Re: HiveContext: cache table not supported for partitioned table?
Cache table works with partitioned table. I guess you’re experimenting with a default local metastore and the metastore_db directory doesn’t exist at the first place. In this case, all metastore tables/views don’t exist at first and will throw the error message you saw when the |PARTITIONS| metastore table is accessed for the first time by Hive client. However, you should also see this line before this error: 14/10/03 10:51:30 ERROR ObjectStore: Direct SQL failed, falling back to ORM And then the table is created on the fly. The cache operation is also performed normally. You can verify this by selecting it and check the Spark UI for cached RDDs. If you try to uncache the table and cache it again, you won’t see this error any more. Normally, in production environment you won’t see this error because metastore database is usually setup ahead of time. On 10/3/14 3:39 AM, Du Li wrote: Hi, In Spark 1.1 HiveContext, I ran a create partitioned table command followed by a cache table command and got a java.sql.SQLSyntaxErrorException: Table/View 'PARTITIONS' does not exist. But cache table worked fine if the table is not a partitioned table. Can anybody confirm that cache of partitioned table is not supported yet in current version? Thanks, Du
Re: What is the best way to build my developing Spark for testing on EC2?
I recommend using the data generators provided with MLlib to generate synthetic data for your scalability tests - provided they're well suited for your algorithms. They let you control things like number of examples and dimensionality of your dataset, as well as number of partitions. As far as cluster set up goes, I usually launch spot instances with the spark-ec2 scripts, and then check out a repo which contains a simple driver application for my code. Then I have something crude like bash scripts running my program and collecting output. You could have a look at the spark-perf repo if you want something a little better principled/automatic. - Evan > On Oct 2, 2014, at 5:37 PM, Yu Ishikawa wrote: > > Hi all, > > I am trying to contribute some machine learning algorithms to MLlib. > I must evaluate their performance on a cluster, changing input data > size, the number of CPU cores and any their parameters. > > I would like to build my develoipng Spark on EC2 automatically. > Is there already a building script for a developing version like spark-ec2 > script? > Or if you have any good idea to evaluate the performance of a developing > MLlib algorithm on a spark cluster like EC2, could you tell me? > > Best, > > > > - > -- Yu Ishikawa > -- > View this message in context: > http://apache-spark-developers-list.1001551.n3.nabble.com/What-is-the-best-way-to-build-my-developing-Spark-for-testing-on-EC2-tp8638.html > Sent from the Apache Spark Developers List mailing list archive at Nabble.com. > > - > To unsubscribe, e-mail: dev-unsubscr...@spark.apache.org > For additional commands, e-mail: dev-h...@spark.apache.org > - To unsubscribe, e-mail: dev-unsubscr...@spark.apache.org For additional commands, e-mail: dev-h...@spark.apache.org
What is the best way to build my developing Spark for testing on EC2?
Hi all, I am trying to contribute some machine learning algorithms to MLlib. I must evaluate their performance on a cluster, changing input data size, the number of CPU cores and any their parameters. I would like to build my develoipng Spark on EC2 automatically. Is there already a building script for a developing version like spark-ec2 script? Or if you have any good idea to evaluate the performance of a developing MLlib algorithm on a spark cluster like EC2, could you tell me? Best, - -- Yu Ishikawa -- View this message in context: http://apache-spark-developers-list.1001551.n3.nabble.com/What-is-the-best-way-to-build-my-developing-Spark-for-testing-on-EC2-tp8638.html Sent from the Apache Spark Developers List mailing list archive at Nabble.com. - To unsubscribe, e-mail: dev-unsubscr...@spark.apache.org For additional commands, e-mail: dev-h...@spark.apache.org
Re: EC2 clusters ready in launch time + 30 seconds
Thanks for the update, Nate. I'm looking forward to seeing how these projects turn out. David, Packer looks very, very interesting. I'm gonna look into it more next week. Nick On Thu, Oct 2, 2014 at 8:00 PM, Nate D'Amico wrote: > Bit of progress on our end, bit of lagging as well. Our guy leading > effort got little bogged down on client project to update hive/sql testbed > to latest spark/sparkSQL, also launching public service so we have been bit > scattered recently. > > Will have some more updates probably after next week. We are planning on > taking our client work around hive/spark, plus taking over the bigtop > automation work to modernize and get that fit for human consumption outside > or org. All our work and puppet modules will be open sourced, documented, > hopefully start to rally some other folks around effort that find it useful > > Side note, another effort we are looking into is gradle tests/support. We > have been leveraging serverspec for some basic infrastructure tests, but > with bigtop switching over to gradle builds/testing setup in 0.8 we want to > include support for that in our own efforts, probably some stuff that can > be learned and leveraged in spark world for repeatable/tested infrastructure > > If anyone has any specific automation questions to your environment you > can drop me a line directly.., will try to help out best I can. Else will > post update to dev list once we get on top of our own product release and > the bigtop work > > Nate > > > -Original Message- > From: David Rowe [mailto:davidr...@gmail.com] > Sent: Thursday, October 02, 2014 4:44 PM > To: Nicholas Chammas > Cc: dev; Shivaram Venkataraman > Subject: Re: EC2 clusters ready in launch time + 30 seconds > > I think this is exactly what packer is for. See e.g. > http://www.packer.io/intro/getting-started/build-image.html > > On a related note, the current AMI for hvm systems (e.g. m3.*, r3.*) has a > bad package for httpd, whcih causes ganglia not to start. For some reason I > can't get access to the raw AMI to fix it. > > On Fri, Oct 3, 2014 at 9:30 AM, Nicholas Chammas < > nicholas.cham...@gmail.com > > wrote: > > > Is there perhaps a way to define an AMI programmatically? Like, a > > collection of base AMI id + list of required stuff to be installed + > > list of required configuration changes. I’m guessing that’s what > > people use things like Puppet, Ansible, or maybe also AWS CloudFormation > for, right? > > > > If we could do something like that, then with every new release of > > Spark we could quickly and easily create new AMIs that have everything > we need. > > spark-ec2 would only have to bring up the instances and do a minimal > > amount of configuration, and the only thing we’d need to track in the > > Spark repo is the code that defines what goes on the AMI, as well as a > > list of the AMI ids specific to each release. > > > > I’m just thinking out loud here. Does this make sense? > > > > Nate, > > > > Any progress on your end with this work? > > > > Nick > > > > > > On Sun, Jul 13, 2014 at 8:53 PM, Shivaram Venkataraman < > > shiva...@eecs.berkeley.edu> wrote: > > > > > It should be possible to improve cluster launch time if we are > > > careful about what commands we run during setup. One way to do this > > > would be to walk down the list of things we do for cluster > > > initialization and see if there is anything we can do make things > > > faster. Unfortunately this might > > be > > > pretty time consuming, but I don't know of a better strategy. The > > > place > > to > > > start would be the setup.sh file at > > > https://github.com/mesos/spark-ec2/blob/v3/setup.sh > > > > > > Here are some things that take a lot of time and could be improved: > > > 1. Creating swap partitions on all machines. We could check if there > > > is a way to get EC2 to always mount a swap partition 2. Copying / > > > syncing things across slaves. The copy-dir script is called too many > > > times right now and each time it pauses for a few milliseconds > > > between slaves [1]. This could be improved by removing unnecessary > > > copies 3. We could make less frequently used modules like Tachyon, > > > persistent > > hdfs > > > not a part of the default setup. > > > > > > [1] https://github.com/mesos/spark-ec2/blob/v3/copy-dir.sh#L42 > > > > > > Thanks > > > Shivaram > > > > > > > > > > > > > > > On Sat, Jul 12, 2014 at 7:02 PM, Nicholas Chammas < > > > nicholas.cham...@gmail.com> wrote: > > > > > > > On Thu, Jul 10, 2014 at 8:10 PM, Nate D'Amico > > wrote: > > > > > > > > > Starting to work through some automation/config stuff for spark > > > > > stack > > > on > > > > > EC2 with a project, will be focusing the work through the apache > > bigtop > > > > > effort to start, can then share with spark community directly as > > things > > > > > progress if people are interested > > > > > > > > > > > > Let us know how that goes. I'm definitely interested in hearing more. > > > > > > > > Nick >
RE: EC2 clusters ready in launch time + 30 seconds
Bit of progress on our end, bit of lagging as well. Our guy leading effort got little bogged down on client project to update hive/sql testbed to latest spark/sparkSQL, also launching public service so we have been bit scattered recently. Will have some more updates probably after next week. We are planning on taking our client work around hive/spark, plus taking over the bigtop automation work to modernize and get that fit for human consumption outside or org. All our work and puppet modules will be open sourced, documented, hopefully start to rally some other folks around effort that find it useful Side note, another effort we are looking into is gradle tests/support. We have been leveraging serverspec for some basic infrastructure tests, but with bigtop switching over to gradle builds/testing setup in 0.8 we want to include support for that in our own efforts, probably some stuff that can be learned and leveraged in spark world for repeatable/tested infrastructure If anyone has any specific automation questions to your environment you can drop me a line directly.., will try to help out best I can. Else will post update to dev list once we get on top of our own product release and the bigtop work Nate -Original Message- From: David Rowe [mailto:davidr...@gmail.com] Sent: Thursday, October 02, 2014 4:44 PM To: Nicholas Chammas Cc: dev; Shivaram Venkataraman Subject: Re: EC2 clusters ready in launch time + 30 seconds I think this is exactly what packer is for. See e.g. http://www.packer.io/intro/getting-started/build-image.html On a related note, the current AMI for hvm systems (e.g. m3.*, r3.*) has a bad package for httpd, whcih causes ganglia not to start. For some reason I can't get access to the raw AMI to fix it. On Fri, Oct 3, 2014 at 9:30 AM, Nicholas Chammas wrote: > Is there perhaps a way to define an AMI programmatically? Like, a > collection of base AMI id + list of required stuff to be installed + > list of required configuration changes. I’m guessing that’s what > people use things like Puppet, Ansible, or maybe also AWS CloudFormation for, > right? > > If we could do something like that, then with every new release of > Spark we could quickly and easily create new AMIs that have everything we > need. > spark-ec2 would only have to bring up the instances and do a minimal > amount of configuration, and the only thing we’d need to track in the > Spark repo is the code that defines what goes on the AMI, as well as a > list of the AMI ids specific to each release. > > I’m just thinking out loud here. Does this make sense? > > Nate, > > Any progress on your end with this work? > > Nick > > > On Sun, Jul 13, 2014 at 8:53 PM, Shivaram Venkataraman < > shiva...@eecs.berkeley.edu> wrote: > > > It should be possible to improve cluster launch time if we are > > careful about what commands we run during setup. One way to do this > > would be to walk down the list of things we do for cluster > > initialization and see if there is anything we can do make things > > faster. Unfortunately this might > be > > pretty time consuming, but I don't know of a better strategy. The > > place > to > > start would be the setup.sh file at > > https://github.com/mesos/spark-ec2/blob/v3/setup.sh > > > > Here are some things that take a lot of time and could be improved: > > 1. Creating swap partitions on all machines. We could check if there > > is a way to get EC2 to always mount a swap partition 2. Copying / > > syncing things across slaves. The copy-dir script is called too many > > times right now and each time it pauses for a few milliseconds > > between slaves [1]. This could be improved by removing unnecessary > > copies 3. We could make less frequently used modules like Tachyon, > > persistent > hdfs > > not a part of the default setup. > > > > [1] https://github.com/mesos/spark-ec2/blob/v3/copy-dir.sh#L42 > > > > Thanks > > Shivaram > > > > > > > > > > On Sat, Jul 12, 2014 at 7:02 PM, Nicholas Chammas < > > nicholas.cham...@gmail.com> wrote: > > > > > On Thu, Jul 10, 2014 at 8:10 PM, Nate D'Amico > wrote: > > > > > > > Starting to work through some automation/config stuff for spark > > > > stack > > on > > > > EC2 with a project, will be focusing the work through the apache > bigtop > > > > effort to start, can then share with spark community directly as > things > > > > progress if people are interested > > > > > > > > > Let us know how that goes. I'm definitely interested in hearing more. > > > > > > Nick > > > > > > - To unsubscribe, e-mail: dev-unsubscr...@spark.apache.org For additional commands, e-mail: dev-h...@spark.apache.org
Re: EC2 clusters ready in launch time + 30 seconds
I think this is exactly what packer is for. See e.g. http://www.packer.io/intro/getting-started/build-image.html On a related note, the current AMI for hvm systems (e.g. m3.*, r3.*) has a bad package for httpd, whcih causes ganglia not to start. For some reason I can't get access to the raw AMI to fix it. On Fri, Oct 3, 2014 at 9:30 AM, Nicholas Chammas wrote: > Is there perhaps a way to define an AMI programmatically? Like, a > collection of base AMI id + list of required stuff to be installed + list > of required configuration changes. I’m guessing that’s what people use > things like Puppet, Ansible, or maybe also AWS CloudFormation for, right? > > If we could do something like that, then with every new release of Spark we > could quickly and easily create new AMIs that have everything we need. > spark-ec2 would only have to bring up the instances and do a minimal amount > of configuration, and the only thing we’d need to track in the Spark repo > is the code that defines what goes on the AMI, as well as a list of the AMI > ids specific to each release. > > I’m just thinking out loud here. Does this make sense? > > Nate, > > Any progress on your end with this work? > > Nick > > > On Sun, Jul 13, 2014 at 8:53 PM, Shivaram Venkataraman < > shiva...@eecs.berkeley.edu> wrote: > > > It should be possible to improve cluster launch time if we are careful > > about what commands we run during setup. One way to do this would be to > > walk down the list of things we do for cluster initialization and see if > > there is anything we can do make things faster. Unfortunately this might > be > > pretty time consuming, but I don't know of a better strategy. The place > to > > start would be the setup.sh file at > > https://github.com/mesos/spark-ec2/blob/v3/setup.sh > > > > Here are some things that take a lot of time and could be improved: > > 1. Creating swap partitions on all machines. We could check if there is a > > way to get EC2 to always mount a swap partition > > 2. Copying / syncing things across slaves. The copy-dir script is called > > too many times right now and each time it pauses for a few milliseconds > > between slaves [1]. This could be improved by removing unnecessary copies > > 3. We could make less frequently used modules like Tachyon, persistent > hdfs > > not a part of the default setup. > > > > [1] https://github.com/mesos/spark-ec2/blob/v3/copy-dir.sh#L42 > > > > Thanks > > Shivaram > > > > > > > > > > On Sat, Jul 12, 2014 at 7:02 PM, Nicholas Chammas < > > nicholas.cham...@gmail.com> wrote: > > > > > On Thu, Jul 10, 2014 at 8:10 PM, Nate D'Amico > wrote: > > > > > > > Starting to work through some automation/config stuff for spark stack > > on > > > > EC2 with a project, will be focusing the work through the apache > bigtop > > > > effort to start, can then share with spark community directly as > things > > > > progress if people are interested > > > > > > > > > Let us know how that goes. I'm definitely interested in hearing more. > > > > > > Nick > > > > > >
Re: EC2 clusters ready in launch time + 30 seconds
Is there perhaps a way to define an AMI programmatically? Like, a collection of base AMI id + list of required stuff to be installed + list of required configuration changes. I’m guessing that’s what people use things like Puppet, Ansible, or maybe also AWS CloudFormation for, right? If we could do something like that, then with every new release of Spark we could quickly and easily create new AMIs that have everything we need. spark-ec2 would only have to bring up the instances and do a minimal amount of configuration, and the only thing we’d need to track in the Spark repo is the code that defines what goes on the AMI, as well as a list of the AMI ids specific to each release. I’m just thinking out loud here. Does this make sense? Nate, Any progress on your end with this work? Nick On Sun, Jul 13, 2014 at 8:53 PM, Shivaram Venkataraman < shiva...@eecs.berkeley.edu> wrote: > It should be possible to improve cluster launch time if we are careful > about what commands we run during setup. One way to do this would be to > walk down the list of things we do for cluster initialization and see if > there is anything we can do make things faster. Unfortunately this might be > pretty time consuming, but I don't know of a better strategy. The place to > start would be the setup.sh file at > https://github.com/mesos/spark-ec2/blob/v3/setup.sh > > Here are some things that take a lot of time and could be improved: > 1. Creating swap partitions on all machines. We could check if there is a > way to get EC2 to always mount a swap partition > 2. Copying / syncing things across slaves. The copy-dir script is called > too many times right now and each time it pauses for a few milliseconds > between slaves [1]. This could be improved by removing unnecessary copies > 3. We could make less frequently used modules like Tachyon, persistent hdfs > not a part of the default setup. > > [1] https://github.com/mesos/spark-ec2/blob/v3/copy-dir.sh#L42 > > Thanks > Shivaram > > > > > On Sat, Jul 12, 2014 at 7:02 PM, Nicholas Chammas < > nicholas.cham...@gmail.com> wrote: > > > On Thu, Jul 10, 2014 at 8:10 PM, Nate D'Amico wrote: > > > > > Starting to work through some automation/config stuff for spark stack > on > > > EC2 with a project, will be focusing the work through the apache bigtop > > > effort to start, can then share with spark community directly as things > > > progress if people are interested > > > > > > Let us know how that goes. I'm definitely interested in hearing more. > > > > Nick > > >
HiveContext: cache table not supported for partitioned table?
Hi, In Spark 1.1 HiveContext, I ran a create partitioned table command followed by a cache table command and got a java.sql.SQLSyntaxErrorException: Table/View 'PARTITIONS' does not exist. But cache table worked fine if the table is not a partitioned table. Can anybody confirm that cache of partitioned table is not supported yet in current version? Thanks, Du