Regarding failure handling in storm-kafka

2014-09-29 Thread Hemanth Yamijala
Hi, We are using Apache-storm 0.9.2 and the storm-kafka (version 0.9.0-wip16a-scala292) which has support for Kafka 0.7. I am trying to understand the failure handling of Kafka spout in a particular scenario. I have 4 workers, 1 running 1 executor of the Kafka spout, 1 running 1 executor of Bolt

Re: Regarding failure handling in storm-kafka

2014-09-29 Thread Hemanth Yamijala
Hi, Few minutes after I sent out this mail, I noticed that the system recovered and all pending messages were processed. However, I can't explain why it took a long time to recover (almost 30 minutes after the erroneous bolt died.) One thing I noticed is that the Kafka spout re-established connec

Re: Rolling Upgrade from 0.91

2014-09-29 Thread Noam Cohen
Thanks Bobby! Noam On 25/09/14 16:35, Bobby Evans wrote: 0.9.1 to 0.9.2 is not possible to do a rolling upgrade. The data structures stored in Zk and on disk have changed in a non-backwards compatible way. Most EBFs we do rolling upgrades. For them we install the new yinst package on all of

Re: Trident Metrics Consumer

2014-09-29 Thread Yuval Oren
You can indeed register metrics in trident. In functions, override prepare(), which gives you a TridentOperationContext. In aggregators, use the IMetricsContext you get in makeState(). -- Yuval Oren N3TWORK > On Sep 26, 2014, at 4:21 PM, Raphael Hsieh wrote: > > I've been following the tutori

RE: storm-deploy

2014-09-29 Thread Gunderson, Richard-CW
Gezim, I manage a storm topology within our AWS env. However I use Chef, not lein. Not sure how much help I can be. To deploy a topology, I execute (from within Chef scripts) a Bash script which runs ../bin/storm kill and then ../bin/storm jar . That’s how I handle it, and so far it’s worked

Re: Trident Metrics Consumer

2014-09-29 Thread Yair Weinberger
On a side note, if you try to register a metric in IRichSpout used in Trident topology, you will get a RunTimeException. See https://issues.apache.org/jira/browse/STORM-197 On Mon, Sep 29, 2014 at 5:39 PM, Yuval Oren wrote: > You can indeed register metrics in trident. In functions, override >

Get the task id of a bolt in a topology

2014-09-29 Thread Nick Katsipoulakis
Hello, How can I get the task ID of a bolt in a topology? I need to get it in order to perform an emitDirect() from a Spout. Thanks, Nick

Re: Get the task id of a bolt in a topology

2014-09-29 Thread Nathan Leung
You get a handle to TopologyContext in the prepare method of your spout/bolt. http://nathanmarz.github.io/storm/doc/backtype/storm/task/TopologyContext.html#getThisTaskId() On Mon, Sep 29, 2014 at 11:37 AM, Nick Katsipoulakis wrote: > Hello, > > How can I get the task ID of a bolt in a topology

Re: Get the task id of a bolt in a topology

2014-09-29 Thread Nick Katsipoulakis
Thank you very much Nathan. Nick On Mon, Sep 29, 2014 at 11:43 AM, Nathan Leung wrote: > You get a handle to TopologyContext in the prepare method of your > spout/bolt. > > > http://nathanmarz.github.io/storm/doc/backtype/storm/task/TopologyContext.html#getThisTaskId() > > On Mon, Sep 29, 2014

Re: Trident Kafka Spout - Ack count increasing even though no messages are processed

2014-09-29 Thread Deepak Subhramanian
Thanks Nathan for the explanation. Is there a way to get the actual no of messages processed in trident from the storm gui or using the storm api ? On Sat, Sep 27, 2014 at 9:20 PM, Nathan Marz wrote: > Trident executes a batch every 500ms (by default). A batch involves a > bunch of coordination

Re: Trident Metrics Consumer

2014-09-29 Thread Raphael Hsieh
Thank you Yuval and Yair, this has been extremely helpful. I am now running in to a problem where Storm thinks that the metric has already been registered. I think this might be because of the parrallelism in my topology, but I'm not sure. Do you have any ideas as to why this might be? I'm receivin

Re: Trident Metrics Consumer

2014-09-29 Thread Raphael Hsieh
Actually nevermind, I think I figured it out. I was calling the same function twice haha. Thanks On Mon, Sep 29, 2014 at 1:19 PM, Raphael Hsieh wrote: > Thank you Yuval and Yair, this has been extremely helpful. > I am now running in to a problem where Storm thinks that the metric has > already

Metrics consumer vs new bolt

2014-09-29 Thread Raphael Hsieh
What is the benefit of using a metrics consumer and generating metrics, as opposed to just creating a new bolt to log metrics ? How does it affect the speed of the topology? from what I can tell, registering a metricsConsumer will effectively spawn a new bolt. Only this bolt has access to metrics

MetricsConsumer vs new Bolt

2014-09-29 Thread Raphael Hsieh
What is the benefit of using a metrics consumer and generating metrics, as opposed to just creating a new bolt to log metrics ? How does it affect the speed of the topology? from what I can tell, registering a metricsConsumer will effectively spawn a new bolt. Only this bolt has access to metrics

答复: Metrics consumer vs new bolt

2014-09-29 Thread vlhf刘海峰
Think about the convenience that storm builtin metric utils brought to you. You don’t need to handle time bucket yourself, or even calculate average. 发件人: Raphael Hsieh [mailto:raffihs...@gmail.com] 发送时间: 2014年9月30日 6:19 收件人: user@storm.apache.org 主题: Metrics consumer vs new bolt What is the ben