StreamBase containers are the building blocks for organizing and connecting multiple StreamBase applications. There is one StreamBase application per container, and each container name serves as a handle for that application's resources. Multiple containers and applications can be run in a single StreamBase Server process, and you can share resources between applications.
Topics:
Containers enable a number of useful features. You can:
-
Dynamically add or remove StreamBase applications from the running StreamBase Server process.
-
Use containers to run multiple applications in a single server process, without necessarily connecting the applications' resources.
-
Connect the output of one StreamBase application to the inputs or outputs of one or more other StreamBase applications, all running within the same server.
-
Establish container connections even for servers running on separate physical servers.
-
Dynamically modify an existing container in a running application, to add or change a container connection, or to start, stop, or pause enqueuing or dequeuing.
-
Specify containers and container connections from the command line, or through an
sbd.sbconfserver configuration file, or with theStreamBaseAdminClientAPI.
The following is an example of command-line container commands:
sbadmin addContainer c1 newapp.sbapp sbadmin addContainer c2 myapp-same-schema.sbapp c2.InputStream=c1.OutputStream
This example defines two containers and associates each with an application. The last
expression specifies that the output of the c1
container is to be consumed by the input stream of the c2 container. Both containers are running in the same
sbd process.
Each container has its own thread of execution. Containers behave much like concurrent (parallel) operators or modules. On a multi-CPU machine, one container does not block the execution of another container.
The following set of conceptual diagrams illustrates the ideas behind containers. The first diagram illustrates a configuration running multiple unrelated StreamBase applications in a single sbd process:
|
In the next diagram, we added two reader applications (B and C) to a running sbd process that was already hosting application A:
|
In the next diagram, we added two writer applications (B and C) to a running sbd process that was already hosting application A. Then we added application D to write its output stream to the output stream of application A.
|
Notes:
-
In the diagrams above, the StreamBase applications are shown for simplicity as having the same components. This is for illustration only; applications in containers are not required to have the same set of components. However, the streams used between applications must have the same schema.
-
If you remove a container and application from a running sbd server process, StreamBase Systems recommends that you remove each one in the reverse order that it was added.
-
You cannot share a Query Table between applications. Each application must use its own Query Table.
In the configuration file for a StreamBase Server process,
the runtime section provides the <application> element. There can be any number of
<application> elements. The application element has two attributes:
-
file- the application's file name, such asmyapp1.ssqlormyapp2.sbapp. -
container- the container's name
The
element can have any number of sub-elements named application. These sub-elements have
container-connectiondest (destination) and source attributes. For example:
<runtime>
<application file="app.ssql" container="A"/>
<application file="app.ssql" container="B">
<container-connection dest="B.InputStream1" source="A.OutputStream1"/>
<container-connection dest="B.InputStream2" source="A.OutputStream2"/>
</application>
</runtime>
Notes:
-
The discussion above does not list all configuration file elements that affect containers. Generate a skeleton configuration file with sbd –s to see the full set of elements.
-
In prior releases, the configuration file used an
application-fileelement, which has been deprecated: the element is still supported, but we recommend that you edit your configuration files to use theapplicationelement instead. -
If you do not define any containers, StreamBase provides a default container, named
.defaultIf you do not specify a container name in a StreamBase command or client, the
defaultcontainer is assumed. For example, the followinglistcommand is run against the default container of the server at the default StreamBase URI:sbc list
To list the entities for a particular container, specify the container name in the StreamBase URI specified with the
-uoption. For example:sbc -u sb://localhost:10000/myContainer list
You can use the StreamBase URI shortcuts to reduce this command to the following:
sbc -u /myContainer list
When an sbd process starts up (with or
without an application), a container named system is
always present and running. The system container has the following output streams:
- error
-
Emits a tuple containing each StreamBase Server runtime error, if any.
- control
-
Emits messages from this StreamBase Server in response to certain system-level events. The control stream is discussed in more detail in Using Control Stream Features.
- stat
-
Emits tuples containing StreamBase Monitor statistics about the running server.
- connections
-
Emits a tuple every time a client connects to or disconnects from this server.
- subscriptions
-
Emits a tuple every time a client subscribes to or unsubscribes from any stream on this server.
StreamBase's formal entity notation is designed to avoid naming
conflicts when you run multiple applications in the same sbd, when you use modules, and when operators are run in
separate threads. StreamBase entity notation can be used to address
application entities when using the sbadmin and sbc commands, or while configuring the runtime section of the server configuration file
(specifically, in the container-connection
element's destination and source attributes). For example, the following command
asks for the XML representation of one particular stream:
sbc describe default.OutputStream1
The entity notation uses the following syntax, with a period separator between each element:
containerName.[moduleRefName...].entityName
- containerName
-
The name of the container hosting the application. Container names are either
system,default, or a name you assigned when creating a container. The following example addresses two operators with the same name in separate containers:container1.ReadOrders default.ReadOrders
When
containerNameis omitted, the container nameddefaultis assumed. Every StreamBase application has a default container named default, which contains the application unless you explicitly create a non-default container. This means you can usually omit thecontainerNameportion of the address unless you are addressing a non-default container, or if a name conflict would otherwise occur. For example, the following commands are equivalent:sbc enqueue default.InputStream1 sbc enqueue InputStream1
- moduleRefName
-
A module that is either referenced explicitly in the application, or referenced implicitly by StreamBase. Implicit references occur when a component runs in one separate or multiple parallel threads. In this case, StreamBase creates a module for each running instance. There can be multiple
moduleRefNamelevels in the address syntax.The following line addresses an operator named
SumOrderswithin a module reference namedMWindowModuleRef1:MWindowModule1Ref1.SumOrders
The next example addresses a group of entities associated with an operator that you have opted to run in a separate thread. In each entity, the prefix references the module that StreamBase creates for the operator's thread.
MyOperator.InputStream1 (input stream) MyOperator.out:MyOperator_1 (output stream) MyOperator.schema:RetrieveLastOrders (schema) MyOperator.MyOperator (operator)
Finally, If you set a component to run in multiple parallel threads, you can address each thread separately. The following addresses two threads for the same operator:
Filter:0.Filter Filter:1.Filter
- entityName
-
The application component. That is, the operator, stream, schema, and so on.
The URI syntax for addressing StreamBase applications allows you to connect to a particular container in a running application. The URI syntax is:
sb://host:port/container-name
For example:
sb://localhost:10000/app2
If you leave off the container-name portion
of the URI, the default container is assumed. Thus, the
following commands are equivalent:
sbc -u sb://localhost:10000/default sbc -u sb://localhost:10000
When you specify a container name in the URI, you can omit the container name from any subsequent entity names. For example, the following commands are equivalent:
sbc -u sb://localhost/app2 enqueue InputStream1 sbc -u sb://localhost enqueue app2.InputStream1
If you start sbd without specifying an
application name, the sbd process
includes a system container, but no default container. In this case, you may see errors when the default
container name is assumed in commands. For example, type the following command in a
terminal window or a StreamBase Command Prompt:
sbd -p 9999
This starts an instance of the server on port 9999 without an application running. In another window, type the following command:
sbc -p 9999 list
The response is:
container system sbc: Container: default does not exist
Some commands that accept StreamBase URIs can specify a comma-separated list of URIs to address two or more StreamBase Servers at once. This feature is used primarily to support a cluster of StreamBase Servers that are using high availability features.
See Multiple URI Syntax for HA Scenarios for a discussion of this subject.
If an sbd process has multiple containers, you can enqueue to or dequeue from multiple containers at the same time. For example, you can dequeue with a command like the following:
sbc dequeue foo.OutputStream1 bar.OutputStream1
If you specify a container in the URI, you can still dequeue or enqueue from different containers. For example:
sbc -u sb://localhost/foo dequeue OutputStream1 bar.OutputStream1
OutputStream1 will be dequeued from the container
foo, while bar.OutputStream1 will be dequeued from the container
bar.
Caution
Because tuples are queued between applications, there is overhead introduced when tuples cross modules. Also note that if the downstream application cannot process tuples as fast as the upstream application can produce them, then the application may become unstable.
There are several ways to add a container to a running sbd process:
-
Using the sbadmin command. The format is:
sbadmin addContainer
container-nameapplication-name.[ssql | sbapp | sbar] -
Using the
StreamBaseAdminClientAPI. -
Specifying containers in the
sbd.sbconffile, with the<runtime> ... <application>element.
You can remove containers from a running sbd server process with the StreamBaseAdminClient API or the sbadmin command. For example:
sbadmin removeContainer container-name
When a container is removed:
-
Enqueuer clients receive an exception:
ContainerManager$NonExistentContainerException
-
Dequeuer clients will get an EOF if they are dequeuing only from that container. If they are connected to multiple containers, then they should continue to receive tuples from that container.
-
Containers that depend on a removed container for input or output will continue to function. Their input or output will disappear.
The sbc list command (and underlying client API) allows you to show all the containers in a running sbd. The following shows example output from sbc list.:
stream .... schema ... operator ... container foo container bar container system
The above example shows three containers. Two are defined by the user, and the third
is provided by StreamBase: foo,
bar, and system. If you
specify a container in the URI, you get a list of the entities for that particular
container.
The sbc describe command shows you the container name and any connections specified when that container was added.
You can use the sbadmin command to manage multiple container-bound applications running in a single sbd process. The relevant sbadmin commands are:
sbadmin {shutdown | suspend | resume} container-name
You can use sbadmin modifyContainer to:
-
add a connection between running containers
-
remove a container connection
-
start, stop, or pause an enqueue to or dequeue from a container
See sbadmin for details on the
modifyContainer command.
The sbmonitor and sbmanager commands can display statistics from containers. While running sbmonitor, press c to toggle the display of container names in the Operator Name column. The following example shows the top portion of a StreamBase Monitor window:
STREAMBASE MONITOR - T+869 sec - press '?' for help Memory: 63.3MB total, 3.1MB used, 60.2MB free, 253.2MB max, Deq: 3 Eng: 1 OPERATOR NAME IN OUT w-us/T %TIME SIZE ______________________________________________________________________________ default.skuGT5000Filter 0 0 N/A 0.00 N/A
See Using StreamBase Monitor and Using StreamBase Manager for more information on the StreamBase monitoring utilities.
Using either the sbadmin addContainer or sbadmin modifyContainer commands, you can dynamically start, stop, and pause enqueuing or dequeuing in containers.
Both commands take the following syntax:
--enqueue=STREAMSTATUS --dequeue=STREAMSTATUS
The values for STREAMSTATUS can be:
-
ENABLED. This is the default.
-
DISABLED. The enqueue or dequeue will actively refuse any enqueue or dequeue requests, and throw an exception.
-
DROP_TUPLES. The enqueue or dequeue will silently drop tuples.
The StreamBase Client API for Java, C++, and .NET supports containers. You can write your own applications to connect to and interact with StreamBase Servers and the containers therein.
See API Guide for more information.
The following StreamBase features are not supported with containers:
-
Multiple applications per container
-
Ad-Hoc queries
