, 0 files staged for uploading (took 838ms)


done

This time the output is slightly different, because the files were found on the remote end and the have the same size as the local files.

So let's modify a file and rerun the sync

$ echo hello > foo/a
$ manta-sync foo/ ~~/stor/foo
building local file list...
local file list built, 4 files found

~~/stor/foo/c... size same as local file, skipping (1/4)
~~/stor/foo/a... size is different, adding to put list (2/4)
~~/stor/foo/d/e... size same as local file, skipping (3/4)
~~/stor/foo/b... size same as local file, skipping (4/4)

upload list built, 1 files staged for uploading (took 999ms)

~~/stor/foo/a... uploaded (1/1)

1 files (6 bytes) put successfully, 0 files failed to put (took 152ms)

done

manta-sync detected one of the files on the local end was a different size than reported by manta, so it staged it for uploading, and PUT the file.

How

manta-sync works in 4 (optionally 5) stages

1. Find all local files

The local module Finder is used to locate (and stat(2)) all local files, to build a list of files that need to be synced.

If -x or --exclude arguments are supplied, they will be used in this step to filter out the local files found. For instance, --exclude ./.git/ will cause manta-sync to skip all files beginning with .git/.

2. Process each local file, figure out if we need to put a new version into Manta

For each local file found, a corresponding remote manta filename is constructed, and then checked for info (HEAD request) to see if it exists, and what its size is if it is found.

If the file is not found (404 / NotFoundError) it is staged for uploading.

If the file is found, and the size reported by manta is different than the size on the filesystem, it is also staged for uploading. This behavior can be modified with the -m or --md5 switch, which tells manta-sync to use the md5 hash of a file instead of the file size.

3. Upload each file that needs to be uploaded, lazily handling directory creation

For each file that has been staged for uploading, a PUT request is made, and all directories that are needed are created lazily (which may result in more than 1 PUT per file).

If -n or --dry-run is supplied, this step is skipped by just printing what actions would have been taken. Note that during a dry-run, HEAD requests are still made.

4. (optional) Delete files found on the remote end not found locally

If --delete is supplied, a walk of the remote file tree is done and compared against the list of local files from step 1. Every file found on the remote end that is not referenced locally is deleted.

Any files skipped (by --exclude) in the first step will be deleted from the remote end if they are found.

5. Print statistics, clean up

manta-sync prints how many files were uploaded, and how many (if any) files failed to upload. Also, any errors that were encountered are displayed again at the bottom of the output.

Possible Future Features

• count number of HEAD and PUT requests done (for billing purposes)

License

MIT