Re: Question about sub-projects and project merging
Hi Jay, Looking at your question, I see the Apache Samza and Apache Kafka *communities* have little overlap(*). The Board looks at communities, and their overlap or lack thereof. Smushing two communities under one TLP is what we have historically called an umbrella TLP, and discourage. Communities should be allowed to operate independently. If you have *one* community, then one TLP makes sense. If you have *two* communities, then increase the overlap. When they look like one community, and that one community votes to merge TLPs ... then ask for that. Cheers, -g (*) 2 common PMC members, 3 common committers. On Mon, Jul 13, 2015 at 12:37 AM, Jay Kreps jay.kr...@gmail.com wrote: Hey board members, There is a longish thread on the Apache Samza mailing list on the relationship between Kafka and Samza and whether they wouldn't make a lot more sense as a single project. This raised some questions I was hoping to get advice on. Discussion thread (warning: super long, I attempt to summarize relevant bits below): http://mail-archives.apache.org/mod_mbox/samza-dev/201507.mbox/%3ccabyby7d_-jcxj7fizsjuebjedgbep33flyx3nrozt0yeox9...@mail.gmail.com%3E Anyhow, some people thought Apache has lot's of sub-projects, that would be a graceful way to step in the right direction. At that point others popped up and said, sub-projects are discouraged by the board. I'm not sure if we understand technically what a subproject is, but I think it means a second repo/committership under the same PMC. A few questions: - Is that what a sub-project is? - Are they discouraged? If so, why? - Assuming it makes sense in this case what is the process for making one? - Putting aside sub-projects as a mechanism what are examples where communities merged successfully? We were pointed towards Lucene/SOLR. Are there others? Relevant background info: - Samza depends on Kafka, but not vice versa - There is some overlap in committers but not extensive (3/11 Samza committers are also Kafka committers) Thanks for the advice! -Jay
Re: Thoughts and obesrvations on Samza
I am leaning to Jay's fifth approach. It is not radical and gives us some time to see the outcome. In addition, I would suggest: 1) Keep the SystemConsumer/SystemProducer API. Because current SystemConsumer/SystemProducer API satisfies the usage (From Joardan, and even Garry's feedback) and is not so broken that we want to deprecate it. Though there are some issues in implemnting the Kinesis, they are not unfixable. Nothing should prevent Samza, as a stream processing system, to support other systems. In addition, there already are some systems exiting besides Kafka: ElasticSearch (committed to the master), HDFS (patch-available), S3( from the mailing list), Kinesis (developing in another repository), ActiveMQ (in two months). We may want to see how those go before we kill them. 2) Can have some Samza devs involved in Kafka's transformer client API. This can not only help the future integration (if any) much easier, because they have knowledge about both systems, but also good for Kafka's community, because Samza devs have the streaming process experience that Kafka devs may miss. 3) Samza's partition management system may still support other systems. Though the partition management logic in samza-kafka will be moved to Kafka, its still useful for other systems that do not have the partition management layer. 4) Start sharing the docs/websites and using the same terminology (though do not know how to do this exactly. :). This will reduce the future confusion and does not hurt Samza's independency. In my opinion, Samza, as a standalone project, still can (and already) heavily replying on Kafka, and even more tuned for Kafka-specific usage. Kafka, also can embed Samza in the document, I do not see anything prevent doing this. Thanks, Fang, Yan yanfang...@gmail.com On Mon, Jul 13, 2015 at 11:25 AM, Jordan Shaw jor...@pubnub.com wrote: Jay, I think doing this iteratively in smaller chunks is a better way to go as new issues arise. As Navina said Kafka is a stream system and Samza is a stream processor and those two ideas should be mutually exclusive. -Jordan On Mon, Jul 13, 2015 at 10:06 AM, Jay Kreps jay.kr...@gmail.com wrote: Hmm, thought about this more. Maybe this is just too much too quick. Overall I think there is some enthusiasm for the proposal but it's not really unanimous enough to make any kind of change this big cleanly. The board doesn't really like the merging stuff, user's are concerned about compatibility, I didn't feel there was unanimous agreement on dropping SystemConsumer, etc. Even if this is the right end state to get to, probably trying to push all this through at once isn't the right way to do it. So let me propose a kind of fifth (?) option which I think is less dramatic and let's things happen gradually. I think this is kind of like combining the first part of Yi's proposal and Jakob's third option, leaving the rest to be figured out incrementally: Option 5: We continue the prototype I shared and propose that as a kind of transformer client API in Kafka. This isn't really a full-fledged stream processing layer, more like a supped up consumer api for munging topics. This would let us figure out some of the technical bits, how to do this on Kafka's group management features, how to integrate the txn feature to do the exactly-once stuff in these transformations, and get all this stuff solid. This api would have valid uses in it's own right, especially when your transformation will be embedded inside an existing service or application which isn't possible with Samza (or other existing systems that I know of). Independently we can iterate on some of the ideas of the original proposal individually and figure out how (if at all) to make use of this functionality. This can be done bit-by-bit: - Could be that the existing StreamTask API ends up wrapping this - Could end up exposed directly in Samza as Yi proposed - Could be that just the lower-level group-management stuff get's used, and in this case it could be either just for standalone mode, or always - Could be that it stays as-is The advantage of this is it is lower risk...we basically don't have to make 12 major decisions all at once that kind of hinge on what amounts to a pretty aggressive rewrite. The disadvantage of this is it is a bit more confusing as all this is getting figured out. As with some of the other stuff, this would require a further discussion in the Kafka community if people do like this approach. Thoughts? -Jay On Sun, Jul 12, 2015 at 10:52 PM, Jay Kreps jay.kr...@gmail.com wrote: Hey Chris, Yeah, I'm obviously in favor of this. The sub-project approach seems the ideal way to take a graceful step in this direction, so I will ping the board folks and see why they are discouraged, it would be good to understand that. If we go that route we would need to do a similar
Re: Question about sub-projects and project merging
Hey Mike, Thanks for sharing, it is helpful to hear the experience that leads to these recommendations. -Jay On Mon, Jul 13, 2015 at 11:01 AM, Mike Kienenberger mkien...@gmail.com wrote: A subproject is one of many projects that fall under the same umbrella project management committee (PMC). It doesn't have to be a separate repo, but it generally has a separate community or a subset of the full community. Speaking as a long-time PMC member for MyFaces, our problem with subprojects (we have 11!) is that it's hard to keep accountability and monitor community health. A subproject starts of being active with some subset of the community, but then reduces activity at some future point. Those who aren't directly involved with the subproject tend not to notice that the particular subproject has fallen to unhealthy levels. Generally, you don't realize something is wrong until after all of the developers have left when you suddenly realize that there's no one answering questions, applying patches, or familiar with the code base. Non-umbrella projects report to the board are expected to evaluate community health each quarter. Umbrella projects are also supposed to do this, but often fail to realize that community health has to be individually evaluated for each subproject each quarter. The PMC chair is likely not directly involved with each subproject, and may not be in a good position to evaluate the sub-project's health. As Hervé mentions, this is particularly true for TLPs which have a main project and optional modules where everyone cares about the main project and only a few care about each module subproject. This is what happened with MyFaces. What tends to happen with umbrella projects is that you end up creating two-tier project management. Those responsible to the board are upper management but may not be directly involved and fail to understand the subproject community health. Those who are supposed to actively manage the project are lower management and are not directly responsible to the board for quarterly reports. Best practice is to have a one-tier PMC. As soon as a subproject is healthy enough to stand on its own, it probably should go TLP. MyFaces successfully spun off DeltaSpike, and DeltaSpike remains healthy. The other alternative is to be certain to address the status of each subproject in the board report, much like the Incubator board report does each time. My advice is the same as others -- keep the two projects separate, but encourage individual Samza committers join as Kafka committers if they feel the need to do so. On Mon, Jul 13, 2015 at 1:37 AM, Jay Kreps jay.kr...@gmail.com wrote: Hey board members, There is a longish thread on the Apache Samza mailing list on the relationship between Kafka and Samza and whether they wouldn't make a lot more sense as a single project. This raised some questions I was hoping to get advice on. Discussion thread (warning: super long, I attempt to summarize relevant bits below): http://mail-archives.apache.org/mod_mbox/samza-dev/201507.mbox/%3ccabyby7d_-jcxj7fizsjuebjedgbep33flyx3nrozt0yeox9...@mail.gmail.com%3E Anyhow, some people thought Apache has lot's of sub-projects, that would be a graceful way to step in the right direction. At that point others popped up and said, sub-projects are discouraged by the board. I'm not sure if we understand technically what a subproject is, but I think it means a second repo/committership under the same PMC. A few questions: - Is that what a sub-project is? - Are they discouraged? If so, why? - Assuming it makes sense in this case what is the process for making one? - Putting aside sub-projects as a mechanism what are examples where communities merged successfully? We were pointed towards Lucene/SOLR. Are there others? Relevant background info: - Samza depends on Kafka, but not vice versa - There is some overlap in committers but not extensive (3/11 Samza committers are also Kafka committers) Thanks for the advice! -Jay
Re: Thoughts and obesrvations on Samza
Jay, I think doing this iteratively in smaller chunks is a better way to go as new issues arise. As Navina said Kafka is a stream system and Samza is a stream processor and those two ideas should be mutually exclusive. -Jordan On Mon, Jul 13, 2015 at 10:06 AM, Jay Kreps jay.kr...@gmail.com wrote: Hmm, thought about this more. Maybe this is just too much too quick. Overall I think there is some enthusiasm for the proposal but it's not really unanimous enough to make any kind of change this big cleanly. The board doesn't really like the merging stuff, user's are concerned about compatibility, I didn't feel there was unanimous agreement on dropping SystemConsumer, etc. Even if this is the right end state to get to, probably trying to push all this through at once isn't the right way to do it. So let me propose a kind of fifth (?) option which I think is less dramatic and let's things happen gradually. I think this is kind of like combining the first part of Yi's proposal and Jakob's third option, leaving the rest to be figured out incrementally: Option 5: We continue the prototype I shared and propose that as a kind of transformer client API in Kafka. This isn't really a full-fledged stream processing layer, more like a supped up consumer api for munging topics. This would let us figure out some of the technical bits, how to do this on Kafka's group management features, how to integrate the txn feature to do the exactly-once stuff in these transformations, and get all this stuff solid. This api would have valid uses in it's own right, especially when your transformation will be embedded inside an existing service or application which isn't possible with Samza (or other existing systems that I know of). Independently we can iterate on some of the ideas of the original proposal individually and figure out how (if at all) to make use of this functionality. This can be done bit-by-bit: - Could be that the existing StreamTask API ends up wrapping this - Could end up exposed directly in Samza as Yi proposed - Could be that just the lower-level group-management stuff get's used, and in this case it could be either just for standalone mode, or always - Could be that it stays as-is The advantage of this is it is lower risk...we basically don't have to make 12 major decisions all at once that kind of hinge on what amounts to a pretty aggressive rewrite. The disadvantage of this is it is a bit more confusing as all this is getting figured out. As with some of the other stuff, this would require a further discussion in the Kafka community if people do like this approach. Thoughts? -Jay On Sun, Jul 12, 2015 at 10:52 PM, Jay Kreps jay.kr...@gmail.com wrote: Hey Chris, Yeah, I'm obviously in favor of this. The sub-project approach seems the ideal way to take a graceful step in this direction, so I will ping the board folks and see why they are discouraged, it would be good to understand that. If we go that route we would need to do a similar discussion in the Kafka list (but makes sense to figure out first if it is what Samza wants). Irrespective of how it's implemented, though, to me the important things are the following: 1. Unify the website, config, naming, docs, metrics, etc--basically fix the product experience so the stream and the processing feel like a single user experience and brand. This seems minor but I think is a really big deal. 2. Make standalone mode a first class citizen and have a real technical plan to be able to support cluster managers other than YARN. 3. Make the config and out-of-the-box experience more usable I think that prototype gives a practical example of how 1-3 could be done and we should pursue it. This is a pretty radical change, so I wouldn't be shocked if people didn't want to take a step like that. Maybe it would make sense to see if people are on board with that general idea, and then try to get some advice on sub-projects in parallel and nail down those details? -Jay On Sun, Jul 12, 2015 at 5:54 PM, Chris Riccomini criccom...@apache.org wrote: Hey all, I want to start by saying that I'm absolutely thrilled to be a part of this community. The amount of level-headed, thoughtful, educated discussion that's gone on over the past ~10 days is overwhelming. Wonderful. It seems like discussion is waning a bit, and we've reached some conclusions. There are several key emails in this threat, which I want to call out: 1. Jakob's summary of the three potential ways forward. http://mail-archives.apache.org/mod_mbox/samza-dev/201507.mbox/%3CCADiKvVu-hxdBfyQ4qm3LDC55cUQbPdmbe4zGzTOOatYF1Pz43A%40mail.gmail.com%3E 2. Julian's call out that we should be focusing on community over code. http://mail-archives.apache.org/mod_mbox/samza-dev/201507.mbox/%3CCAPSgeESZ_7bVFbwN%2Bzqi5MH%3D4CWu9MZUSanKg0-1woMqt55Fvg%40mail.gmail.com%3E 3. Martin's
Re: Thoughts and obesrvations on Samza
Hmm, thought about this more. Maybe this is just too much too quick. Overall I think there is some enthusiasm for the proposal but it's not really unanimous enough to make any kind of change this big cleanly. The board doesn't really like the merging stuff, user's are concerned about compatibility, I didn't feel there was unanimous agreement on dropping SystemConsumer, etc. Even if this is the right end state to get to, probably trying to push all this through at once isn't the right way to do it. So let me propose a kind of fifth (?) option which I think is less dramatic and let's things happen gradually. I think this is kind of like combining the first part of Yi's proposal and Jakob's third option, leaving the rest to be figured out incrementally: Option 5: We continue the prototype I shared and propose that as a kind of transformer client API in Kafka. This isn't really a full-fledged stream processing layer, more like a supped up consumer api for munging topics. This would let us figure out some of the technical bits, how to do this on Kafka's group management features, how to integrate the txn feature to do the exactly-once stuff in these transformations, and get all this stuff solid. This api would have valid uses in it's own right, especially when your transformation will be embedded inside an existing service or application which isn't possible with Samza (or other existing systems that I know of). Independently we can iterate on some of the ideas of the original proposal individually and figure out how (if at all) to make use of this functionality. This can be done bit-by-bit: - Could be that the existing StreamTask API ends up wrapping this - Could end up exposed directly in Samza as Yi proposed - Could be that just the lower-level group-management stuff get's used, and in this case it could be either just for standalone mode, or always - Could be that it stays as-is The advantage of this is it is lower risk...we basically don't have to make 12 major decisions all at once that kind of hinge on what amounts to a pretty aggressive rewrite. The disadvantage of this is it is a bit more confusing as all this is getting figured out. As with some of the other stuff, this would require a further discussion in the Kafka community if people do like this approach. Thoughts? -Jay On Sun, Jul 12, 2015 at 10:52 PM, Jay Kreps jay.kr...@gmail.com wrote: Hey Chris, Yeah, I'm obviously in favor of this. The sub-project approach seems the ideal way to take a graceful step in this direction, so I will ping the board folks and see why they are discouraged, it would be good to understand that. If we go that route we would need to do a similar discussion in the Kafka list (but makes sense to figure out first if it is what Samza wants). Irrespective of how it's implemented, though, to me the important things are the following: 1. Unify the website, config, naming, docs, metrics, etc--basically fix the product experience so the stream and the processing feel like a single user experience and brand. This seems minor but I think is a really big deal. 2. Make standalone mode a first class citizen and have a real technical plan to be able to support cluster managers other than YARN. 3. Make the config and out-of-the-box experience more usable I think that prototype gives a practical example of how 1-3 could be done and we should pursue it. This is a pretty radical change, so I wouldn't be shocked if people didn't want to take a step like that. Maybe it would make sense to see if people are on board with that general idea, and then try to get some advice on sub-projects in parallel and nail down those details? -Jay On Sun, Jul 12, 2015 at 5:54 PM, Chris Riccomini criccom...@apache.org wrote: Hey all, I want to start by saying that I'm absolutely thrilled to be a part of this community. The amount of level-headed, thoughtful, educated discussion that's gone on over the past ~10 days is overwhelming. Wonderful. It seems like discussion is waning a bit, and we've reached some conclusions. There are several key emails in this threat, which I want to call out: 1. Jakob's summary of the three potential ways forward. http://mail-archives.apache.org/mod_mbox/samza-dev/201507.mbox/%3CCADiKvVu-hxdBfyQ4qm3LDC55cUQbPdmbe4zGzTOOatYF1Pz43A%40mail.gmail.com%3E 2. Julian's call out that we should be focusing on community over code. http://mail-archives.apache.org/mod_mbox/samza-dev/201507.mbox/%3CCAPSgeESZ_7bVFbwN%2Bzqi5MH%3D4CWu9MZUSanKg0-1woMqt55Fvg%40mail.gmail.com%3E 3. Martin's summary about the benefits of merging communities. http://mail-archives.apache.org/mod_mbox/samza-dev/201507.mbox/%3CBFB866B6-D9D8-4578-93C0-FFAEB1DF00FC%40kleppmann.com%3E 4. Jakob's comments about the distinction between community and code paths. http://mail-archives.apache.org/mod_mbox/samza-dev/201507.mbox/%3CCADiKvVtWPjHLLDsmxvz9KggVA5DfBi-nUvfqB6QdA-du%2B_a9Ng%40mail.gmail.com%3E I
Re: Review Request 36274: SAMZA-401: getCpuTime to truly calculate duty cycle of the event loop
--- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/36274/#review91527 --- samza-core/src/main/scala/org/apache/samza/container/RunLoop.scala (lines 75 - 76) https://reviews.apache.org/r/36274/#comment144953 for more accurate, I think the activeNs should go before totalNs. - Yan Fang On July 7, 2015, 7:08 p.m., Luis De Pombo wrote: --- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/36274/ --- (Updated July 7, 2015, 7:08 p.m.) Review request for samza. Repository: samza Description --- SAMZA-401: getCpuTime to truly calculate duty cycle of the event loop Diffs - samza-core/src/main/scala/org/apache/samza/container/RunLoop.scala c292ae47cd89ef0f25dc682c02dd288e2ba6dcc5 samza-core/src/main/scala/org/apache/samza/util/TimerUtils.scala 1643070dd710efb9ade9eb5812dabd6fa60ce023 samza-core/src/main/scala/org/apache/samza/util/Util.scala 2feb65b729b45fbc3b83a75c4072527e3c4e60be samza-core/src/test/scala/org/apache/samza/container/TestRunLoop.scala 64a5844bdb343a3c509cba059b9f3b9a19dc9eff Diff: https://reviews.apache.org/r/36274/diff/ Testing --- Thanks, Luis De Pombo
You're invited to an Apache Samza meetup hosted at LinkedIn on July 21 @ 6PM
Hi all - The Samza development team invites you to join us at an Apache Samzahttp://samza.apache.org/ meetup hosted in the Unite conference room at LinkedIn's Mountain View campus on Tuesday, July 21 at 6PM. Food/drinks and streaming video will be provided. Please RSVP herehttp://www.meetup.com/Bay-Area-Samza-Meetup/events/223768847/ if you plan to attend the event in-person. Here are two topics we’ll be covering. We’ll probably have a 3rd topic which is still under discussion. Harvesting the Power of Samza in LinkedIn's Feed - LinkedIn's Feed is the entry point for hundreds of millions of members who seek to stay informed about their professional interests. The feed strives to provide relevant content to members that's also new and fresh. How does the feed solve this problem at scale? What role does Samza play in this? Join us to find out. Athena - Stream porcessing platform @ Uber - We present Athena - a stream processing platform for Uber's near real time analytics applications, built using Samza. We will be discussing some of the existing and upcoming use cases and how they impact the Uber partners / riders. We will be talking about the tooling built around Samza for easier user on-boarding - such as deployment manager, integration with typesafe config system, unit test framework, Graphite integration, metric whitelisting and so on. We'll also go over some of the issues observed during this process. Hope to see you there! Ed Yakabosky (TPM, Samza @ LinkedIn)
Re: Thoughts and obesrvations on Samza
Hi, Garry, Just want to chime in to state our experience in LinkedIn. In LinkedIn, we have a lot of aggregation/transformation stream processing jobs that falls into the transformation category. That's also the motivation for us to develop the SQL layer on top of streams to allow easy programming model for data transformation on streams. Ingestion from wide-range of sources and egress to some serving tier are important, but I would argue that w/o the transformation in between, there is not much value added by stream processing. Just my 2-cents. On Mon, Jul 13, 2015 at 12:56 PM, Garry Turkington g.turking...@improvedigital.com wrote: Hi, I'm also supportive of Jay's option 5. There is a risk the transformer API -- I'd have preferred Metamorphosis but it's too hard to type! -- takes on a life of its own and we end up with two very different things but given how good the Kafka community has been at introducing new producer and consumer clients and giving very clear guidance on when they are production ready this is a danger I believe can be managed. It'd also be excellent to get some working code to beat around the notions of stream processing atop a system with transacdtional messages. On the question of whether to keep or deprecate SystemConsumer/Producer I believe we need get a better understanding over the next while of just what the Samza community is looking for in such connectivity. For my own use cases I have been looking to add additional implementations primarily to use Samza as the data ingress and egress component around Kafka. Writing external clients that require their own reliability and scalability management gets old real fast and pushing this into a simple Samza job that reads from system X and pushes into Kafka (or vice versa) was the obvious choice for me in the current model. For this type of usage though copycat is likely much superior (obviously needs proven) and the question then is if most Samza users look to the system implementations to also act as a front-end into Kafka or if significant usage is indeed intended to have the alternative systems as the primary message source. That understanding will I think give much clarity in just what value the abstraction overhead of the current model brings. Garry -Original Message- From: Yan Fang [mailto:yanfang...@gmail.com] Sent: 13 July 2015 19:58 To: dev@samza.apache.org Subject: Re: Thoughts and obesrvations on Samza I am leaning to Jay's fifth approach. It is not radical and gives us some time to see the outcome. In addition, I would suggest: 1) Keep the SystemConsumer/SystemProducer API. Because current SystemConsumer/SystemProducer API satisfies the usage (From Joardan, and even Garry's feedback) and is not so broken that we want to deprecate it. Though there are some issues in implemnting the Kinesis, they are not unfixable. Nothing should prevent Samza, as a stream processing system, to support other systems. In addition, there already are some systems exiting besides Kafka: ElasticSearch (committed to the master), HDFS (patch-available), S3( from the mailing list), Kinesis (developing in another repository), ActiveMQ (in two months). We may want to see how those go before we kill them. 2) Can have some Samza devs involved in Kafka's transformer client API. This can not only help the future integration (if any) much easier, because they have knowledge about both systems, but also good for Kafka's community, because Samza devs have the streaming process experience that Kafka devs may miss. 3) Samza's partition management system may still support other systems. Though the partition management logic in samza-kafka will be moved to Kafka, its still useful for other systems that do not have the partition management layer. 4) Start sharing the docs/websites and using the same terminology (though do not know how to do this exactly. :). This will reduce the future confusion and does not hurt Samza's independency. In my opinion, Samza, as a standalone project, still can (and already) heavily replying on Kafka, and even more tuned for Kafka-specific usage. Kafka, also can embed Samza in the document, I do not see anything prevent doing this. Thanks, Fang, Yan yanfang...@gmail.com On Mon, Jul 13, 2015 at 11:25 AM, Jordan Shaw jor...@pubnub.com wrote: Jay, I think doing this iteratively in smaller chunks is a better way to go as new issues arise. As Navina said Kafka is a stream system and Samza is a stream processor and those two ideas should be mutually exclusive. -Jordan On Mon, Jul 13, 2015 at 10:06 AM, Jay Kreps jay.kr...@gmail.com wrote: Hmm, thought about this more. Maybe this is just too much too quick. Overall I think there is some enthusiasm for the proposal but it's not really unanimous enough to make any kind of change this big cleanly. The board doesn't really like the merging stuff, user's are concerned about
Re: Review Request 36089: SAMZA-670 Allow easier access to JMX port
--- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/36089/#review91533 --- samza-core/src/main/scala/org/apache/samza/container/SamzaContainer.scala (line 634) https://reviews.apache.org/r/36089/#comment144956 I think a better way, which requires much fewer changes, is to call something like jmxServer.getJmxUrl, jmxServer.jmxTunelingUrl. jmxServer can be a variable of SamzaContainer Object. Then we do not need to change ContainerModel, JobModel, SamzaContext. Because there is no reason that we want to contain jmx information into those three objects. - Yan Fang On July 1, 2015, 2:07 p.m., József Márton Jung wrote: --- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/36089/ --- (Updated July 1, 2015, 2:07 p.m.) Review request for samza. Repository: samza Description --- JMX address of application master and the containers is available through AM UI Diffs - checkstyle/import-control.xml 3374f0c samza-api/src/main/java/org/apache/samza/container/SamzaContainerContext.java fd7333b samza-core/src/main/java/org/apache/samza/container/LocalityManager.java e661e12 samza-core/src/main/java/org/apache/samza/coordinator/stream/CoordinatorStreamMessage.java 6c1e488 samza-core/src/main/java/org/apache/samza/job/model/ContainerModel.java 98a34bc samza-core/src/main/java/org/apache/samza/job/model/JobModel.java 95a2ce5 samza-core/src/main/scala/org/apache/samza/container/SamzaContainer.scala cbacd18 samza-core/src/main/scala/org/apache/samza/coordinator/JobCoordinator.scala 8ee034a samza-core/src/main/scala/org/apache/samza/metrics/JmxServer.scala f343faf samza-core/src/test/scala/org/apache/samza/container/TestSamzaContainer.scala 9fb1aa9 samza-core/src/test/scala/org/apache/samza/container/TestTaskInstance.scala 7caad28 samza-test/src/main/scala/org/apache/samza/test/performance/TestKeyValuePerformance.scala 1ce7d25 samza-yarn/src/main/resources/scalate/WEB-INF/views/index.scaml cf0d2fc samza-yarn/src/main/scala/org/apache/samza/job/yarn/SamzaAppMaster.scala 20aa373 samza-yarn/src/main/scala/org/apache/samza/job/yarn/SamzaAppMasterState.scala 1445605 Diff: https://reviews.apache.org/r/36089/diff/ Testing --- Thanks, József Márton Jung
Re: Thoughts and obesrvations on Samza
Hi, Jay, Given all the user concerns, the board disagreement on sub-projects, I am supporting your 5th option as well. As you said, even the end goal is the same, it might help to pave a smooth path forward. One thing I learned over the years is that what we planned for may not be the final product, and the unexpected product may be even better if we learn and adapt along the way. :) So, since I assume that in option 5, Samza will fully embrace the new Kafka Streams API as the core and heavily depends on it, I want to raise up some detailed logistic questions: 1. How do Samza community contribute to the design and development of the new Kafka Streams API? As Kartik mentioned, if there is a model for Samza community to contribute to just this part of Kafka code base, it would be a huge plus point to the integration. 2. What's the scope of the new Kafka Streams API? Is it just focused on message consumption, producing, Kafka-based partition distribution, offset management, message selection and delivery to StreamProcessor? In other words, I have a question regarding to whether we should put samza-kv-store in the scope? The reasons that I think that it might be better to stay in Samza initially are: a) KV-store libraries does not directly interact w/ Kafka brokers, it optionally uses Kafka consumers and producers like a client program; b) there are a tons of experiments / tune-ups on RocksDB that we want to have a faster iteration on this library (e.g. there is an experimental time-sequence KV store implementation from LinkedIn we also want to experiment on in window operator in SQL). The down-side I can see is that w/o this in Kafka Streams API, the as-a-library mode may not get the state management support. If we can find a way to make sure that the current Samza community can contribute to this library in a faster velocity, I can be convinced otherwise as well. What's your opinion on this? Overall, thanks a lot for pushing forward the whole discussion! -Yi On Mon, Jul 13, 2015 at 12:56 PM, Garry Turkington g.turking...@improvedigital.com wrote: Hi, I'm also supportive of Jay's option 5. There is a risk the transformer API -- I'd have preferred Metamorphosis but it's too hard to type! -- takes on a life of its own and we end up with two very different things but given how good the Kafka community has been at introducing new producer and consumer clients and giving very clear guidance on when they are production ready this is a danger I believe can be managed. It'd also be excellent to get some working code to beat around the notions of stream processing atop a system with transacdtional messages. On the question of whether to keep or deprecate SystemConsumer/Producer I believe we need get a better understanding over the next while of just what the Samza community is looking for in such connectivity. For my own use cases I have been looking to add additional implementations primarily to use Samza as the data ingress and egress component around Kafka. Writing external clients that require their own reliability and scalability management gets old real fast and pushing this into a simple Samza job that reads from system X and pushes into Kafka (or vice versa) was the obvious choice for me in the current model. For this type of usage though copycat is likely much superior (obviously needs proven) and the question then is if most Samza users look to the system implementations to also act as a front-end into Kafka or if significant usage is indeed intended to have the alternative systems as the primary message source. That understanding will I think give much clarity in just what value the abstraction overhead of the current model brings. Garry -Original Message- From: Yan Fang [mailto:yanfang...@gmail.com] Sent: 13 July 2015 19:58 To: dev@samza.apache.org Subject: Re: Thoughts and obesrvations on Samza I am leaning to Jay's fifth approach. It is not radical and gives us some time to see the outcome. In addition, I would suggest: 1) Keep the SystemConsumer/SystemProducer API. Because current SystemConsumer/SystemProducer API satisfies the usage (From Joardan, and even Garry's feedback) and is not so broken that we want to deprecate it. Though there are some issues in implemnting the Kinesis, they are not unfixable. Nothing should prevent Samza, as a stream processing system, to support other systems. In addition, there already are some systems exiting besides Kafka: ElasticSearch (committed to the master), HDFS (patch-available), S3( from the mailing list), Kinesis (developing in another repository), ActiveMQ (in two months). We may want to see how those go before we kill them. 2) Can have some Samza devs involved in Kafka's transformer client API. This can not only help the future integration (if any) much easier, because they have knowledge about both systems, but also good for Kafka's community, because Samza devs have the streaming process
Review Request 36471: added stream for auto scaling, consumer to read from the stream in profiler, sliding window metric
--- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/36471/ --- Review request for samza and Navina Ramesh. Repository: samza Description --- Autoscaling for samza (work in progress) This work is for SAMZA-719. Currently, a fixed number of containers is assigned to a job as an input configuration parameter. However, with this design jobs can fail due to lack of enough resources (such as memory), or they can become a bottleneck in a workflow containing many jobs. While auto-scaling is much broader term, the goal of this project will be to enable a Samza job to automatically scale its containers such that there is improved job performance. Based on the design, we need a profiler, analyser, optimizer and deployer module. -currently profiler and analyzer added. -tests not added for those components, and further testing is needed. Diffs - checkstyle/import-control.xml 3374f0c432e61ac4cda275377604cfd481f0cddf samza-api/src/main/java/org/apache/samza/autoScaling/AutoScalingMode.java PRE-CREATION samza-api/src/main/java/org/apache/samza/autoScaling/Profiler.java PRE-CREATION samza-api/src/main/java/org/apache/samza/autoScaling/analyzer/Analyzer.java PRE-CREATION samza-core/src/main/java/org/apache/samza/autoScaling/AutoScalingSystem.java PRE-CREATION samza-core/src/main/java/org/apache/samza/autoScaling/SnapshotReporterProfiler.java PRE-CREATION samza-core/src/main/java/org/apache/samza/autoScaling/analyzer/MemoryAnalyzer.java PRE-CREATION samza-core/src/main/java/org/apache/samza/autoScaling/metrics/MemoryMetrics.java PRE-CREATION samza-core/src/main/java/org/apache/samza/autoScaling/metrics/SlidingWindowMetric.java PRE-CREATION samza-core/src/main/java/org/apache/samza/autoScaling/stream/AutoScalingMetricsSystemConsumer.java PRE-CREATION samza-core/src/main/scala/org/apache/samza/autoScaling/stream/AutoScalingMetricsSystemFactory.scala PRE-CREATION samza-core/src/main/scala/org/apache/samza/config/AutoScalingConfigRewriter.scala PRE-CREATION samza-core/src/main/scala/org/apache/samza/config/JobConfig.scala e4b14f4da6649eb78753ba3b3f529373b6f2dbe4 samza-core/src/main/scala/org/apache/samza/metrics/JvmMetrics.scala a95a0ecde300f6576fe46b37d5898e3d21634126 samza-core/src/main/scala/org/apache/samza/util/Util.scala 2feb65b729b45fbc3b83a75c4072527e3c4e60be samza-core/src/test/java/org/apache/samza/autoScaling/SlidingWindowMetricTest.java PRE-CREATION Diff: https://reviews.apache.org/r/36471/diff/ Testing --- Thanks, Shadi A. Noghabi
Re: Question about sub-projects and project merging
From peanut gallery; a. It looks to me that there is no overwhelming reason to merge the communities. In fact, IF it already was a single community, it might be time to split Samza out. Ask this question; If the active Samza devs lay down their tools, how many Kafka devs would care about (and further the dev of) Samza? b. Having hard dependency on another upstream project is common place in ASF. Take a look at the Hadoop echo system for many examples. c. To me, it sounds more like a technical issue of design, where Samza is more flexible than needed, perhaps because the original intent was to allow integration with more messaging systems than Kafka. Redesigning seems to be a driver, and that doesn't need to lead to merging the communities. d. Is there actually other underlying community issue? I haven't seen any worrying signs from Board reports, but I am asking anyway... These kind of questions often surface when the most active members of the community feel somewhat burned out and looking for other active devs to help out. Cheers Niclas On Mon, Jul 13, 2015 at 8:37 AM, Jay Kreps jay.kr...@gmail.com wrote: Hey board members, There is a longish thread on the Apache Samza mailing list on the relationship between Kafka and Samza and whether they wouldn't make a lot more sense as a single project. This raised some questions I was hoping to get advice on. Discussion thread (warning: super long, I attempt to summarize relevant bits below): http://mail-archives.apache.org/mod_mbox/samza-dev/201507.mbox/%3ccabyby7d_-jcxj7fizsjuebjedgbep33flyx3nrozt0yeox9...@mail.gmail.com%3E Anyhow, some people thought Apache has lot's of sub-projects, that would be a graceful way to step in the right direction. At that point others popped up and said, sub-projects are discouraged by the board. I'm not sure if we understand technically what a subproject is, but I think it means a second repo/committership under the same PMC. A few questions: - Is that what a sub-project is? - Are they discouraged? If so, why? - Assuming it makes sense in this case what is the process for making one? - Putting aside sub-projects as a mechanism what are examples where communities merged successfully? We were pointed towards Lucene/SOLR. Are there others? Relevant background info: - Samza depends on Kafka, but not vice versa - There is some overlap in committers but not extensive (3/11 Samza committers are also Kafka committers) Thanks for the advice! -Jay -- Niclas Hedhman, Software Developer http://zest.apache.org - New Energy for Java
Re: Question about sub-projects and project merging
some remarks on what a sub-project is? taken from my experience working on this exact topic for https://projects.apache.org/ first: see facts at https://projects.apache.org/projects.html?pmc for a complete list of projects (as documented by PMCs, then there are a lot of software that is not described) grouped by PMCs. I came to the conclusion that this is a question of semantic around project term, with 2 competing visions at ASF: - either you talk of TLPs + sub-projects - or you talk about committees + projects After trying both visions for https://projects.apache.org/ , which started on the TLP + sub-projects vision because TLP is pretty much used by all of us, I finally preferred committees + projects since it avoided the question of classifying projects in Top Level Projects and sub-projects, with the bad impression it puts on sub-ones, and the fact that in some committees, there is no project that is more top or sub: see Commons or Logging. But for some committees, there is really a main project and other projects are more like extensions or plugin: see Ant or Velocity IMHO, talking about committees and projects is the best way to avoid bad passion that comes from TLPs + sub-projects vision. With that terms, your question of merging 2 TLPs becomes merging 2 committees, ie their communities, and putting 2 projects under the management of this merged committee: IMHO, the description is more verbose but the debate is less passionated and focused on the main question = is this really the same community, then that should be managed by one committee only? I don't have any opinion on Kafka and Samza case: I just hope these explanations will help for the discussion. Regards, Hervé Le dimanche 12 juillet 2015 22:37:55 Jay Kreps a écrit : Hey board members, There is a longish thread on the Apache Samza mailing list on the relationship between Kafka and Samza and whether they wouldn't make a lot more sense as a single project. This raised some questions I was hoping to get advice on. Discussion thread (warning: super long, I attempt to summarize relevant bits below): http://mail-archives.apache.org/mod_mbox/samza-dev/201507.mbox/%3CCABYbY7d_- jcxj7fizsjuebjedgbep33flyx3nrozt0yeox9...@mail.gmail.com%3E Anyhow, some people thought Apache has lot's of sub-projects, that would be a graceful way to step in the right direction. At that point others popped up and said, sub-projects are discouraged by the board. I'm not sure if we understand technically what a subproject is, but I think it means a second repo/committership under the same PMC. A few questions: - Is that what a sub-project is? - Are they discouraged? If so, why? - Assuming it makes sense in this case what is the process for making one? - Putting aside sub-projects as a mechanism what are examples where communities merged successfully? We were pointed towards Lucene/SOLR. Are there others? Relevant background info: - Samza depends on Kafka, but not vice versa - There is some overlap in committers but not extensive (3/11 Samza committers are also Kafka committers) Thanks for the advice! -Jay