Re: Invoking actions via websockets
Re-posting from the external issue... This would be an excellent feature of the API Gateway . Websocket support is something developers ask about a lot... - We would need to define the error handling process when the the triggers are unavailable or rate-limited. How would we signify to a client that a message wasn't delivered? Another (more difficult) thing to consider is how to make it possible for messages to be sent to listening clients outside of responding to an incoming event. If we take one of the most common examples for a websocket application, a chat application, how could we enable this feature? Ideally, an action could call the API Gateway service with a message to pass to one or more clients currently connected to pass messages that arrived from other sources. On Fri, 26 Oct 2018 at 22:58, Carlos Santana wrote: > Hi Matt > > Do you require changes in the controller, invoker, or runtimes for this? > > Or you think first implementation would be all apigateway side holding the > connecting open as a websocket server to the client, as long the client > keeps sending messages/frames, the apigateway calls web actions with a > single connection to one of the edge/controllers with a keep alive. > > > -- Carlos > > On Fri, Oct 26, 2018 at 2:10 PM Matt Hamann > wrote: > > > I just opened an issue[1] in the `apigateway` repo to discuss the > > ability for the gateway to handle websocket connections from a client > > that could invoke a whisk action `onMessage`. > > > > I'd like to get feedback from the community at large as to this > > proposal. Interested in your thoughts, ideas, etc. > > > > [1] https://github.com/apache/incubator-openwhisk-apigateway/issues/325 > > > > -Matt > > matthew.ham...@gmail.com > > > -- Regards, James Thomas
Re: A question about concurrency in OpenWhisk
Ok so my assumption is correct and we can release the runtime. The runtime can be extended to handle a pool of underlying action distributing the requests to the available handler. There will be separated channels for each action, without overlap. In the current shape with concurrent actions there can be an "out-of-band" behavior. -- Michele Sciabarra mich...@sciabarra.com - Original message - From: Rodric Rabbah To: dev@openwhisk.apache.org Subject: Re: A question about concurrency in OpenWhisk Date: Mon, 29 Oct 2018 07:09:02 -0400 Today there is at most one activation at a time per container we. A PR from Tyson will allow a runtime to opt into intro container concurrency and so the invoker will send up to N invocations to that the container when possible. https://github.com/apache/incubator-openwhisk/pull/2795 -r > On Oct 29, 2018, at 7:06 AM, Michele Sciabarra wrote: > > I would like to understand a bit more how concurrency is handled by the > Invoker. > > If there are, say, 2 concurrent requests, can the invoker send the second > request to the same runtime BEFORE it has completed the first request? I have > seen some tests related in the nodejs runtime but none in the basic tests. > > Also, Carlos mentioned the need to increase the concurrency in the same > runtime. It can be done but I need to know the requirements. So far I assumed > that until a request is not completed, the invoker will spawn another runtime > to serve them. > > > > -- > Michele Sciabarra > mich...@sciabarra.com
Re: A question about concurrency in OpenWhisk
Today there is at most one activation at a time per container we. A PR from Tyson will allow a runtime to opt into intro container concurrency and so the invoker will send up to N invocations to that the container when possible. https://github.com/apache/incubator-openwhisk/pull/2795 -r > On Oct 29, 2018, at 7:06 AM, Michele Sciabarra wrote: > > I would like to understand a bit more how concurrency is handled by the > Invoker. > > If there are, say, 2 concurrent requests, can the invoker send the second > request to the same runtime BEFORE it has completed the first request? I have > seen some tests related in the nodejs runtime but none in the basic tests. > > Also, Carlos mentioned the need to increase the concurrency in the same > runtime. It can be done but I need to know the requirements. So far I assumed > that until a request is not completed, the invoker will spawn another runtime > to serve them. > > > > -- > Michele Sciabarra > mich...@sciabarra.com
A question about concurrency in OpenWhisk
I would like to understand a bit more how concurrency is handled by the Invoker. If there are, say, 2 concurrent requests, can the invoker send the second request to the same runtime BEFORE it has completed the first request? I have seen some tests related in the nodejs runtime but none in the basic tests. Also, Carlos mentioned the need to increase the concurrency in the same runtime. It can be done but I need to know the requirements. So far I assumed that until a request is not completed, the invoker will spawn another runtime to serve them. -- Michele Sciabarra mich...@sciabarra.com
Re: Profiling OpenWhisk operation
If you just want Graphana dashboards you can use the instructions in the docs that point to an image with StastD here https://github.com/apache/incubator-openwhisk/blob/master/docs/metrics.md If you want to get the metrics from StatsD to Prometheus I found this looks like it can be use, it even says you can skip StatsD server and just point directly the the exporter https://github.com/prometheus/statsd_exporter If you have success would be awesome for you to add a reference to the metrics doc. — Carlos On Mon, Oct 29, 2018 at 6:55 AM Erez Hadad wrote: > Hi folks, > > I've found several indications in the OpenWhisk code, showing that it uses > Kamon (kamon.io) for reporting internal metrics and profiling. > Can someone please explain (or direct to an explanation) how to set up > OpenWhisk profiling with Kamon and Prometheus? > > Regards, > -- Erez > > Erez Hadad, PhD > Cloud System Technologies > IBM Research - Haifa > email: er...@il.ibm.com > phone: +972-4-829-6509 > > > >
Profiling OpenWhisk operation
Hi folks, I've found several indications in the OpenWhisk code, showing that it uses Kamon (kamon.io) for reporting internal metrics and profiling. Can someone please explain (or direct to an explanation) how to set up OpenWhisk profiling with Kamon and Prometheus? Regards, -- Erez Erez Hadad, PhD Cloud System Technologies IBM Research - Haifa email: er...@il.ibm.com phone: +972-4-829-6509
