12 KiB

Raw Blame History

Configuring the server

The server's config file at $HOME/.config/perkeep/server-config.json is JSON. It can either be in simple mode (for basic configurations), or in low-level mode (for any sort of crazy configuration).

Configuration Keys & Values

Note, if you can't find what you're looking for here, check the API docs: /pkg/types/serverconfig.

auth: the authentication mechanism to use. Example values include:
- none: No authentication.
- localhost: Accept connections coming from localhost. On Linux, this means connections from localhost that are also from the same user as the user running the server.
- userpass:alice:secret: HTTP basic authentication. Username "alice", password "secret". Only recommended if using HTTPS.
- userpass:alice:secret:+localhost: Same as above, but also accept localhost auth
- userpass:alice:secret:vivify=othersecret: Alice has password "secret", but her Android phone can use password "othersecret" to do a minimal set of operations (upload new things, but not access anything).
baseURL: Optional. If non-empty, this is the root of your URL prefix for your Perkeep server. Useful for when running behind a reverse proxy. Should not end in a slash. e.g. https://yourserver.example.com
https: if "true", HTTPS is used.
- httpsCert: path to the HTTPS certificate file. This is the public file. It should include the concatenation of any required intermediate certs as well.
- httpsKey: path to the HTTPS private key file.
- If an explicit certificate and key are not provided, a certificate from Let's Encrypt is requested automatically if the following conditions apply:
  - A fully qualified domain name is specified in either baseURL or listen.
  - Perkeep listens on port 443 in order to answer the TLS-SNI challenge from Let's Encrypt.
- As a fallback, if no FQDN is found, a self-signed certificate is generated.
camliNetIP: the optional internet-facing IP address for this Perkeep instance. If set, a name in the camlistore.net domain for that IP address will be requested on startup. The obtained domain name will then be used as the host name in the base URL. For now, the protocol to get the name requires receiving a challenge on port 443. Also, this option implies https, and that the HTTPS certificate is obtained from Let's Encrypt. For these reasons, this option is mutually exclusive with baseURL, listen, httpsCert, and httpsKey. On cloud instances (Google Compute Engine only for now), this option is automatically used.
identity: your GPG fingerprint. A keypair is created for new users on start, but this may be changed if you know what you're doing.
identitySecretRing: your GnuPG secret keyring file. A new keyring is created on start for new users, but may be changed if you know what you're doing.
listen: The port (like "80" or ":80") or IP & port (like "10.0.0.2:8080") to listen for HTTP(s) connections on.
shareHandler: if true, the server's sharing functionality is enabled, letting your friends have access to any content you've specifically shared. Its URL prefix path defaults to "/share/".
shareHandlerPath: Optional. If non-empty, it specifies the URL prefix path to the share handler, and the shareHandler value is ignored (i.e the share handler is enabled). Example: "/public/".
runIndex: defaults to true. If "false", no search, no UI, no indexing. (These can be controlled at a more granular level by writing a low-level config file)
copyIndexToMemory: defaults to true. If "false", don't slurp the whole index into memory on start-up. Specifying false will result in certain queries being slow, unavailable, or unsorted (work in progress). This option may be unsupported in the future. Keeping this set to "true" is recommended.
sourceRoot: Optional. If non-empty, it specifies the path to an alternative Perkeep source tree, in order to override the embedded UI and/or Closure resources. The UI files will be expected in <sourceRoot>/server/perkeepd/ui and the Closure library in <sourceRoot>/third_party/closure/lib.

Storage options

At least one of these must be set:

memoryStorage: if true, blobs will be stored in memory only. This is generally only useful for debugging & development.
blobPath: local disk path to store blobs. (valid for diskpacked too).
s3: "key:secret:bucket[/optional/dir]" or "key:secret:bucket[/optional/dir]:hostname" (with colons, but no quotes).
b2: "account_id:application_key:bucket[/optional/dir]".
googlecloudstorage: "clientId:clientSecret:refreshToken:bucketName[/optional/dir]"

The s3 storage option's hostname value may be set to use an S3-compatible endpoint instead of AWS S3, such as my-minio-server.example.com. A specific region may be specified by using Low-level Configuration, though the bucket's region will generally be detected automatically.

Additionally, there are two mutually exclusive options which only apply if blobPath is set:

packRelated: if true, blobs are automatically repacked for fast read access.
packBlobs: if true, diskpacked is used instead of the default filestorage. This gives better write throughput, at the cost of slower read access.

For now, if more than one storage option is set, one of them is the primary storage and the other ones are set up as mirrors. The precedence order is the same as the order they are listed above.

Others aren't yet supported by the simple config mode. Patches to pkg/serverinit welcome.

Examples for configuring storage backends

Indexing options

Unless runIndex is set to false, exactly one of these must be set:

sqlite: path to SQLite database file to use for indexing
kvIndexFile: path to kv (https://github.com/cznic/kv) database file to use for indexing
levelDB: path to levelDB (https://github.com/syndtr/goleveldb) database file to use for indexing
mongo: user:password@host
mysql: user@host:password
postgres: user@host:password
memoryIndex: if true, a memory-only indexer is used.

Database-related options

dbname: optional name of the index database if MySQL, PostgreSQL, or MongoDB, is used. If empty, dbUnique is used as part of the database name.
dbUnique: optionally provides a unique value to differentiate databases on a DBMS shared by multiple Perkeep instances. It should not contain spaces or punctuation. If empty, identity is used instead. If the latter is absent, the current username (provided by the operating system) is used instead. For the index database, dbname takes priority.

When using MariaDB or MySQL, the user will need to be able to create a schema in addition to the default schema. You will need grant create, insert, update, delete, alter, show databases on *.* permissions for your database user.

You can use the pk dbinit command to initialize your database, and see dbinit.go and dbschema.go if you're curious about the details.

Publishing options

Perkeep uses Go html templates to publish pages, and publishing can be configured through the publish key. There is already support for an image gallery view, which can be enabled similarly to the example below (obviously, the camliRoot will be different).

"publish": {
  "/pics/": {
    "camliRoot": "mypics",
    "cacheRoot": "/home/joe/var/perkeep/blobs/cache",
    "goTemplate": "gallery.html"
  }
}

See the serverconfig.Publish type for all the configuration parameters.

One can create any permanode with pk-put or the web UI, and set its camliRoot attribute to the value set in the config, to use it as the root permanode for publishing.

One common use-case is for Perkeep (and the publisher app) to run behind a reverse-proxy (such as Nginx), which takes care of the TLS termination, and where therefore it might be acceptable for both perkeepd and publisher to listen for non-TLS HTTP connections. In that case, the app handler configuration parameters should be specified, such as in the example below. In addition, please note that the reverse-proxy should not modify the Host header of the incoming requests.

"publish": {
    "/pics/": {
        "camliRoot": "mypics",
        "cacheRoot": "/home/joe/var/perkeep/blobs/cache",
        "goTemplate": "gallery.html",
        "apiHost": "http://localhost:3179/",
        "listen": ":44352",
        "backendURL": "http://localhost:44352/"
    }
},

Please see the publishing README for further details on how to set up permanodes for publishing, or if you want to make/contribute more publishing views.

Importers

Perkeep has several built-in importers, including:

Feeds (RSS, Atom, and RDF)
Flickr
Foursquare
Picasa
Pinboard
Twitter
Instapaper

These can be setup by visiting the "/importer/" URL prefix path, e.g. http://localhost:3179/importer/

Windows

The default configuration comes with SQLite for the indexer. However, getting mattn go-sqlite3 to work on windows is not straightforward, so we suggest using one of the other indexers, like MySQL.

The following steps should get you started with MySQL:

Dowload and install MariaDB or MySQL (the latter requires .NET).
Edit your server configuration file (if it does not exit yet, running perkeepd will automatically create it):
- Remove the sqlite option.
- Add a dbname option. (ex: "dbname": "camliprod")
- Add a mysql option. (ex: "mysql": "foo@localhost:bar")
Create a dedicated user/password for your mysql server.
Initialize the database with pk: pk dbinit --user=foo --password=bar --host=localhost --dbname=camliprod --wipe

Setting up MongoDB is even simpler, but the MongoDB indexer is not as well tested as the MySQL one.

Low-level configuration

You can specify a low-level configuration file to perkeepd with the same -configfile option that is used to specify the simple mode configuration file. Perkeep tests for the presence of the "handlerConfig": true key/value pair to determine whether the configuration should be considered low-level.

As the low-level configuration needs to be much more detailed and precise, it is not advised to write one from scratch. Therefore, the easiest way to get started is to first run Perkeep with a simple configuration (or none, as one will be automatically generated), and to download the equivalent low-level configuration that can be found at /debug/config on your Perkeep instance.

In the following are examples of features that can only be achieved through low-level configuration, for now.

Replication to another Perkeep instance

If "/bs" is the storage for your primary instance, such as for example:

    "/bs/": {
        "handler": "storage-blobpacked",
        "handlerArgs": {
            "largeBlobs": "/bs-packed/",
            "metaIndex": {
                "file": "/home/you/var/perkeep/blobs/packed/packindex.leveldb",
                "type": "leveldb"
            },
            "smallBlobs": "/bs-loose/"
        }
    },

then instead of "/bs", you can use everywhere else instead in the config the prefix "/bsrepl/", which can be defined as:

    "/bsrepl/": {
        "handler": "storage-replica",
        "handlerArgs": {
            "backends": [
                "/bs/",
                "/r1/"
            ]
        }
    },

where "/r1/" is the blobserver for your other Perkeep instance, such as:

	"/r1/": {
		"handler": "storage-remote",
		"handlerArgs": {
			"url": "https://example.com:3179",
			"auth": "userpass:foo:bar",
			"skipStartupCheck": false
		}
	},

12 KiB Raw Blame History