The Autoupdate Service is part of the OpenSlides environment. It is a http endpoint where the clients can connect to get the actual data and also get updates, when the requested data changes.
IMPORTANT: The data are sent via an open http-connection. All browsers limit the amount of open http1.1 connections to a domain. For this service to work, the browser has to connect to the service with http2 and therefore needs https.
The service needs some secrets to run. You can create them with:
mkdir secrets
printf "password" > secrets/postgres_password
printf "my_token_key" > secrets/auth_token_key
printf "my_cookie_key" > secrets/auth_cookie_key
It also needs a running postgres and redis instance. You can start one with:
docker run --network host -e POSTGRES_PASSWORD=password -e POSTGRES_USER=openslides -e POSTGRES_DB=openslides postgres:13
docker run --network host redis
export DATABASE_PASSWORD_FILE=secrets/postgres_password
export AUTH_TOKEN_KEY_FILE=secrets/auth_token_key
export AUTH_COOKIE_KEY_FILE=secrets/auth_cookie_key
go build
./openslides-autoupdate-service
Make sure the service inside the docker container can connect to the auth
service, postgres and redis, for example with the docker argument --network host
.
docker build . --tag openslides-autoupdate
docker run --network host -v $PWD/secrets:/run/secrets openslides-autoupdate
To restart the service when ever a source file has shanged, the tool CompileDaemon can help.
go install github.com/githubnemo/CompileDaemon@latest
CompileDaemon -log-prefix=false -build "go build" -command "./openslides-autoupdate-service"
The make target build-dev
creates a docker image that uses this tool. The
environment varialbe OPENSLIDES_DEVELOPMENT
is used to use default auth keys.
make build-dev
docker run --network host --env OPENSLIDES_DEVELOPMENT=true openslides-autoupdate-dev
go test ./...
There is a make target, that creates and runs the docker-test-container:
make run-tests
Curl needs the flag -N / --no-buffer
or it can happen, that the output is not
printed immediately.
When the server is started, clients can listen for keys to do so, they have to send a keyrequest in the body of the request. An example request is:
curl -N localhost:9012/system/autoupdate -d '[{"ids": [1], "collection": "user", "fields": {"username": null}}]'
To see a list of possible json-strings see the file internal/keysbuilder/keysbuilder_test.go.
Keys can also defined with the query parameter k
:
curl -N localhost:9012/system/autoupdate?k=user/1/username,user/2/username
With this query method, it is not possible to request related keys.
A request can have a body and the k
-query parameter.
After the request is send, the values to the keys are returned as a json-object without a newline:
{"user/1/username":"value","user/2/username":"value"}
With the query parameter single
the server writes the first response and
closes the request immediately. So there are not autoupdates:
curl -N localhost:9012/system/autoupdate?k=user/1/username&single=1
With the query parameter position=XX
it is possible to request the data at a
specific position from the datastore. This implieds single
:
curl -N localhost:9012/system/autoupdate?k=user/1/username&position=42
Values are updated via redis:
xadd ModifiedFields * user/1/username newName user/1/password newPassword
The data for a projector can be accessed with autoupdate requests. For example use:
curl -N localhost:9012/system/autoupdate -d '
[
{
"ids": [1],
"collection": "projector",
"fields": {
"current_projection_ids": {
"type": "relation-list",
"collection": "projection",
"fields": {
"content": null,
"content_object_id": null,
"stable": null,
"type": null,
"options": null
}
}
}
}
]'
To get all history information for an fqid call:
curl localhost:9012/system/autoupdate/history_information?fqid=motion/42
It returns a list of all changes to the requested fqid. Each element in the list is an object like this:
{
"position": 23,
"user_id": 5,
"information": "motion was created",
"timestamp: 1234567
}
To get the data at a position, use the normal autoupdate request with the
attribute position
. See above.
The autoupdate service provides an internal route to return fields for a defined user.
curl "localhost:9012/internal/autoupdate?user_id=42&k=user/1/username"
It also supports the attributes single=1
and the normal autoupdate body.
The autoupdate services saves how many connections are currently open to each user.
The save interval can be defined with the environment variable
METRIC_SAVE_INTERVAL
. The default is 5 minutes.
The values are saved for each instance of the autoupdate service. So it is possible to access all open connection for every instance of the autoupdate service in the same cloud.
curl "localhost:9012/service/autoupdate/connection_count"
It returns a JSON dictonary like this:
{"0":15,"1":4,"2":3}
The key is a user ID and the value is the amount of currently open connections.
User ID 0
is for connections, that are not logged in (public access). It the
example above, there are 15 open connections from public access, the user with
the ID 1 has 4 open connections and the user with the ID 2 has 3 open
connection.
Users can only access this page if they have the organization management level or higher.
The autoupdate service logs some metric values. The interval can be set with the
environment variable METRIC_INTERVAL
.
The logged metric is a json dictonary like:
{
"connections_longpolling_connected_connections_public_access": 0,
"connections_longpolling_connected_users_average_connections": 3,
"connections_longpolling_connected_users_current": 1,
"connections_longpolling_connected_users_current_local": 1,
"connections_longpolling_connected_users_total": 1,
"connections_longpolling_connected_users_total_local": 1,
"connections_longpolling_current_connections": 3,
"connections_longpolling_current_connections_local": 3,
"connections_stream_connected_connections_public_access": 0,
"connections_stream_connected_users_average_connections": 6,
"connections_stream_connected_users_current": 2,
"connections_stream_connected_users_current_local": 2,
"connections_stream_connected_users_total": 3,
"connections_stream_connected_users_total_local": 3,
"connections_stream_current_connections": 13,
"connections_stream_current_connections_local": 13,
"datastore_cache_key_len": 236478,
"datastore_cache_size": 1722987,
"runtime_goroutines": 68
}
The prefix connections_stream
are for "normal" connections.
connections_longpolling
are for connections, that use the longpolling
fallback.
connections_stream_connected_connections_public_access
: Number of connections from non logged in users (public access) from all autoupdate instances.connections_stream_connected_users_average_connections
: Average connection count for each logged in user (no public access).connections_stream_connected_users_current
: Amount of connected users that have at least one open connection.connections_stream_connected_users_current_local
: Amount of connected users that have at least one open connection of this instance.connections_stream_connected_users_total
: Amount of different users that are currently connected or were connected since the autoupdate service was started.connections_stream_connected_users_total_local
: Same asconnected_users_total
, but only for this instance.connections_stream_current_connections
: Amount of all connections.connections_stream_current_connections_local
: Amount of all connections of this instance.datastore_cache_key_len
: Amount of keys in the cache.datastore_cache_size
: Combined size of all values in the cache.runtime_goroutines
: Current goroutines used by the instance.
The service is configurated with environment variables. See all environment varialbes.
To use a new models.yml update the meta repository in meta
.
Afterwards call go generate ./...
to update the generated files.