Re: Propose to combine module gearpump-daemon into gearpump-core
+1 as I can’t find another better solution. 在 2016/10/11 下午4:29,“Karol Brejna”写入: +1 I think moving EmbeddedCluster capabilities to gearpump-core makes sense and simplifies things. On Tue, Oct 11, 2016 at 7:04 AM, Manu Zhang wrote: > +1. Integration tests for Beam Gearpump runner heavily depends on > EmbeddedCluster. It's quite weird to add `gearpump-daemon` dependency to > run gearpump examples. > > Thanks, > Manu > > On Tue, Oct 11, 2016 at 10:17 AM Vincent Wang wrote: > >> Hi all, >> >> Currently we have an EmbeddedCluster in gearpump-daemon module which >> allows user to run their gearpump application in a local JVM. >> Now suppose user want to implement an streaming application using >> Gearpump and use the EmbeddedCluster to test it locally or even in an IDE. >> So the application needs to have extra dependency on the gearpump-daemon >> module to access EmbeddedCluster, and user have to write code like >> * if (debugMode) {* >> * localCluster = new EmbeddedCluster(akkaConf);* >> * localCluster.start();* >> *}* >> *ClientContext masterClient = null;* >> *if (localCluster != null) {* >> * masterClient = localCluster.newClientContext();* >> *} else {* >> * masterClient = new ClientContext(akkaConf);* >> *}* >> *masterClient.submit(app);* >> >> Which is obviously tedious and inconvenient. >> >> So we'd like to provide user with a implementation just like what Flink >> did. Basically we want to integrate EmbeddedCluster into ClientContext so >> that user can switch the application running mode easily by some >> configuration or decision made by ClientConext automatically. >> However, currently the ClientContext is in module gearpump-core and >> EmbeddedCluster >> has dependencies on Worker and Master daemon actors, which means these >> classes need to be moved along with EmbeddedCluster and finally almost the >> whole gearpump-daemon module need to be integrated into gearpump-core >> module. >>That's the side effect of achieving a more powerful ClientContext >> integrated with a EmbeddedCluster and I'd like to hear your opinions about >> on this proposal. >> >> Thanks, >> Huafeng >>
Re: Propose to combine module gearpump-daemon into gearpump-core
+1 I think moving EmbeddedCluster capabilities to gearpump-core makes sense and simplifies things. On Tue, Oct 11, 2016 at 7:04 AM, Manu Zhangwrote: > +1. Integration tests for Beam Gearpump runner heavily depends on > EmbeddedCluster. It's quite weird to add `gearpump-daemon` dependency to > run gearpump examples. > > Thanks, > Manu > > On Tue, Oct 11, 2016 at 10:17 AM Vincent Wang wrote: > >> Hi all, >> >> Currently we have an EmbeddedCluster in gearpump-daemon module which >> allows user to run their gearpump application in a local JVM. >> Now suppose user want to implement an streaming application using >> Gearpump and use the EmbeddedCluster to test it locally or even in an IDE. >> So the application needs to have extra dependency on the gearpump-daemon >> module to access EmbeddedCluster, and user have to write code like >> * if (debugMode) {* >> * localCluster = new EmbeddedCluster(akkaConf);* >> * localCluster.start();* >> *}* >> *ClientContext masterClient = null;* >> *if (localCluster != null) {* >> * masterClient = localCluster.newClientContext();* >> *} else {* >> * masterClient = new ClientContext(akkaConf);* >> *}* >> *masterClient.submit(app);* >> >> Which is obviously tedious and inconvenient. >> >> So we'd like to provide user with a implementation just like what Flink >> did. Basically we want to integrate EmbeddedCluster into ClientContext so >> that user can switch the application running mode easily by some >> configuration or decision made by ClientConext automatically. >> However, currently the ClientContext is in module gearpump-core and >> EmbeddedCluster >> has dependencies on Worker and Master daemon actors, which means these >> classes need to be moved along with EmbeddedCluster and finally almost the >> whole gearpump-daemon module need to be integrated into gearpump-core >> module. >>That's the side effect of achieving a more powerful ClientContext >> integrated with a EmbeddedCluster and I'd like to hear your opinions about >> on this proposal. >> >> Thanks, >> Huafeng >>
Re: Propose to combine module gearpump-daemon into gearpump-core
+1. Integration tests for Beam Gearpump runner heavily depends on EmbeddedCluster. It's quite weird to add `gearpump-daemon` dependency to run gearpump examples. Thanks, Manu On Tue, Oct 11, 2016 at 10:17 AM Vincent Wangwrote: > Hi all, > > Currently we have an EmbeddedCluster in gearpump-daemon module which > allows user to run their gearpump application in a local JVM. > Now suppose user want to implement an streaming application using > Gearpump and use the EmbeddedCluster to test it locally or even in an IDE. > So the application needs to have extra dependency on the gearpump-daemon > module to access EmbeddedCluster, and user have to write code like > * if (debugMode) {* > * localCluster = new EmbeddedCluster(akkaConf);* > * localCluster.start();* > *}* > *ClientContext masterClient = null;* > *if (localCluster != null) {* > * masterClient = localCluster.newClientContext();* > *} else {* > * masterClient = new ClientContext(akkaConf);* > *}* > *masterClient.submit(app);* > > Which is obviously tedious and inconvenient. > > So we'd like to provide user with a implementation just like what Flink > did. Basically we want to integrate EmbeddedCluster into ClientContext so > that user can switch the application running mode easily by some > configuration or decision made by ClientConext automatically. > However, currently the ClientContext is in module gearpump-core and > EmbeddedCluster > has dependencies on Worker and Master daemon actors, which means these > classes need to be moved along with EmbeddedCluster and finally almost the > whole gearpump-daemon module need to be integrated into gearpump-core > module. >That's the side effect of achieving a more powerful ClientContext > integrated with a EmbeddedCluster and I'd like to hear your opinions about > on this proposal. > > Thanks, > Huafeng >