Dne 5.4.2017 v 18:13 Cleber Rosa napsal(a):
This is mainly what I dislike about the `get_implementation`. The `get_implementation` should IMO get all the available info about the stream and based on that it should decide which implementation is it. I don't see a point in having dict-like-method to map "remote" to "RemoteStream" class, I can do that by `getattr(streams, type)`.On 04/05/2017 03:29 AM, Lukáš Doktor wrote:Dne 3.4.2017 v 15:48 Cleber Rosa napsal(a):Note: this document can be view in rendered format at: https://github.com/clebergnu/avocado/blob/RFC_multi_stream_v1/docs/source/rfcs/multi_stream.rst =========================== Multi Stream Test Support =========================== Introduction ============ Avocado currently does not provide test writers with standard tools or guidelines for developing tests that spawn multiple machines. Since these days the concept of a "machine" is blurring really quickly, this proposal for Avocado's version of "multi machine" test support is more abstract (that's an early and quick explanation of what a "stream" means). One of the major goal is to be more flexible and stand the "test" (pun intended) of time. This is a counter proposal to a previous RFC posted and discussed on Avocado development mailing list. Many of the concepts detailed here were introduced there: * https://www.redhat.com/archives/avocado-devel/2016-March/msg00025.html * https://www.redhat.com/archives/avocado-devel/2016-March/msg00035.html * https://www.redhat.com/archives/avocado-devel/2016-April/msg00042.html * https://www.redhat.com/archives/avocado-devel/2016-April/msg00072.html Background ========== The prior art that influences Avocado the most is Autotest. The reason is that many of the Avocado developers worked on Autotest before, and both share various common goals. Let's use Autotest, which provided support for multiple machine test support as a basis for comparison. Back in the Autotest days, a test that would spawn multiple machines was a very particular type of test. To write such a test, one would write a **different** type of "control file" (a server one). Then, by running a "server control file" with an **also different** command line application (``autoserv``, A.K.A. ``autotest-remote``), the server control file would have access to some special variables, such as the ``machines`` one. By using an **also different** type of job implementation, the control file could run a given **Python function** on these various ``machines``. An actual sample server control file (``server/samples/reboot.srv``) for Autotest looks like this:: 1 def run(machine): 2 host = hosts.create_host(machine) 3 host.reboot() 4 5 job.parallel_simple(run, machines) Line #5 makes use of the different (server) job implementation to run function ``run`` (defined in line #1) in parallel on machines given by the special variable ``machines`` (made available by the also special ``autoserv`` tool). This quick background check shows two important facts: 1) The functionality is not scoped to tests. It's not easy to understand where a test begins or ends by looking at such a control file. 2) Users (and most importantly test writers) have to learn about different tools and APIs when writing "multi machine" code; 3) The machines are defined outside the test itself (in the form of arguments to the ``autoserv`` command line application); Please keep these Autotest characteristics in mind: Avocado's multi stream test support goals will be presented shortly, and will detail how they contrast with those. Avocado's Multi Stream Test Support Goals ========================================= This is a hopefully complete summary of our goals: 1) To not require a different type of test, that is, allow users to *write* a plain `avocado.Test` while still having access to multi stream goodies; 2) To allow for clear separation between the test itself and its execution environment (focus here on the execution streams environment); 3) To allow increased flexibility by abstracting the "machines" concept into "excution streams"; 4) To allow for even increased flexibility by allowing test writers to use not only Python functions, but other representations of code to be executed on those separate streams; Comparison with prior art ------------------------- When compared to the Autotest version of multiple machine support for tests, Avocado's version is similar in that it keeps the separation of machine and test definition. That means that tests written in accordance to the official guidelines, will not contain reference to the machines ("execution streams") on which they will have portions of themselves executed on. But, a major difference from the Autotest version is that this proposal attempts to provide the **same basic tools and test APIs** to the test writers needing the multiple stream support. Of course, additional tools and APIs will be available, but they will not incompatible with traditional Avocado INSTRUMENTED tests. Core concepts ============= Because the first goal of this RFC is to set the general scope and approach to Multi Stream test support, it's important to properly describe each of the core concepts (usually abstractions) that will be used in later parts of this document. Execution Stream ---------------- An *Execution Stream* is defined as a disposable execution environment, different and ideally isolated from the main test execution environment. A simplistic but still valid implementation of an execution environment could be based on an Operating System level process. Another valid implementation would be based on a lightweight container. Yet another valid example could be based on a remote execution interface (such as a secure shell connection). These examples makes it clear that level of isolation is determined solely by the implementation. .. note:: Even though the idea is very similar, the term *thread* was intentionally avoided here, so that readers are not led to think that the architecture is based on an OS level thread. An execution stream is the *"where"* to execute a "Block Of Code" (which is the *"what"*). Block of Code ------------- A *Block of Code* is defined as computer executable code that can run from start to finish under a given environment and is able to report its outcome. For instance, a command such as ``grep -q vmx /proc/cpuinfo; echo $?`` is valid computer executable code that can run under various shell implementations. A Python function or module, a shell command, or even an Avocado INSTRUMENTED test could qualify as a block of code, given that an environment knows how to run them. Again, this is the *what* to be run on a "Execution Streams" (which, in turn, is *"where"* it can be run). Basic interface =============== Without initial implementation attempts, it's unreasonable to document interfaces at this point and do not expect them to change. Still, the already existing understanding of use cases suggests an early view of the interfaces that would be made available. Execution Stream Interface -------------------------- One individual execution stream, within the context of a test, should allow its users (test writers) to control it with a clean interface. Actions that an execution stream implementation should provide: * ``run``: Starts the execution of the given block of code (async, non-blocking). * ``wait``: Block until the execution of the block of code has finished. ``run`` can be given a ``wait`` parameter that will automatically block until the execution of code has finished. * ``terminate``: Terminate the execution stream, interrupting the execution of the block of code and freeing all resources associated with this disposable environment The following properties should be provided to let users monitor the progress and outcome of the execution: * ``active``: Signals with True or False wether the block of code given on ``run`` has finished executing. This will always return False if ``wait`` is used, but can return either True or False when running in async mode. * ``success``: A simplistic but precise view of the outcome of the execution. * ``output``: A dictionary of various outputs that may have been created by ``run``, keyed by a descriptive name. The following properties could be provided to transport block of code payloads to the execution environment: * ``send``: Sends the given content to the execution stream environment.This is ambitious. Arbitrary code chunks have dependencies, be it binary, script, module, whatever... But sure, each implementation (Bash, PythonModule, PythonCode) can support various means to allow deps.Writing many different implementations is indeed ambitious. The major goal here, though, is to define a sane interface that works for the basic and most urgent use cases while still not locking out doors to future needs. Basically, making sure we don't loose the investment here and rewriting large portions of Avocado just because we now find that running, say, Ansible playbooks, in different streams is now a very important thing for our tests.Anyway for the python script I'd recommend looking at `avocado-vt/virttest/remote_commander` which is a generic interface to execute tasks remotely to get the sight of complexity.Yep, I'm aware of it, but thanks for the pointer since it does indeed make sense to keep it in mind.Block of Code Interface for test writers ---------------------------------------- When a test writer intends to execute a block code, he must choose from one of the available implementations. Since the test writer must know what type of code it's executing, the user inteface with the implementation can be much more flexible. For instance, suppose a Block Of Code implementation called ``PythonModule`` exists. This implementation would possibly run something like ``python -m <modulename>`` and collect its outcome. A user of such an implementation could write a test such as:: from avocado import Test from avocado.streams.code import PythonModule class ModuleTest(Test): def test(self): self.streams[1].run(PythonModule("mymodule", path=["/opt/myproject"])) The ``path`` interface in this example is made available and supported by the ``PythonModule`` implementation alone and will not be used the execution stream implementations. As a general rule, the "payload" should be the first argument to all block of code implementations. Other arguments can follow. Another possibility related to parameters is to have the Avocado's own test parameters ``self.params`` passed through to the block of code implementations, either all of them, or a subset based on path. This could allow for example, a parameter signaling a "debug" condition to be passed on to the execution of the block of code. Example:: from avocado import Test from avocado.streams.code import PythonModule class ModuleTest(Test): def test(self): self.streams[1].run(PythonModule("mymodule", path=["/opt/myproject"], params=self.params)) Block of Code Interface for Execution Stream usage -------------------------------------------------- Another type of public interface, in the sense that it's well known and documented, is the interface that Execution Stream implementations will use to interact with Block of Code implementations. This is not intended to be used by test writers, though. Again, it's too early to define a frozen implementation, but this is how it could look like: * ``send_self``: uses the Execution Stream's ``send`` interface to properly populate the payload or other necessary assets for its execution. * ``run``: Starts the execution of the payload, and waits for the outcome in a synchronous way. The asynchronous support is handled at the Execution Stream side. * ``success``: Reports the positive or negative outcome in a simplistic but precise way. * ``output``: A dictionary of various outputs that may be generated by the execution of the code. The Execution Stream implementation may merge this content with its own ``output`` dictionary, given an unified view of the output produced there. Advanced topics and internals ============================= Execution Streams ----------------- An execution stream was defined as a "disposable execution environment". A "disposable execution environment", currently in the form of a fresh and separate process, is exactly what the Avocado test runner gives to a test in execution. While there may be similarities between the Avocado Test Process (created by the test runner) and execution streams, please note that the execution streams are created *by* one's test code. The following diagram may help to make the roles clearer:: +-----------------------------------+ | Avocado Test Process | <= created by the test runner | +-------------------------------+ | | | main execution stream | | <= executes your `test*()` method | +-------------------------------+ | | | execution stream #1 | | <= initialized on demand by one's | | ... | | test code. utilities to do so | | execution stream #n | | are provided by the framework | +-------------------------------+ | +-----------------------------------+ Even though the proposed mechanism is to let the framework create the execution lazily (on demand), the use of the execution stream is the definitive trigger for its creation. With that in mind, it's accurate to say that the execution streams are created by one's test code (running on the "main execution stream"). Synchronous, asynchronous and synchronized execution ---------------------------------------------------- As can be seen in the interface proposal for ``run``, the default behavior is to have asynchronous executions, as most observed use cases seem to fit this execution mode. Still, it may be useful to also have synchronous execution. For that, it'd be a matter of setting the ``wait`` option to ``run``. Another valid execution mode is synchronized execution. This has been thoroughly documented by the previous RFCs, under sections named "Synchronization". In theory, both synchronous and asynchronous execution modes could be combined with a synchronized execution, since the synchronization would happen among the execution streams themselves. The synchronization mechanism, usually called a "barrier", won't be given too much focus here, since on the previous RFCs, it was considered a somehow agreed and understood point. Termination ----------- By favoring asynchronous execution, execution streams need to also have a default behavior for handling termination of termination of resources. For instance, for a process based execution stream, if the following code is executed:: from avocado import Test from avocado.streams.code import shell import time class MyTest(avocado.Test): def test(self): self.streams[0].run(shell("sleep 100")) time.sleep(10) The process created as part of the execution stream would run for 10 seconds, and not 100 seconds. This reflects that execution streams are, by definition, **disposable** execution environments. Execution streams are thus limited to the scope of one test, so implementations will need to terminate and clean up all associated resources. .. note:: based on initial experiments, this will usually mean that a ``__del__`` method will be written to handle the cleanup.I'd suggest the runner should after the test execution run: for stream in test_instance.streams: stream.close() which should explicitly take care of closing the streams (and terminating it's processes) as `__del__` might not be executed on all occasions and I'd suggest doing this during/after `tearDown`.Right. Let's keep that in mind. At the end of the day, it's an implementation level detail, and as long as the cleanup is tested and works, we're good to go.Avocado Utility Libraries ------------------------- Based on initial evaluation, it looks like most of the features necessary to implement multi stream execution support can be architected as a setto implement multi stream execution support can be __designed__ as a setof utility libraries. One example of pseudo code that could be possible with this design:: from avocado import Test from avocado.streams import get_implementation from avocado.streams.code import shell class Remote(Test): def test_filtering(self): klass = get_implementation("remote")
So as mentioned later I'd suggest creating `get_stream` method which would guess the most appropriate implementation based on the input and report already initialized stream.
This is not really about the automatic creation. As for today it's impossible to map tests-to-variants and I saw some people are actually writing tests which depends on certain path so they can supply different params to different tests:Well this is not really scalable. Imagine that in one execution you want to use local, in next remote and then docker container. In your example you'd have to change the source code of the test. How about this: get_implementation(params.get("first_stream")) get_implementation(params.get("second_stream"))Your suggestion is really a "how to write an Avocado test". Yes, we absolutely should use parameters in tests, and that's why the "klass" parameters come from parameters. The use of "remote" here is just to make things clearer with regards to how one would refer to implementations (by name).where: first_stream = "localhost" second_stream = "test:[email protected]" third_stream = "docker://create=yes,image=fedora,remove=always" fourth_stream = "libvirt://create=no,domain=test_machine,start=yes" ...These URLs are just painful to read IMO.Another option would be to allow `path` of the Avocado Test `params` and the class would get the details from the provided params+path. The cons would be it'd be hard to use anything but `params` for that: streams: first: type: localhost second: hostname: 192.168.122.10 user: test password: 123456 third: type: docker create: yes ...I like this structure better.get_implementation(self.params, "/streams/first/*") => would use `params.get(..., "/streams/first/*")` to get all necessary parametersThe idea of `get_implementation` is pretty simple: get an Execution Stream implementation by its name. What you're describing here (with regards to getting the "right" parameters and instantiating/activating the execution stream easily/automatically) is exactly what I propose later.
boot:
timeout: 10
check:
timeout: 1
shutdown:
timeout: 5
and in boot.py:
timeout = self.params.get("timeout", "*/boot/*")
in check.py:
timeout = self.params.get("timeout", "*/check/*")
...
So the comment above tried to describe how one would create descriptions
of streams on different places and be still able to get them in various
tests from different locations. Imagine in test1.py:
get_implementation(self.params, "*/test1/streams/first/*")
in test2.py:
get_implementation(self.params, "*/test2/streams/server/*")
and so on. The automatic way would be shared for all tests, while this
would allow the same but using different locations. Anyway this is
really painful and I'd prefer using the single-line definition, or the
list of tuples as described later...
Yes, by `get_implementation("docker://create=yes,image=fedora,remove=always") or `get_implementation(host=host, username=username, ...)`, which returns already initialized object of the detected type. Alternatively the user would import directly the specific class and use `DockerStream(image="fedora")`.if klass is not None: stream = klass(host=self.params.get("remote_hostname"), username=self.params.get("remote_username") password=self.params.get("remote_password")) cmd = "ping -c 1 %s" % self.params.get("test_host_hostname") stream.run(shell(cmd))I do like the rest of the example (only the klass would be already initialized by the `get_implementation`.This conflicts with the idea of "utility library to be used by test developer that wants full control of the individual creation and use of the execution streams". I mean, if some implementations are made available at the `avocado.utils` namespace, then they could even be referred to directly by module/name.
Again, the better name for `get_implementation` in this sense would be `get_stream`.
Again, since we're talking about making whatever makes sense in the
utility namespace, there could a function such as
"get_stream_parameters(name)" that would return the parameters to an
Execution Stream, that is:
stream_name = "server"
path = "/avocado/streams/%s/*" % stream_name
impl = self.params.get("type", path=path)
stream = get_implementation(impl)(**get_stream_parameters(path))
Actually it should probably be `self.streams.get_stream()` as we do want to register it in the `Streams` to get it cleaned automatically. This also works for libraries as you'd first initialize `Streams` object and then do the `get_stream()`, `close()` and so on directly on the `Streams` object.
Please note that this is not the intended end result of this proposal, but a side effect of implementing it using different software layers. Most users should favor the simplified (higher level) interface. Writing a Multi-Stream test =========================== As mentioned before, users have not yet been given tools **and guidelines** for writing multi-host (multi-stream in Avocado lingo) tests. By setting a standard and supported way to use the available tools, we can certainly expect advanced multi-stream tests to become easier to write and then much more common, robust and better supported by Avocado itself. Mapping from parameters ----------------------- The separation of stream definitions and test is a very important goal of this proposal. Avocado already has a advanced parameter system, inof this proposal. Avocado already has __an__ advanced parameter system, inwhich a test received parameters from various sources.The most common way of passing parameters at this point is by means of YAML files, so these will be used as the example format.Well this might be quite hard to understand, how about just saying: "Avocado supports test parametrisation via Test Parameters system and the most common way is to use a YAML file by using `yaml_to_mux` plugin."Parameters that match a predefined schema (based on paths and node names) will be by evaluated by a tests' ``streams`` instance (available as ``self.streams`` within a test). For instance, the following snippet of test code:: from avocado import Test class MyTest(Test): def test(self): self.streams[1].run(python("import mylib; mylib.action()")) Together with the following YAML file fed as input to the parameter system:: avocado: streams: - 1: type: remote host: foo.example.comThis is currently not supported by our yaml parser as any dictionary is mapped to multiplex structure and I'm not sure it'd be possible (in a sane manner) to treat dict inside lists differently. Anyway as I mentioned earlier we could use:Oops, I may have used an incorrect syntax or idea (or both).avocado: streams: 1: ssh://foo.example.com or: avocado: streams: 1: type: remote host: foo.example.comThis one looks good IMO. I'm all in for more explicitly naming **when** you go to the lengths of defining them.
Yes, plus we do use the OrderedDicts so we can still let people use:
self.streams[0:4]
together with:
self.streams["server"]
just beware the:
self.streams["1"]
is different then:
self.streams[1]
and the node-names (therefor even the 1: in streams) becomes string in
Avocado, so it actually works well and in the example above the:
self.streams["1"] == self.streams[0]
or: avocado: streams: - type: remote host: foo.example.com Another thing is I'd probably prefer names to ints so "1" or "server" or "worker1" etc, which goes nicely with the first 2 examples. The last example goes well with indexes, but it starts with 0 (which would be my recommendation anyway if we decided to go with indexes).I think I mentioned somewhere "if only integers"... I actually share your fondness of names. I did not say it explicitly, but I think we can support both, as the slicing examples make a lot of sense to me. But, the slicing examples can be expanded to its own dialect, such as supporting regexes. For instance, `self.streams["client-\d+"]` makes a lot of sense IMO.
Yes, this makes sense and is quite simple to do...
Yes, that is the point. `__len__` reports the "defined", not necessarily "initialized" streams.Would result in the execution of ``import mylib; mylib.action()`` in a Python interpreter on host ``foo.example.com``. If test environments are refered to on a test, but have not been definedIf test environments are __referred__ to on a test, but have not been definedOK, thanks.in the outlined schema, Avocado's ``streams`` attribute implementation can use a default Execution Stream implementation, such as a local process based one. This default implementation can, of course, also be configured at the system and user level by means of configuration files, command line arguments and so on. Another possibility is an "execution stream strict mode", in which no default implementation would be used, but an error condition would be generated. This may be useful on environments or tests that are really tied to their execution stream types.I'd solve this by supporting `__len__` where `len(self.streams)` should report number of defined streams. Note the number of defined streams changes based on how many streams are defined __OR__ used by the test. So: avocado: streams: - 0: - 1: len(self.streams) => 2 self.streams[5].run(cmd) len(self.streams) => 6 where the streams 2-5 are the default streams.I have mixed feelings here. In my understanding, all of the streams are created on demand, that is, when they're used. A configuration that defines thousands of them will not cause an empty test (think of `passtest.py`) to initialize them.
Adding streams dynamically is definitely must-have-feature and one note the code above would actually only have the `streams[5]` initialized, the rest would wait for the first usage.
Sure, another possibility would be to support `streams.defined_streams` but I think the `__len__` would make more sense. It's similar to a list:But, the meaning of `len(self.streams)`, that is, defined or initialized, is something we can further discuss later.
streams = list(params.get("streams", "/avocado/*")
len(streams) == 2
streams.extend([2,3,4,5])
len(streams) == 6
Intercommunication Test Example ------------------------------- This is a simple example that exercises the most important aspects proposed here. The use case is to check that different hosts can communicate among themselves. To do that, we define two streams as parameters (using YAML here), backed by a "remote" implementation:: avocado: streams: - 1: type: remote host: foo.example.com - 2: type: remote host: bar.example.com Then, the following Avocado Test code makes use of them:: from avocado import Test from avocado.streams.code import shell class InterCommunication(Test): def test(self): self.streams[1].run(shell("ping -c 1 %s" % self.streams[2].host)) self.streams[2].run(shell("ping -c 1 %s" % self.streams[1].host)) self.streams.wait() self.assertTrue(self.streams.success)Brainstorming here, how about letting `wait` raise exception when it fails unless we use `wait(ignore_failure)`. The exception would contain all the information so it'd be THE exception which failed the test?Yep, this is a valid question. I think the answer will depend on how much we want the **test** result to be bound to what happens on the streams. Right now it's obvious that I decided to keep them pretty much separate.
Sure, the more I think about it I also share your vision.
Oh you mean that the `synchronized=True` means it'd make a `barrier` mechanism available between those two streams and you'd be able to use this barrier from the streams? That would be rather limiting as you might want to create multiple barriers between several streams. I think you should re-describe your synchronization here as well as the structure changed a bit...As for the `streams.success`, I guess it'd be a property, which would go through all streams results, and report `any(_.failure for _ in self.streams)`, right?Exactly.The ``streams`` attribute provide a aggregated interface for all the streams. Calling ``self.streams.wait()`` waits for all execution streams (and their block of code) to finish execution. Support for slicing, if execution streams names based on integers only could be added, allowing for writing tests such as:: avocado: streams: - 1: type: remote host: foo.example.com - 2: type: remote host: bar.example.com - 3: type: remote host: blackhat.example.com - 4: type: remote host: pentest.example.com from avocado import Test from avocado.streams.code import shell class InterCommunication(Test): def test(self): self.streams[1].run(shell("ping -c 1 %s" % self.streams[2].host)) self.streams[2].run(shell("ping -c 1 %s" % self.streams[1].host)) self.streams[3].run(shell("ping -c 1 %s" % self.streams[1].host)) self.streams[4].run(shell("ping -c 1 %s" % self.streams[1].host)) self.streams.wait() self.assertTrue(self.streams[1:2].success) self.assertFalse(self.streams[3:4].success)As mentioned earlier I'd prefer names to indexes, anyway I see the indexes useful as well. How about supporting a name or index?Yep, also thought of that.As for the slices, I'd prefer list-like slice to stream-like slices as it'd be more natural to me to interact with a list of individual streams rather than a Stream object with a limited subset of streams. Anyway that's a matter of taste and I can definitely live with this as well.See my previous comments.Now about this example, it's really limited. Again you are hard-coding the scenario and changing it is really complicated. I'd prefer something like: self.streams[0].run(server_cmd) self.streams[1:].run(contact_server_cmd) self.assertTrue(self.streams.success)Real tests will probably (hopefully) use better (symbolic) names. The goal here is to focus on the mechanisms, which on yours and on my version are identical.Support for synchronized execution also maps really well to the slicing example. For instance, consider this:: from avocado import Test from avocado.streams.code import shell class InterCommunication(Test): def test(self): self.streams[1].run(shell("ping -c 60 %s" % self.streams[2].host) self.streams[2].run(shell("ping -c 60 %s" % self.streams[1].host)) ddos = shell("ddos --target %s" self.streams[1].host) self.streams[3:4].run(ddos, synchronized=True) self.streams[1:2].wait() self.assertTrue(self.streams.success) This instructs streams 1 and 2 to start connectivity checks as soon as they **individually** can, while, for a full DDOS effect, streams 3 and 4 would start only when they are both ready to do so.OK so this is about before-start-synchronisation. Well again, I'm not much fond of boundling the streams so I'd prefer allowing to define the workload (which returns when the workload is ready), trigger it (which triggers it and reports immediately) and then wait for it. The difference in usage is: self.streams[3].run(ddos, stopped=True) self.streams[4].run(ddos, stopped=True) self.streams[3].start() self.streams[4].start() The result is the same (unless you create processes per each stream and in `self.streams[3:4]` you use signals to synchronize the execution) but it allows greater flexibility like synchronizing other tasks then just streams... Or we can create methods `establish(cmd)`, `start()` and `wait()` which might better describe the actions.I think you missed my point here. Streams #3 and #4, in my example, wait for *each other*. Using a mechanism such as barriers.
In my RFC I basically passed the information about barriers by test params, which is obviously not available here as the Code could be anything. Anyway a barrier requires:
1. name 2. server 3. number of clientsWhen creating synchronized streams by a slice, you have all of these available, but it might not be always like that and you might want to define different groups of barriers, for example:
ssh:
start_ssh()
barrier("ssh_started", worker, 2)
barrier("worker_finished", main, 3)
http:
start_http()
barrier("http_started", worker, 2)
barrier("worker_finished", main, 3)
worker:
barrier("ssh_started", worker, 2)
barrier("http_started", worker, 2)
connect_to_both()
barrier("worker_finished", main, 3)
Here you see 3 streams, they use 3 different barriers, 2 are
stream->stream and 1 is stream->main. You might ask in this simple
example why would you care about where the synchronization happens as
you could always use `main` (which is the main Avocado host) but imagine
testing private network between the 2 interfaces which might result in
something like:
stop_eth0()
# now main can't communicate with this stream
start_ssh()
barrier("ssh_started", worker, 2)
barrier("worker_finished", worker, 2)
start_eth0()
# now this stream reports the status to main
where temporarily the connection stream->main is impossible. This was
supported and used in Autotest.
Now the tricky part, how to support it. In the original RFC I used params to do that, which works well but is quite demanding on the test writer (especially with dynamic number of streams). I thought about it and one way is to create a streams representation and pass it to all streams. Now this is simple for pre-defined streams, but for ad-hock streams it would require either post-execution synchronization (which is available for python processes, but unless we do some master-slave architecture for bash or other scripts, it's not possible), or we could allow streams preparation before the actual execution (something like the `establish(cmd)` and `start()`. Let's use the 3 tests I described earlier and write a main test for it, deliberately using ad-hock streams rather then pre-defined configuration:
no_workers = len(self.streams) # all pre-defined streams are workers
assert no_workers > 0, "Not enought workers to run this test"
# define servers
server_types = params.get("server_types", default=["ssh", "http"])
for type in server_types:
self.streams.get_stream(stream_definition % type)
for stream in self.streams:
stream.start() # which forwards the definition of all defined
streams
where the streams definition is something like `Streams` only with static data, allowing:
streams.barrier(name, no_clients) => contact main test to enter the barrier streams[0].barrier(name, no_clients) => contact stream0 to enter the barrier len(streams["client.*"]) => get number of "client.*" matching streams (to be able to calculate how many clients should enter the barrier)
...
Basically from the Code worker I expect it to establish the connection
on `get_stream()`, get some info like `ip_addr` and so on. Then on
`start()` I expect it to get the important info about streams and send
that along with the command where it is executed (for BASH as env
variables or a file stored somewhere) so the executed stream knows what
is going on around.
Quite important note I don't expect all of this to do for all streams as quite often no synchronization is needed, sometimes stream->main synchronization is enough and sometimes all_streams->main synchronization is needed. All of these can be treated as a special situations and could be way simpler to develop, anyway the ultimate goal should be the free-flexibility in synchronization between all streams and our solution should scale to that (yes, as in my previous PR we can leave it up to the test writers and only allow passing params to the streams - env-variables, dict, ...).
PS: The stream->main sync should actually be as simple as waiting till the main test reports `self.streams[0].barrier("name")`, which puts additional requirements on the main test's code, but is really simple to develop. The only reason why we need worker->worker (or worker->some_server) synchronization is when we start disabling network.
Sure, I becoming to like even the `Streams` object with slicing support, basically I see this as an `avocado.utils` library which would be (lazily) available as `self.streams` in each test along with a first-time initialization of streams from `self.params` (during the lazy init).Feedback and future versions ============================ This being an RFC, feedback is extremely welcome. Also, exepect new versions based on feedback, discussions and further development of the ideas initially exposed here.Overall I like it, I don't really see it as a counter-proposal as I think they are pretty similar, only you defined the block-of-code to be more generic and refined the high-level interface.I guess it's a good thing that you like it and that this is just a "only" (small and simple) kind of thing. I guess my goal was attained at some level :).
I'm a bit afraid regarding the Code blocks as I think simple bash scripts can be better dealt with by aexpect-like libraries, python ones with python's distributed computing (yes, in python2 there is no barrier mechanism, but we can add utils for that), but for the complex tasks I see the benefit.
I see some issues regarding the synchronization and flexibility as you are focused on rather simple example. I remember some multi-host tests from Autotest which disabled/hanged networks during the test executions and I believe we ought to support it as well. We can leave the params passing on the test writers, but maybe we can pass the definitions automatically.
Complex tasks are basically all the written tests with setUp test and tearDown (note the stream does not know about them, it's just the inner structure which results to success/failure including cleaned environment). One benefit in being able to run these is the code-reusability and the other is uniform way of writing units.As for the implementation I think the `self.stream` description and all the details about comes a bit too early. I'd start with the low level interface, which is a library allowing to create `Streams()`, defines `Code()` and `Stream()` objects independently on Avocado and later when we see the usage I'd standardized the usage and embedded it into the `Test` class. Anyway I know we don't share this vision and I'm fine with doing it the other way around as the result is the same, only we might found some limits later which might be hard to solve in the current schema. But based on what I know from virtualization this (together with the barrier synchronization) should be enough to support the tests we know from Autotest which is a good start.Sure. I mentioned that it's impossible to define a "freeze" of any sort on the interfaces. Still, talking about them, helps to shape the features they may have and how they'd be used. But the most important thing here is your acknowledgment that this seems to fit the needs we have on virtualization tests.Last remark regarding my and your RFC is that I deliberately defined the block-of-code like (not as) Test, because executing scripts is possible nowadays via `aexpect`, `Remoter`, `remote_commander` or other standard python libraries, but combining existing tests and get not just theRight, but in no standard and supported way. This is one of the major goals here. The very first line in this RFC is: "Avocado currently does not provide test writers with standard tools or guidelines for developing tests that spawn multiple machines."executed results but also to gather remote environment and so on, that I see beneficial. Anyway if I understand this implementation correctly it'd be possible to create `AvocadoTest` inherited from `Code` which would allow such executions and the `Streams` could support environment gathering (optionally) and that is all I care about. For simpler stuffRight.I'd simply use `aexpect` (as I personally like it a lot) and I know of people who use `remote_commander` to synchronize and distribute tasks across multiple machines in `avocado-vt` so I assume they'll stick to their working solution as well. I see the `Streams` as a library toOur goal is to come up with useful innovations that will motivate users to adopt them, so this is a bit negative. Unless you don't really believe in the value of this proposal, I would expect the opposite attitude.support complex tasks, not just a simple command execution, even though some abstraction might be useful.Can you describe complex tasks? I imagine that the ideas your have are based on "interacting with a remote shell/machine/console/application" as aexpect and other tools allow, right?
You can do most of this via some interactive shell (as a lot of framework does) but when the test itself starts having many lines embedding it into another one makes it harder to follow.
Yep, I agree. I'm saying that for the simple blocks of code (like shell script) this could be actually a better technology. For a binary execution I'd chose Remoter. But for the complex tasks which includes several lines of preps and some cleanups I'd definitely want something like this RFC. This note here was more a brainstorm about a possible improvement of aexpect (and Remoter-like implementations) which should probably share the same definition like streams in order to be able to learn it once and re-use it everywhere (yes, I do think all of these could share eg. the connection-code, container-creation-code and so on..)Actually now after the summary I noticed that what I'd really need is either the `AvocadoTest`-like command to be able to combine basic tests into a complex scenarios and then I'd need a unique way of interacting with different streams. So what I'd probably need is an `aexpect`-concentrator which would allow me to ask for a session based on the description: aexpect.RemoteShellSession(url) where `url` is something like: "localhost" "test:[email protected]" "docker://create=yes,image=fedora,remove=always" "libvirt://create=no,domain=test_machine,start=yes" which would establish the connection (creating the container/vm first if asked for) and than I'd interact with it as with other `aexpect.ShellSession` (therefor not just a single command, but full expect-like behavior)OK, this matches my previous comment. Yes, what you're describing is probably not something this RFC contemplates, but at least parts of could be common for both use cases.
Btw for the `Avocado-test`-like code-block (if developed as it can simply be executed as s standalone script) I'd like to see sharing the sources with the remote-test-runner, so most of the Avocado sources would actually benefit from a clear reusable interfaces.
Well they are utils so I don't see a point in propagating them in `Test`. Anyway yes, we should share the code of this RFC and then expand our documentation to help users to pick the right technology for their use-case and they all should be similar to use with some benefits regarding their usages.This last note is probably outside the scope of multi-stream tests and could (hopefully should) be implemented in parallel to serve different purpose. Anyway with this in mind I don't see much point in having `Bash` or `PythonModule`-like code-blocks in parallel test execution and I'd only focus on the full-blown complex parallel tasks.Sure, something like `self.interactives` available at the test level, working similar to the streams could be a nice addition here.
Yes, as mentioned earlier we are (probably) on the same boat, the main concern here is regarding the synchronization...Anyway, hopefully this feedback is understandable, I have been writing it for 2 days so feel free to ask for some hints ... LukášYes, I think I understood all your points. Let me know if my response was clear enough.
And this for the feedback!
signature.asc
Description: OpenPGP digital signature
