Fellas,
I have noticed a weird thing while working on upgrading Ignite in Zeppelin. One of the tests failed because of Caused by: class org.apache.ignite.spi.IgniteSpiException: Local node's marshaller differs from remote node's marshaller (to make sure all nodes in topology have identical marshaller, configure marshaller explicitly in configuration) [locMarshaller=org.apache.ignite.marshaller.optimized.OptimizedMarshaller, rmtMarshaller=org.apache.ignite.internal.binary.BinaryMarshaller, locNodeAddrs=[testing-worker-linux-docker-a1225458-3234-linux-7.prod.travis-ci.org/0:0:0:0:0:0:0:1%1, /127.0.0.1, /172.17.12.239], locPort=0, rmtNodeAddr=[testing-worker-linux-docker-a1225458-3234-linux-7.prod.travis-ci.org/0:0:0:0:0:0:0:1%1, /127.0.0.1, /172.17.12.239], locNodeId=8fa598a2-9986-4f29-ba8d-8f857048e227, rmtNodeId=3aa33408-afea-48c7-941c-8ea87429f44d] At the glance, it seems like running nodes in server and client modes somehow trigger different defaults for the marshallers, but I wasn't able to quickly find anything in the code proving my point. I've ended up forcing OptimizedMarshaller for the server node in the Zep's ignite/src/test/java/org/apache/zeppelin/ignite/IgniteSqlInterpreterTest.java and it solved the immediate issue. However, the situation seems worthy noting on dev@ Cheers, Cos |
In the interest of full disclosure, I have fixed the failing test as shown in
http://is.gd/HsrHGf but I am fairly sure it might/will pop-up in the actual deployment, because the client node will be started with OptimizedMarshaller anyway. Cos On Mon, Jan 04, 2016 at 08:40PM, Konstantin Boudnik wrote: > Fellas, > > I have noticed a weird thing while working on upgrading Ignite in Zeppelin. > One of the tests failed because of > > Caused by: class org.apache.ignite.spi.IgniteSpiException: Local node's > marshaller differs from remote node's marshaller (to make sure all nodes in > topology have identical marshaller, configure marshaller explicitly in > configuration) > [locMarshaller=org.apache.ignite.marshaller.optimized.OptimizedMarshaller, > rmtMarshaller=org.apache.ignite.internal.binary.BinaryMarshaller, > locNodeAddrs=[testing-worker-linux-docker-a1225458-3234-linux-7.prod.travis-ci.org/0:0:0:0:0:0:0:1%1, > /127.0.0.1, /172.17.12.239], locPort=0, > rmtNodeAddr=[testing-worker-linux-docker-a1225458-3234-linux-7.prod.travis-ci.org/0:0:0:0:0:0:0:1%1, > /127.0.0.1, /172.17.12.239], locNodeId=8fa598a2-9986-4f29-ba8d-8f857048e227, > rmtNodeId=3aa33408-afea-48c7-941c-8ea87429f44d] > > At the glance, it seems like running nodes in server and client modes somehow > trigger different defaults for the marshallers, but I wasn't able to quickly > find anything in the code proving my point. > > I've ended up forcing OptimizedMarshaller for the server node in the Zep's > ignite/src/test/java/org/apache/zeppelin/ignite/IgniteSqlInterpreterTest.java > and it solved the immediate issue. However, the situation seems worthy noting > on dev@ > > Cheers, > Cos > |
Cos, I think the proper fix would be to use the default binary marshaller
on both ends, no? On Mon, Jan 4, 2016 at 11:05 PM, Konstantin Boudnik <[hidden email]> wrote: > In the interest of full disclosure, I have fixed the failing test as shown > in > http://is.gd/HsrHGf but I am fairly sure it might/will pop-up in the > actual > deployment, because the client node will be started with > OptimizedMarshaller > anyway. > > Cos > > On Mon, Jan 04, 2016 at 08:40PM, Konstantin Boudnik wrote: > > Fellas, > > > > I have noticed a weird thing while working on upgrading Ignite in > Zeppelin. > > One of the tests failed because of > > > > Caused by: class org.apache.ignite.spi.IgniteSpiException: Local node's > > marshaller differs from remote node's marshaller (to make sure all > nodes in > > topology have identical marshaller, configure marshaller explicitly in > > configuration) > > > [locMarshaller=org.apache.ignite.marshaller.optimized.OptimizedMarshaller, > > rmtMarshaller=org.apache.ignite.internal.binary.BinaryMarshaller, > > locNodeAddrs=[ > testing-worker-linux-docker-a1225458-3234-linux-7.prod.travis-ci.org/0:0:0:0:0:0:0:1%1 > , > > /127.0.0.1, /172.17.12.239], locPort=0, > > rmtNodeAddr=[ > testing-worker-linux-docker-a1225458-3234-linux-7.prod.travis-ci.org/0:0:0:0:0:0:0:1%1 > , > > /127.0.0.1, /172.17.12.239], > locNodeId=8fa598a2-9986-4f29-ba8d-8f857048e227, > > rmtNodeId=3aa33408-afea-48c7-941c-8ea87429f44d] > > > > At the glance, it seems like running nodes in server and client modes > somehow > > trigger different defaults for the marshallers, but I wasn't able to > quickly > > find anything in the code proving my point. > > > > I've ended up forcing OptimizedMarshaller for the server node in the > Zep's > > > ignite/src/test/java/org/apache/zeppelin/ignite/IgniteSqlInterpreterTest.java > > and it solved the immediate issue. However, the situation seems worthy > noting > > on dev@ > > > > Cheers, > > Cos > > > > > |
In reply to this post by Konstantin Boudnik-2
I don't think server or client mode setting anyhow affects default
marshaller. Binary should be default for 1.5. Alex Goncharuk, can you please comment? --Yakov 2016-01-05 7:40 GMT+03:00 Konstantin Boudnik <[hidden email]>: > Fellas, > > I have noticed a weird thing while working on upgrading Ignite in Zeppelin. > One of the tests failed because of > > Caused by: class org.apache.ignite.spi.IgniteSpiException: Local node's > marshaller differs from remote node's marshaller (to make sure all nodes > in > topology have identical marshaller, configure marshaller explicitly in > configuration) > [locMarshaller=org.apache.ignite.marshaller.optimized.OptimizedMarshaller, > rmtMarshaller=org.apache.ignite.internal.binary.BinaryMarshaller, > locNodeAddrs=[ > testing-worker-linux-docker-a1225458-3234-linux-7.prod.travis-ci.org/0:0:0:0:0:0:0:1%1 > , > /127.0.0.1, /172.17.12.239], locPort=0, > rmtNodeAddr=[ > testing-worker-linux-docker-a1225458-3234-linux-7.prod.travis-ci.org/0:0:0:0:0:0:0:1%1 > , > /127.0.0.1, /172.17.12.239], > locNodeId=8fa598a2-9986-4f29-ba8d-8f857048e227, > rmtNodeId=3aa33408-afea-48c7-941c-8ea87429f44d] > > At the glance, it seems like running nodes in server and client modes > somehow > trigger different defaults for the marshallers, but I wasn't able to > quickly > find anything in the code proving my point. > > I've ended up forcing OptimizedMarshaller for the server node in the Zep's > > ignite/src/test/java/org/apache/zeppelin/ignite/IgniteSqlInterpreterTest.java > and it solved the immediate issue. However, the situation seems worthy > noting > on dev@ > > Cheers, > Cos > > |
Server/client mode does not (and should not) change the marshaller
settings, to my knowledge there are no such places in the code that might change this behavior. There might be something in the Zep integration test, but I need some time to install Zep and take a look at the code. |
I believe Andrey Gura introduced the Ignite-Zeppelin integration.
Andrey, can you please comment? D. On Tue, Jan 5, 2016 at 2:46 AM, Alexey Goncharuk <[hidden email] > wrote: > Server/client mode does not (and should not) change the marshaller > settings, to my knowledge there are no such places in the code that might > change this behavior. > > There might be something in the Zep integration test, but I need some time > to install Zep and take a look at the code. > |
In reply to this post by Alexey Goncharuk
Test is pretty simple and doesn't do anything crazy at least in my uneducated
opinion. To Yakov's point about using default binary marshaller: it is unclear why the client node won't use the default then. The test was working with 1.4 and then stopped. It hasn't been touched for a while, which tells me that something got changed in the underlying library. Thanks for looking into this! cos On Tue, Jan 05, 2016 at 01:46PM, Alexey Goncharuk wrote: > Server/client mode does not (and should not) change the marshaller > settings, to my knowledge there are no such places in the code that might > change this behavior. > > There might be something in the Zep integration test, but I need some time > to install Zep and take a look at the code. |
On Tue, Jan 5, 2016 at 8:12 PM, Konstantin Boudnik <[hidden email]> wrote:
> Test is pretty simple and doesn't do anything crazy at least in my > uneducated > opinion. > > To Yakov's point about using default binary marshaller: it is unclear why > the > client node won't use the default then. The test was working with 1.4 and > then > stopped. It hasn't been touched for a while, which tells me that something > got > changed in the underlying library. > Cos, I think you are right. Most likely the client code in Zeppelin sets up the marshaller. We need to update it to work with 1.5. We should also should allow to specify a default cache per notebook, so users could override the default cache setting in the interpreter. > > Thanks for looking into this! > cos > > On Tue, Jan 05, 2016 at 01:46PM, Alexey Goncharuk wrote: > > Server/client mode does not (and should not) change the marshaller > > settings, to my knowledge there are no such places in the code that might > > change this behavior. > > > > There might be something in the Zep integration test, but I need some > time > > to install Zep and take a look at the code. > |
On Tue, Jan 05, 2016 at 10:40PM, Dmitriy Setrakyan wrote:
> On Tue, Jan 5, 2016 at 8:12 PM, Konstantin Boudnik <[hidden email]> wrote: > > > Test is pretty simple and doesn't do anything crazy at least in my > > uneducated > > opinion. > > > > To Yakov's point about using default binary marshaller: it is unclear why > > the > > client node won't use the default then. The test was working with 1.4 and > > then > > stopped. It hasn't been touched for a while, which tells me that something > > got > > changed in the underlying library. > > > > Cos, I think you are right. Most likely the client code in Zeppelin sets up > the marshaller. We need to update it to work with 1.5. We should also > should allow to specify a default cache per notebook, so users could > override the default cache setting in the interpreter. set the marshaller from the test. But I let ppl more familiar with the matter to chime in. For now the test is fixed, although clumsily. > > Thanks for looking into this! > > cos > > > > On Tue, Jan 05, 2016 at 01:46PM, Alexey Goncharuk wrote: > > > Server/client mode does not (and should not) change the marshaller > > > settings, to my knowledge there are no such places in the code that might > > > change this behavior. > > > > > > There might be something in the Zep integration test, but I need some > > time > > > to install Zep and take a look at the code. > > |
Andrey Gura, can you please clarify?
--Yakov 2016-01-06 22:06 GMT+03:00 Konstantin Boudnik <[hidden email]>: > On Tue, Jan 05, 2016 at 10:40PM, Dmitriy Setrakyan wrote: > > On Tue, Jan 5, 2016 at 8:12 PM, Konstantin Boudnik <[hidden email]> > wrote: > > > > > Test is pretty simple and doesn't do anything crazy at least in my > > > uneducated > > > opinion. > > > > > > To Yakov's point about using default binary marshaller: it is unclear > why > > > the > > > client node won't use the default then. The test was working with 1.4 > and > > > then > > > stopped. It hasn't been touched for a while, which tells me that > something > > > got > > > changed in the underlying library. > > > > > > > Cos, I think you are right. Most likely the client code in Zeppelin sets > up > > the marshaller. We need to update it to work with 1.5. We should also > > should allow to specify a default cache per notebook, so users could > > override the default cache setting in the interpreter. > > Looking at the IgniteInterpreter code, again just briefly, there's no way > to > set the marshaller from the test. But I let ppl more familiar with the > matter > to chime in. For now the test is fixed, although clumsily. > > > > Thanks for looking into this! > > > cos > > > > > > On Tue, Jan 05, 2016 at 01:46PM, Alexey Goncharuk wrote: > > > > Server/client mode does not (and should not) change the marshaller > > > > settings, to my knowledge there are no such places in the code that > might > > > > change this behavior. > > > > > > > > There might be something in the Zep integration test, but I need some > > > time > > > > to install Zep and take a look at the code. > > > > |
Cos,
the problem is that default config (incubator-zeppelin/ignite/src/main/resources/default-ignite-jdbc.xml) contains `marshaller` bean definition of OptimizedMarshaller type. IgniteSqlInterpreter class uses this config. Just remove this bean definition and default marshaller will be used. On Mon, Jan 11, 2016 at 11:57 AM, Yakov Zhdanov <[hidden email]> wrote: > Andrey Gura, can you please clarify? > > --Yakov > > 2016-01-06 22:06 GMT+03:00 Konstantin Boudnik <[hidden email]>: > > > On Tue, Jan 05, 2016 at 10:40PM, Dmitriy Setrakyan wrote: > > > On Tue, Jan 5, 2016 at 8:12 PM, Konstantin Boudnik <[hidden email]> > > wrote: > > > > > > > Test is pretty simple and doesn't do anything crazy at least in my > > > > uneducated > > > > opinion. > > > > > > > > To Yakov's point about using default binary marshaller: it is unclear > > why > > > > the > > > > client node won't use the default then. The test was working with 1.4 > > and > > > > then > > > > stopped. It hasn't been touched for a while, which tells me that > > something > > > > got > > > > changed in the underlying library. > > > > > > > > > > Cos, I think you are right. Most likely the client code in Zeppelin > sets > > up > > > the marshaller. We need to update it to work with 1.5. We should also > > > should allow to specify a default cache per notebook, so users could > > > override the default cache setting in the interpreter. > > > > Looking at the IgniteInterpreter code, again just briefly, there's no way > > to > > set the marshaller from the test. But I let ppl more familiar with the > > matter > > to chime in. For now the test is fixed, although clumsily. > > > > > > Thanks for looking into this! > > > > cos > > > > > > > > On Tue, Jan 05, 2016 at 01:46PM, Alexey Goncharuk wrote: > > > > > Server/client mode does not (and should not) change the marshaller > > > > > settings, to my knowledge there are no such places in the code that > > might > > > > > change this behavior. > > > > > > > > > > There might be something in the Zep integration test, but I need > some > > > > time > > > > > to install Zep and take a look at the code. > > > > > > > -- Andrey Gura GridGain Systems, Inc. www.gridgain.com |
Andrey,
Any chance we could upgrade Zeppelin to work with Ignite 1.5? D. On Mon, Jan 11, 2016 at 7:44 AM, Andrey Gura <[hidden email]> wrote: > Cos, > > the problem is that default config > (incubator-zeppelin/ignite/src/main/resources/default-ignite-jdbc.xml) > contains `marshaller` bean definition of OptimizedMarshaller type. > IgniteSqlInterpreter class uses this config. > Just remove this bean definition and default marshaller will be used. > > On Mon, Jan 11, 2016 at 11:57 AM, Yakov Zhdanov <[hidden email]> > wrote: > > > Andrey Gura, can you please clarify? > > > > --Yakov > > > > 2016-01-06 22:06 GMT+03:00 Konstantin Boudnik <[hidden email]>: > > > > > On Tue, Jan 05, 2016 at 10:40PM, Dmitriy Setrakyan wrote: > > > > On Tue, Jan 5, 2016 at 8:12 PM, Konstantin Boudnik <[hidden email]> > > > wrote: > > > > > > > > > Test is pretty simple and doesn't do anything crazy at least in my > > > > > uneducated > > > > > opinion. > > > > > > > > > > To Yakov's point about using default binary marshaller: it is > unclear > > > why > > > > > the > > > > > client node won't use the default then. The test was working with > 1.4 > > > and > > > > > then > > > > > stopped. It hasn't been touched for a while, which tells me that > > > something > > > > > got > > > > > changed in the underlying library. > > > > > > > > > > > > > Cos, I think you are right. Most likely the client code in Zeppelin > > sets > > > up > > > > the marshaller. We need to update it to work with 1.5. We should also > > > > should allow to specify a default cache per notebook, so users could > > > > override the default cache setting in the interpreter. > > > > > > Looking at the IgniteInterpreter code, again just briefly, there's no > way > > > to > > > set the marshaller from the test. But I let ppl more familiar with the > > > matter > > > to chime in. For now the test is fixed, although clumsily. > > > > > > > > Thanks for looking into this! > > > > > cos > > > > > > > > > > On Tue, Jan 05, 2016 at 01:46PM, Alexey Goncharuk wrote: > > > > > > Server/client mode does not (and should not) change the > marshaller > > > > > > settings, to my knowledge there are no such places in the code > that > > > might > > > > > > change this behavior. > > > > > > > > > > > > There might be something in the Zep integration test, but I need > > some > > > > > time > > > > > > to install Zep and take a look at the code. > > > > > > > > > > > > > > -- > Andrey Gura > GridGain Systems, Inc. > www.gridgain.com > |
Dmitry,
If I understand correctly this ticket ( https://issues.apache.org/jira/browse/ZEPPELIN-545) assigned to Cos and has Fixed status. The fix is ok. But removing of marshaller bean definition is better solution. So I can provide new PR or we can wait for Cos. On Mon, Jan 11, 2016 at 8:55 PM, Dmitriy Setrakyan <[hidden email]> wrote: > Andrey, > > Any chance we could upgrade Zeppelin to work with Ignite 1.5? > > D. > > On Mon, Jan 11, 2016 at 7:44 AM, Andrey Gura <[hidden email]> wrote: > > > Cos, > > > > the problem is that default config > > (incubator-zeppelin/ignite/src/main/resources/default-ignite-jdbc.xml) > > contains `marshaller` bean definition of OptimizedMarshaller type. > > IgniteSqlInterpreter class uses this config. > > Just remove this bean definition and default marshaller will be used. > > > > On Mon, Jan 11, 2016 at 11:57 AM, Yakov Zhdanov <[hidden email]> > > wrote: > > > > > Andrey Gura, can you please clarify? > > > > > > --Yakov > > > > > > 2016-01-06 22:06 GMT+03:00 Konstantin Boudnik <[hidden email]>: > > > > > > > On Tue, Jan 05, 2016 at 10:40PM, Dmitriy Setrakyan wrote: > > > > > On Tue, Jan 5, 2016 at 8:12 PM, Konstantin Boudnik <[hidden email] > > > > > > wrote: > > > > > > > > > > > Test is pretty simple and doesn't do anything crazy at least in > my > > > > > > uneducated > > > > > > opinion. > > > > > > > > > > > > To Yakov's point about using default binary marshaller: it is > > unclear > > > > why > > > > > > the > > > > > > client node won't use the default then. The test was working with > > 1.4 > > > > and > > > > > > then > > > > > > stopped. It hasn't been touched for a while, which tells me that > > > > something > > > > > > got > > > > > > changed in the underlying library. > > > > > > > > > > > > > > > > Cos, I think you are right. Most likely the client code in Zeppelin > > > sets > > > > up > > > > > the marshaller. We need to update it to work with 1.5. We should > also > > > > > should allow to specify a default cache per notebook, so users > could > > > > > override the default cache setting in the interpreter. > > > > > > > > Looking at the IgniteInterpreter code, again just briefly, there's no > > way > > > > to > > > > set the marshaller from the test. But I let ppl more familiar with > the > > > > matter > > > > to chime in. For now the test is fixed, although clumsily. > > > > > > > > > > Thanks for looking into this! > > > > > > cos > > > > > > > > > > > > On Tue, Jan 05, 2016 at 01:46PM, Alexey Goncharuk wrote: > > > > > > > Server/client mode does not (and should not) change the > > marshaller > > > > > > > settings, to my knowledge there are no such places in the code > > that > > > > might > > > > > > > change this behavior. > > > > > > > > > > > > > > There might be something in the Zep integration test, but I > need > > > some > > > > > > time > > > > > > > to install Zep and take a look at the code. > > > > > > > > > > > > > > > > > > > > > -- > > Andrey Gura > > GridGain Systems, Inc. > > www.gridgain.com > > > -- Andrey Gura GridGain Systems, Inc. www.gridgain.com |
Andrey, I think Cos won’t mind if you will provide a new PR.
On Mon, Jan 11, 2016 at 10:08 AM, Andrey Gura <[hidden email]> wrote: > Dmitry, > > If I understand correctly this ticket ( > https://issues.apache.org/jira/browse/ZEPPELIN-545) assigned to Cos and > has > Fixed status. > > The fix is ok. But removing of marshaller bean definition is better > solution. So I can provide new PR or we can wait for Cos. > > > On Mon, Jan 11, 2016 at 8:55 PM, Dmitriy Setrakyan <[hidden email]> > wrote: > > > Andrey, > > > > Any chance we could upgrade Zeppelin to work with Ignite 1.5? > > > > D. > > > > On Mon, Jan 11, 2016 at 7:44 AM, Andrey Gura <[hidden email]> wrote: > > > > > Cos, > > > > > > the problem is that default config > > > (incubator-zeppelin/ignite/src/main/resources/default-ignite-jdbc.xml) > > > contains `marshaller` bean definition of OptimizedMarshaller type. > > > IgniteSqlInterpreter class uses this config. > > > Just remove this bean definition and default marshaller will be used. > > > > > > On Mon, Jan 11, 2016 at 11:57 AM, Yakov Zhdanov <[hidden email]> > > > wrote: > > > > > > > Andrey Gura, can you please clarify? > > > > > > > > --Yakov > > > > > > > > 2016-01-06 22:06 GMT+03:00 Konstantin Boudnik <[hidden email]>: > > > > > > > > > On Tue, Jan 05, 2016 at 10:40PM, Dmitriy Setrakyan wrote: > > > > > > On Tue, Jan 5, 2016 at 8:12 PM, Konstantin Boudnik < > [hidden email] > > > > > > > > wrote: > > > > > > > > > > > > > Test is pretty simple and doesn't do anything crazy at least in > > my > > > > > > > uneducated > > > > > > > opinion. > > > > > > > > > > > > > > To Yakov's point about using default binary marshaller: it is > > > unclear > > > > > why > > > > > > > the > > > > > > > client node won't use the default then. The test was working > with > > > 1.4 > > > > > and > > > > > > > then > > > > > > > stopped. It hasn't been touched for a while, which tells me > that > > > > > something > > > > > > > got > > > > > > > changed in the underlying library. > > > > > > > > > > > > > > > > > > > Cos, I think you are right. Most likely the client code in > Zeppelin > > > > sets > > > > > up > > > > > > the marshaller. We need to update it to work with 1.5. We should > > also > > > > > > should allow to specify a default cache per notebook, so users > > could > > > > > > override the default cache setting in the interpreter. > > > > > > > > > > Looking at the IgniteInterpreter code, again just briefly, there's > no > > > way > > > > > to > > > > > set the marshaller from the test. But I let ppl more familiar with > > the > > > > > matter > > > > > to chime in. For now the test is fixed, although clumsily. > > > > > > > > > > > > Thanks for looking into this! > > > > > > > cos > > > > > > > > > > > > > > On Tue, Jan 05, 2016 at 01:46PM, Alexey Goncharuk wrote: > > > > > > > > Server/client mode does not (and should not) change the > > > marshaller > > > > > > > > settings, to my knowledge there are no such places in the > code > > > that > > > > > might > > > > > > > > change this behavior. > > > > > > > > > > > > > > > > There might be something in the Zep integration test, but I > > need > > > > some > > > > > > > time > > > > > > > > to install Zep and take a look at the code. > > > > > > > > > > > > > > > > > > > > > > > > > > > > -- > > > Andrey Gura > > > GridGain Systems, Inc. > > > www.gridgain.com > > > > > > > > > -- > Andrey Gura > GridGain Systems, Inc. > www.gridgain.com > |
I'll just create new ticket and provide PR.
On Mon, Jan 11, 2016 at 9:11 PM, Dmitriy Setrakyan <[hidden email]> wrote: > Andrey, I think Cos won’t mind if you will provide a new PR. > > On Mon, Jan 11, 2016 at 10:08 AM, Andrey Gura <[hidden email]> wrote: > > > Dmitry, > > > > If I understand correctly this ticket ( > > https://issues.apache.org/jira/browse/ZEPPELIN-545) assigned to Cos and > > has > > Fixed status. > > > > The fix is ok. But removing of marshaller bean definition is better > > solution. So I can provide new PR or we can wait for Cos. > > > > > > On Mon, Jan 11, 2016 at 8:55 PM, Dmitriy Setrakyan < > [hidden email]> > > wrote: > > > > > Andrey, > > > > > > Any chance we could upgrade Zeppelin to work with Ignite 1.5? > > > > > > D. > > > > > > On Mon, Jan 11, 2016 at 7:44 AM, Andrey Gura <[hidden email]> > wrote: > > > > > > > Cos, > > > > > > > > the problem is that default config > > > > > (incubator-zeppelin/ignite/src/main/resources/default-ignite-jdbc.xml) > > > > contains `marshaller` bean definition of OptimizedMarshaller type. > > > > IgniteSqlInterpreter class uses this config. > > > > Just remove this bean definition and default marshaller will be used. > > > > > > > > On Mon, Jan 11, 2016 at 11:57 AM, Yakov Zhdanov <[hidden email] > > > > > > wrote: > > > > > > > > > Andrey Gura, can you please clarify? > > > > > > > > > > --Yakov > > > > > > > > > > 2016-01-06 22:06 GMT+03:00 Konstantin Boudnik <[hidden email]>: > > > > > > > > > > > On Tue, Jan 05, 2016 at 10:40PM, Dmitriy Setrakyan wrote: > > > > > > > On Tue, Jan 5, 2016 at 8:12 PM, Konstantin Boudnik < > > [hidden email] > > > > > > > > > > wrote: > > > > > > > > > > > > > > > Test is pretty simple and doesn't do anything crazy at least > in > > > my > > > > > > > > uneducated > > > > > > > > opinion. > > > > > > > > > > > > > > > > To Yakov's point about using default binary marshaller: it is > > > > unclear > > > > > > why > > > > > > > > the > > > > > > > > client node won't use the default then. The test was working > > with > > > > 1.4 > > > > > > and > > > > > > > > then > > > > > > > > stopped. It hasn't been touched for a while, which tells me > > that > > > > > > something > > > > > > > > got > > > > > > > > changed in the underlying library. > > > > > > > > > > > > > > > > > > > > > > Cos, I think you are right. Most likely the client code in > > Zeppelin > > > > > sets > > > > > > up > > > > > > > the marshaller. We need to update it to work with 1.5. We > should > > > also > > > > > > > should allow to specify a default cache per notebook, so users > > > could > > > > > > > override the default cache setting in the interpreter. > > > > > > > > > > > > Looking at the IgniteInterpreter code, again just briefly, > there's > > no > > > > way > > > > > > to > > > > > > set the marshaller from the test. But I let ppl more familiar > with > > > the > > > > > > matter > > > > > > to chime in. For now the test is fixed, although clumsily. > > > > > > > > > > > > > > Thanks for looking into this! > > > > > > > > cos > > > > > > > > > > > > > > > > On Tue, Jan 05, 2016 at 01:46PM, Alexey Goncharuk wrote: > > > > > > > > > Server/client mode does not (and should not) change the > > > > marshaller > > > > > > > > > settings, to my knowledge there are no such places in the > > code > > > > that > > > > > > might > > > > > > > > > change this behavior. > > > > > > > > > > > > > > > > > > There might be something in the Zep integration test, but I > > > need > > > > > some > > > > > > > > time > > > > > > > > > to install Zep and take a look at the code. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > -- > > > > Andrey Gura > > > > GridGain Systems, Inc. > > > > www.gridgain.com > > > > > > > > > > > > > > > -- > > Andrey Gura > > GridGain Systems, Inc. > > www.gridgain.com > > > -- Andrey Gura GridGain Systems, Inc. www.gridgain.com |
On Tue, Jan 12, 2016 at 1:54 AM, Andrey Gura <[hidden email]> wrote:
> I'll just create new ticket and provide PR. > Thanks! > > On Mon, Jan 11, 2016 at 9:11 PM, Dmitriy Setrakyan <[hidden email]> > wrote: > > > Andrey, I think Cos won’t mind if you will provide a new PR. > > > > On Mon, Jan 11, 2016 at 10:08 AM, Andrey Gura <[hidden email]> > wrote: > > > > > Dmitry, > > > > > > If I understand correctly this ticket ( > > > https://issues.apache.org/jira/browse/ZEPPELIN-545) assigned to Cos > and > > > has > > > Fixed status. > > > > > > The fix is ok. But removing of marshaller bean definition is better > > > solution. So I can provide new PR or we can wait for Cos. > > > > > > > > > On Mon, Jan 11, 2016 at 8:55 PM, Dmitriy Setrakyan < > > [hidden email]> > > > wrote: > > > > > > > Andrey, > > > > > > > > Any chance we could upgrade Zeppelin to work with Ignite 1.5? > > > > > > > > D. > > > > > > > > On Mon, Jan 11, 2016 at 7:44 AM, Andrey Gura <[hidden email]> > > wrote: > > > > > > > > > Cos, > > > > > > > > > > the problem is that default config > > > > > > > (incubator-zeppelin/ignite/src/main/resources/default-ignite-jdbc.xml) > > > > > contains `marshaller` bean definition of OptimizedMarshaller type. > > > > > IgniteSqlInterpreter class uses this config. > > > > > Just remove this bean definition and default marshaller will be > used. > > > > > > > > > > On Mon, Jan 11, 2016 at 11:57 AM, Yakov Zhdanov < > [hidden email] > > > > > > > > wrote: > > > > > > > > > > > Andrey Gura, can you please clarify? > > > > > > > > > > > > --Yakov > > > > > > > > > > > > 2016-01-06 22:06 GMT+03:00 Konstantin Boudnik <[hidden email]>: > > > > > > > > > > > > > On Tue, Jan 05, 2016 at 10:40PM, Dmitriy Setrakyan wrote: > > > > > > > > On Tue, Jan 5, 2016 at 8:12 PM, Konstantin Boudnik < > > > [hidden email] > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > > > > Test is pretty simple and doesn't do anything crazy at > least > > in > > > > my > > > > > > > > > uneducated > > > > > > > > > opinion. > > > > > > > > > > > > > > > > > > To Yakov's point about using default binary marshaller: it > is > > > > > unclear > > > > > > > why > > > > > > > > > the > > > > > > > > > client node won't use the default then. The test was > working > > > with > > > > > 1.4 > > > > > > > and > > > > > > > > > then > > > > > > > > > stopped. It hasn't been touched for a while, which tells me > > > that > > > > > > > something > > > > > > > > > got > > > > > > > > > changed in the underlying library. > > > > > > > > > > > > > > > > > > > > > > > > > Cos, I think you are right. Most likely the client code in > > > Zeppelin > > > > > > sets > > > > > > > up > > > > > > > > the marshaller. We need to update it to work with 1.5. We > > should > > > > also > > > > > > > > should allow to specify a default cache per notebook, so > users > > > > could > > > > > > > > override the default cache setting in the interpreter. > > > > > > > > > > > > > > Looking at the IgniteInterpreter code, again just briefly, > > there's > > > no > > > > > way > > > > > > > to > > > > > > > set the marshaller from the test. But I let ppl more familiar > > with > > > > the > > > > > > > matter > > > > > > > to chime in. For now the test is fixed, although clumsily. > > > > > > > > > > > > > > > > Thanks for looking into this! > > > > > > > > > cos > > > > > > > > > > > > > > > > > > On Tue, Jan 05, 2016 at 01:46PM, Alexey Goncharuk wrote: > > > > > > > > > > Server/client mode does not (and should not) change the > > > > > marshaller > > > > > > > > > > settings, to my knowledge there are no such places in the > > > code > > > > > that > > > > > > > might > > > > > > > > > > change this behavior. > > > > > > > > > > > > > > > > > > > > There might be something in the Zep integration test, > but I > > > > need > > > > > > some > > > > > > > > > time > > > > > > > > > > to install Zep and take a look at the code. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > -- > > > > > Andrey Gura > > > > > GridGain Systems, Inc. > > > > > www.gridgain.com > > > > > > > > > > > > > > > > > > > > > -- > > > Andrey Gura > > > GridGain Systems, Inc. > > > www.gridgain.com > > > > > > > > > -- > Andrey Gura > GridGain Systems, Inc. > www.gridgain.com > |
In reply to this post by Andrey Gura
Argh... the spring configs, of course ;( Thanks, will do.
Cos On Mon, Jan 11, 2016 at 06:44PM, Andrey Gura wrote: > Cos, > > the problem is that default config > (incubator-zeppelin/ignite/src/main/resources/default-ignite-jdbc.xml) > contains `marshaller` bean definition of OptimizedMarshaller type. > IgniteSqlInterpreter class uses this config. > Just remove this bean definition and default marshaller will be used. > > On Mon, Jan 11, 2016 at 11:57 AM, Yakov Zhdanov <[hidden email]> wrote: > > > Andrey Gura, can you please clarify? > > > > --Yakov > > > > 2016-01-06 22:06 GMT+03:00 Konstantin Boudnik <[hidden email]>: > > > > > On Tue, Jan 05, 2016 at 10:40PM, Dmitriy Setrakyan wrote: > > > > On Tue, Jan 5, 2016 at 8:12 PM, Konstantin Boudnik <[hidden email]> > > > wrote: > > > > > > > > > Test is pretty simple and doesn't do anything crazy at least in my > > > > > uneducated > > > > > opinion. > > > > > > > > > > To Yakov's point about using default binary marshaller: it is unclear > > > why > > > > > the > > > > > client node won't use the default then. The test was working with 1.4 > > > and > > > > > then > > > > > stopped. It hasn't been touched for a while, which tells me that > > > something > > > > > got > > > > > changed in the underlying library. > > > > > > > > > > > > > Cos, I think you are right. Most likely the client code in Zeppelin > > sets > > > up > > > > the marshaller. We need to update it to work with 1.5. We should also > > > > should allow to specify a default cache per notebook, so users could > > > > override the default cache setting in the interpreter. > > > > > > Looking at the IgniteInterpreter code, again just briefly, there's no way > > > to > > > set the marshaller from the test. But I let ppl more familiar with the > > > matter > > > to chime in. For now the test is fixed, although clumsily. > > > > > > > > Thanks for looking into this! > > > > > cos > > > > > > > > > > On Tue, Jan 05, 2016 at 01:46PM, Alexey Goncharuk wrote: > > > > > > Server/client mode does not (and should not) change the marshaller > > > > > > settings, to my knowledge there are no such places in the code that > > > might > > > > > > change this behavior. > > > > > > > > > > > > There might be something in the Zep integration test, but I need > > some > > > > > time > > > > > > to install Zep and take a look at the code. > > > > > > > > > > > > > > -- > Andrey Gura > GridGain Systems, Inc. > www.gridgain.com |
Commented on PR - you were faster ;) Thanks!
On Tue, Jan 12, 2016 at 12:47PM, Konstantin Boudnik wrote: > Argh... the spring configs, of course ;( Thanks, will do. > > Cos > > On Mon, Jan 11, 2016 at 06:44PM, Andrey Gura wrote: > > Cos, > > > > the problem is that default config > > (incubator-zeppelin/ignite/src/main/resources/default-ignite-jdbc.xml) > > contains `marshaller` bean definition of OptimizedMarshaller type. > > IgniteSqlInterpreter class uses this config. > > Just remove this bean definition and default marshaller will be used. > > > > On Mon, Jan 11, 2016 at 11:57 AM, Yakov Zhdanov <[hidden email]> wrote: > > > > > Andrey Gura, can you please clarify? > > > > > > --Yakov > > > > > > 2016-01-06 22:06 GMT+03:00 Konstantin Boudnik <[hidden email]>: > > > > > > > On Tue, Jan 05, 2016 at 10:40PM, Dmitriy Setrakyan wrote: > > > > > On Tue, Jan 5, 2016 at 8:12 PM, Konstantin Boudnik <[hidden email]> > > > > wrote: > > > > > > > > > > > Test is pretty simple and doesn't do anything crazy at least in my > > > > > > uneducated > > > > > > opinion. > > > > > > > > > > > > To Yakov's point about using default binary marshaller: it is unclear > > > > why > > > > > > the > > > > > > client node won't use the default then. The test was working with 1.4 > > > > and > > > > > > then > > > > > > stopped. It hasn't been touched for a while, which tells me that > > > > something > > > > > > got > > > > > > changed in the underlying library. > > > > > > > > > > > > > > > > Cos, I think you are right. Most likely the client code in Zeppelin > > > sets > > > > up > > > > > the marshaller. We need to update it to work with 1.5. We should also > > > > > should allow to specify a default cache per notebook, so users could > > > > > override the default cache setting in the interpreter. > > > > > > > > Looking at the IgniteInterpreter code, again just briefly, there's no way > > > > to > > > > set the marshaller from the test. But I let ppl more familiar with the > > > > matter > > > > to chime in. For now the test is fixed, although clumsily. > > > > > > > > > > Thanks for looking into this! > > > > > > cos > > > > > > > > > > > > On Tue, Jan 05, 2016 at 01:46PM, Alexey Goncharuk wrote: > > > > > > > Server/client mode does not (and should not) change the marshaller > > > > > > > settings, to my knowledge there are no such places in the code that > > > > might > > > > > > > change this behavior. > > > > > > > > > > > > > > There might be something in the Zep integration test, but I need > > > some > > > > > > time > > > > > > > to install Zep and take a look at the code. > > > > > > > > > > > > > > > > > > > > > -- > > Andrey Gura > > GridGain Systems, Inc. > > www.gridgain.com |
Free forum by Nabble | Edit this page |