couchsnap
v1.2.1
Published
Very minimal CouchDB snapshot tool
Downloads
14
Readme
couchsnap
Super-simple CouchDB snapshotting tool for creating incremental snapshots of the winning revisions of documents in a Apache CouchDB or Cloudant database.
- winning revisions only of documents and design documents
- no deletions (unless
--deletions true
is supplied). - no attachments
- no conflicts
Installation
Install on your machine (Node.js & npm required):
$ npm install -g couchsnap
Reference
Environment variables:
COUCH_URL
- the URL of your CouchDB service e.g.http://user:[email protected]
COUCH_DATABASE
- (optional) the name of the database to work with e.g.orders
IAM_API_KEY
- (optional) if using IBM IAM for authentication
Usage
--url/-u
- same asCOUCH_URL
environment variable.--database/-db/-d
- same as `COUCH_DATABASE environment variable.--deletions
- include deleted documents in the output. (Default:false
).
Put your CouchDB URL (with credentials) in an environment variable:
$ export COUCH_URL="https://username:[email protected]"
Create a snapshot:
$ couchsnap --db mydb
spooling changes for mydb since 0
Finished fetching changes
mydb-snapshot-2022-11-09T160406.195Z.jsonl
mydb-meta.json
At a later date, another snapshot can be taken:
$ couchsnap --db mydb
spooling changes for mydb since 23597-*****
mydb-snapshot-2022-11-09T160451.041Z.jsonl
mydb-meta.json
Ad infinitum.
You may elect to include deleted documents by adding --deletions
e.g.
$ couchsnap --db mydb --deletions
...
Finding a document's history
For a known document id e.g. abc123
:
grep -h "abc123" mydb-snapshot-*
Restoring a database
Each backup file contains one document per line so we can feed this data to couchimport using its 'jsonl' mode. To ensure that we insert the newest data first, we can concatenate the snapshots in newest-first order:
# list the files in reverse time order, "cat" them and send them to couchimport
ls -t mydb-snapshot-* | xargs cat | couchimport --db mydb2 --type jsonl
# or use "tac" to reverse the order of each file
ls -t mydb-snapshot-* | xargs tac | couchimport --db mydb2 --type jsonl
Some caveats:
- This only restores to a new empty database.
- Deleted documents are neither backed-up nor restored (unless
--deletions
is supplied). - The restored documents will have a new
_rev
token. e.g.1-abc123
. i.e. the restored database would be unsuitable for a replicating relationship with the original database (as they have different revision histories). - Attachments are neither backud-up or restored.
- Conflicting document revisions are neither backed-up nor restored.
- Secondary index definitions (in design documents) are backed up but will need to be rebuilt on restore.
How does it work?
couchsnap
simply spools the changes feed storing the winning revisions of each non-deleted document to a file - one row line per document. The documents are stored without their revision tokens (_rev
) to avoid creating conflicts on restore.
When snapshotting a database, two files are created:
<database>-snapshot-<timestamp>.jsonl
<database>-meta.json
The meta file contains meta data about where the last snapshot left off, so that a new snapshot can resume from the location.
Note: The nature of the CouchDB changes feed means that some snapshots may contain duplicate changes as the changes feed only guarantees "at least once" delivery.