Server-side intake

By the end of this page you will know how to get PST files onto the server without keeping your own computer on, and how to have them assigned and imported automatically.

Three ways, all ending in the usual analysis (a PST becomes valid or invalid):

• the drop folder: copy files into it with rsync, scp, a mount or a file manager;
• server-side fetch: the server downloads the file itself (HTTP(S), SFTP, S3-compatible);
• the browser upload described in Migrate a mailbox, which still works.

Drop folder

The worker watches DATA_DIR/inbox (./data/inbox with the default compose file). Set INBOX_DIR to use another folder, for example a mount.

rsync -av --partial *.pst user@server:/path/to/data/inbox/

• .pst and .ost files are taken, and also .mbox, .mbx, .eml, .msg, .ics, .vcf and .zip files, which become sources for the files connector (see Protocols and file formats). Keep unrelated archives out of the folder. Dotfiles (rsync temporary names are .name.XXXXXX) and names ending in .part, .partial, .tmp, .temp, .crdownload are ignored. Sub-folders are scanned, except hidden ones.
• A file is picked up when its size and modification time have not changed for inbox.settle_seconds (60 by default). A file still being copied is left alone.
• The file is moved into the PST folder: an atomic rename on the same filesystem, otherwise a copy followed by removal of the source. If the worker restarts in the middle, it finishes the move.
• A file with the same content as a PST already known is left in the folder and not registered again. mccb never deletes it.
• The PST row shows source: "inbox".

SMB shares are not read directly. Mount the share on the server (mount -t cifs) and set INBOX_DIR to the mount point.

Fetch from a URL, SFTP or S3

curl -b cookies -X POST http://localhost:22362/api/psts/fetch -H 'Content-Type: application/json' -d '{"kind":"http","url":"https://files.example.com/jane.pst","token":"..."}'

kind	fields
http	url, optionally token (Bearer) or username + password (Basic)
sftp	host, port (22), path, username, password or private_key (+ passphrase), optionally host_key_sha256
s3	endpoint (for example https://s3.eu-west-1.amazonaws.com or a MinIO URL), bucket, key, region (us-east-1), access_key, secret_key

• The transfer runs in the worker. The PST row has status fetching, and received_bytes grows (size is known once the server answers).
• HTTP and S3 resume with a Range request; SFTP resumes from the size already on disk. On a network error the worker retries up to 5 times with a growing pause, resuming each time. After that the status is fetch_failed with the reason in error; POST /api/psts/{id}/retry-fetch tries again.
• POST /api/psts/{id}/cancel stops the transfer and removes the partial file and the row.
• Free space is checked with the same rule as uploads (size plus UPLOAD_DISK_MARGIN).
• Credentials are encrypted at rest and never returned by the API.
• Link-local and cloud metadata addresses are refused; private ranges are allowed. See Security.
• SFTP without host_key_sha256: the first key seen is remembered and then enforced.

Automatic assignment and import

Off by default. In PUT /api/intake/settings:

{"auto_chain": {"enabled": true, "start_import": true, "default_domains": {"1": "example.com"}}}

After a PST is analysed as valid, mccb looks for the mailbox whose address matches the file name:

• [email protected] matches the mailbox [email protected];
• jane.doe.pst matches [email protected] on server 1 when default_domains has "1": "example.com".

If exactly one mailbox matches and it has no PST yet, the PST is assigned to it. With start_import the import is queued too, when the mailbox has a password and no active job. An existing assignment is never changed. When several mailboxes match, nothing is done. The outcome is in the chain field of the PST (assigned, ambiguous, no_match, account_has_pst, already_assigned). Only PSTs analysed after you enable the option are considered.

Pair this with Notifications to be told when everything is done.