Thanks for the links. The difficulty with building DR for HDFS is the distributed nature of HDFS. If each DataNode had a mirror copy in DR via something similar to SRDF (assuming NameNode and others taken care of), then there would not be an issue. The fail-over would be starting the mirror HDFS in DR site.
However, I agree with the points made that if your active cluster is busy, then the job becomes more challenging due to latency observed. Points also need to be observed that in an Enterprise like a Bank that Prod-DR WAN is shared among many applications some transactional (Oracle, Sybase , MSSQL) and others as DW including HDFS. May be a smart solution would be to replicate active partitions using streaming technologies and leave the dormant ones as they hardly change. However, we are still talking about potentially Terabytes of data through Gigabits WAN. The problem from my experience is that if you replicate few hundred Gigabytes of data daily, then you may just live with it. As your data grows, the task of streaming data is going to be much challenging. I have seen these issues with replicating large rows of CLOBS and BLOBS columns with Oracle and Sybase trying to push data from London to Singapore. It can become a nightmare. Dr Mich Talebzadeh LinkedIn * https://www.linkedin.com/profile/view?id=AAEAAAAWh2gBxianrbJd6zP6AcPCCdOABUrV8Pw <https://www.linkedin.com/profile/view?id=AAEAAAAWh2gBxianrbJd6zP6AcPCCdOABUrV8Pw>* http://talebzadehmich.wordpress.com *Disclaimer:* Use it at your own risk. Any and all responsibility for any loss, damage or destruction of data or any other property which may arise from relying on this email's technical content is explicitly disclaimed. The author will in no case be liable for any monetary damages arising from such loss, damage or destruction. On 12 November 2016 at 17:17, Timur Shenkao <t...@timshenkao.su> wrote: > Hi guys! > > 1) Though it's quite interesting, I believe that this discussion is not > about Spark :) > 2) If you are interested, there is solution by Cloudera > https://www.cloudera.com/documentation/enterprise/5-5- > x/topics/cm_bdr_replication_intro.html (requires that *source cluster* > has Cloudera Enterprise license, so it's not for free). > Correct me but I don't remember specialized replication solution by > Hortonworks (Atlas, Falcon, etc. are not precisely about inter-custer > replication). > Some solutions from Hadoop Ecosystem try to implement replication of > their own: https://cwiki.apache.org/confluence/pages/viewpage. > action?pageId=62687462 , http://highscalability.com/ > blog/2016/8/1/how-to-setup-a-highly-available-multi-az- > cassandra-cluster-o.html , > 3) Read this discussion https://community.hortonworks. > com/questions/29645/hdfs-replication-for-dr.html > 4) I prefer bash scripts / Python scripts / Oozie jobs + distcp - it's for > free & I control what's going on precisely. But, in case of huge clusters & > sophisticated logic, this approach become cumbersome. > 5) Don't forget about security & encryption: your sensitive data may be > read by third-party agents during replication > > On Sat, Nov 12, 2016 at 6:05 PM, Mich Talebzadeh < > mich.talebza...@gmail.com> wrote: > >> Thanks Jorn. >> >> The way WanDisco promotes itself is doing block level replication. as I >> understand you modify core-file.xml and add couple of network server >> locations there. they call this tool Fusion. there are at least 2 fusion >> servers for high availability. each one among other things has a database >> of its own. Once the client interacts with HDFS the fusion server behaves >> like a sniffer with its own port. As soon as the first HTFS block of >> 256MBout of say a file of 30GB is written, it starts sending that block to >> recipient. the laws of physics, the pipeline size etc applies here. That is >> up to the consumer. it can 10 files at the same time etc. so that is all. >> It is a known technology now labeled as streaming. so in summary it does >> not have to wait for the full file to be written to HDFS before replicating >> blocks. that is where it scores. >> >> It helps WAN work. Say the primary/active HDFS is in London and the >> replicate is in Singapore. so users in Singapore can see replicated data >> (eventually) when it gets there. It can obviously be used for DR in that >> case it is like Hot standby (borrowing a terminology from Sybase). In >> contrast one can do the same with period loads with homemade tools or tools >> like BDR from Cloudera. >> >> I mentioned that Hive is going to have its metastore on Hbase as well and >> that can be potential problems. The site is here >> <https://www.wandisco.com/> >> >> They are claiming there is no competitors in the market for their >> streaming HA product. >> >> HTH >> >> Dr Mich Talebzadeh >> >> >> >> LinkedIn * >> https://www.linkedin.com/profile/view?id=AAEAAAAWh2gBxianrbJd6zP6AcPCCdOABUrV8Pw >> <https://www.linkedin.com/profile/view?id=AAEAAAAWh2gBxianrbJd6zP6AcPCCdOABUrV8Pw>* >> >> >> >> http://talebzadehmich.wordpress.com >> >> >> *Disclaimer:* Use it at your own risk. Any and all responsibility for >> any loss, damage or destruction of data or any other property which may >> arise from relying on this email's technical content is explicitly >> disclaimed. The author will in no case be liable for any monetary damages >> arising from such loss, damage or destruction. >> >> >> >> On 12 November 2016 at 11:17, Jörn Franke <jornfra...@gmail.com> wrote: >> >>> What is wrong with the good old batch transfer for transferring data >>> from a cluster to another? I assume your use case is only business >>> continuity in case of disasters such as data center loss, which are >>> unlikely to happen (well it does not mean they do not happen) and where you >>> could afford to loose one day (or hour) of data (depends!). >>> >>> Nevertheless, I assume he refers to the Hadoop storage policies: >>> https://hadoop.apache.org/docs/current/hadoop-proj >>> ect-dist/hadoop-hdfs/ArchivalStorage.html , but this still only works >>> for the same cluster. >>> >>> You could also develop a custom secondary file system, similar to the >>> Ignite Cache filesystem, that sits on top of HDFS and as soon as it >>> receives data it sends them to another cluster and provides it to HDFS. Not >>> knowing Wandisco, I assume what it does. Given the prices (and the fact >>> that clusters tend to grow) you may want to evaluate if buying or making >>> makes sense. In any case, it also requires evaluation of network >>> throughput, because this may become the bottleneck somewhere (either within >>> the cluster or more likely between data centers). >>> >>> As you mentioned, Hbase & Co may require a special consideration for the >>> case that data is in-memory and not yet persisted. >>> >>> On Sat, Nov 12, 2016 at 12:04 PM, Mich Talebzadeh < >>> mich.talebza...@gmail.com> wrote: >>> >>>> thanks Vince >>>> >>>> can you provide more details on this pls >>>> >>>> Dr Mich Talebzadeh >>>> >>>> >>>> >>>> LinkedIn * >>>> https://www.linkedin.com/profile/view?id=AAEAAAAWh2gBxianrbJd6zP6AcPCCdOABUrV8Pw >>>> <https://www.linkedin.com/profile/view?id=AAEAAAAWh2gBxianrbJd6zP6AcPCCdOABUrV8Pw>* >>>> >>>> >>>> >>>> http://talebzadehmich.wordpress.com >>>> >>>> >>>> *Disclaimer:* Use it at your own risk. Any and all responsibility for >>>> any loss, damage or destruction of data or any other property which may >>>> arise from relying on this email's technical content is explicitly >>>> disclaimed. The author will in no case be liable for any monetary damages >>>> arising from such loss, damage or destruction. >>>> >>>> >>>> >>>> On 12 November 2016 at 09:52, vincent gromakowski < >>>> vincent.gromakow...@gmail.com> wrote: >>>> >>>>> A Hdfs tiering policy with good tags should be similar >>>>> >>>>> Le 11 nov. 2016 11:19 PM, "Mich Talebzadeh" <mich.talebza...@gmail.com> >>>>> a écrit : >>>>> >>>>>> I really don't see why one wants to set up streaming replication >>>>>> unless for situations where similar functionality to transactional >>>>>> databases is required in big data? >>>>>> >>>>>> Dr Mich Talebzadeh >>>>>> >>>>>> >>>>>> >>>>>> LinkedIn * >>>>>> https://www.linkedin.com/profile/view?id=AAEAAAAWh2gBxianrbJd6zP6AcPCCdOABUrV8Pw >>>>>> <https://www.linkedin.com/profile/view?id=AAEAAAAWh2gBxianrbJd6zP6AcPCCdOABUrV8Pw>* >>>>>> >>>>>> >>>>>> >>>>>> http://talebzadehmich.wordpress.com >>>>>> >>>>>> >>>>>> *Disclaimer:* Use it at your own risk. Any and all responsibility >>>>>> for any loss, damage or destruction of data or any other property which >>>>>> may >>>>>> arise from relying on this email's technical content is explicitly >>>>>> disclaimed. The author will in no case be liable for any monetary damages >>>>>> arising from such loss, damage or destruction. >>>>>> >>>>>> >>>>>> >>>>>> On 11 November 2016 at 17:24, Mich Talebzadeh < >>>>>> mich.talebza...@gmail.com> wrote: >>>>>> >>>>>>> I think it differs as it starts streaming data through its own port >>>>>>> as soon as the first block is landed. so the granularity is a block. >>>>>>> >>>>>>> however, think of it as oracle golden gate replication or sap >>>>>>> replication for databases. the only difference is that if the >>>>>>> corruption in >>>>>>> the block with hdfs it will be freplicated much like srdf. >>>>>>> >>>>>>> whereas with oracle or sap it is log based replication which stops >>>>>>> when it encounters corruption. >>>>>>> >>>>>>> replication depends on the block. so can replicate hive metadata and >>>>>>> fsimage etc. but cannot replicate hbase memstore if hbase crashes. >>>>>>> >>>>>>> so that is the gist of it. streaming replication as opposed to >>>>>>> snapshot. >>>>>>> >>>>>>> sounds familiar. think of it as log shipping in oracle old days >>>>>>> versus goldengate etc. >>>>>>> >>>>>>> hth >>>>>>> >>>>>>> Dr Mich Talebzadeh >>>>>>> >>>>>>> >>>>>>> >>>>>>> LinkedIn * >>>>>>> https://www.linkedin.com/profile/view?id=AAEAAAAWh2gBxianrbJd6zP6AcPCCdOABUrV8Pw >>>>>>> <https://www.linkedin.com/profile/view?id=AAEAAAAWh2gBxianrbJd6zP6AcPCCdOABUrV8Pw>* >>>>>>> >>>>>>> >>>>>>> >>>>>>> http://talebzadehmich.wordpress.com >>>>>>> >>>>>>> >>>>>>> *Disclaimer:* Use it at your own risk. Any and all responsibility >>>>>>> for any loss, damage or destruction of data or any other property which >>>>>>> may >>>>>>> arise from relying on this email's technical content is explicitly >>>>>>> disclaimed. The author will in no case be liable for any monetary >>>>>>> damages >>>>>>> arising from such loss, damage or destruction. >>>>>>> >>>>>>> >>>>>>> >>>>>>> On 11 November 2016 at 17:14, Deepak Sharma <deepakmc...@gmail.com> >>>>>>> wrote: >>>>>>> >>>>>>>> Reason being you can set up hdfs duplication on your own to some >>>>>>>> other cluster . >>>>>>>> >>>>>>>> On Nov 11, 2016 22:42, "Mich Talebzadeh" <mich.talebza...@gmail.com> >>>>>>>> wrote: >>>>>>>> >>>>>>>>> reason being ? >>>>>>>>> >>>>>>>>> Dr Mich Talebzadeh >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> LinkedIn * >>>>>>>>> https://www.linkedin.com/profile/view?id=AAEAAAAWh2gBxianrbJd6zP6AcPCCdOABUrV8Pw >>>>>>>>> <https://www.linkedin.com/profile/view?id=AAEAAAAWh2gBxianrbJd6zP6AcPCCdOABUrV8Pw>* >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> http://talebzadehmich.wordpress.com >>>>>>>>> >>>>>>>>> >>>>>>>>> *Disclaimer:* Use it at your own risk. Any and all responsibility >>>>>>>>> for any loss, damage or destruction of data or any other property >>>>>>>>> which may >>>>>>>>> arise from relying on this email's technical content is explicitly >>>>>>>>> disclaimed. The author will in no case be liable for any monetary >>>>>>>>> damages >>>>>>>>> arising from such loss, damage or destruction. >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> On 11 November 2016 at 17:11, Deepak Sharma <deepakmc...@gmail.com >>>>>>>>> > wrote: >>>>>>>>> >>>>>>>>>> This is waste of money I guess. >>>>>>>>>> >>>>>>>>>> On Nov 11, 2016 22:41, "Mich Talebzadeh" < >>>>>>>>>> mich.talebza...@gmail.com> wrote: >>>>>>>>>> >>>>>>>>>>> starts at $4,000 per node per year all inclusive. >>>>>>>>>>> >>>>>>>>>>> With discount it can be halved but we are talking a node itself >>>>>>>>>>> so if you have 5 nodes in primary and 5 nodes in DR we are talking >>>>>>>>>>> about >>>>>>>>>>> $40K already. >>>>>>>>>>> >>>>>>>>>>> HTH >>>>>>>>>>> >>>>>>>>>>> Dr Mich Talebzadeh >>>>>>>>>>> >>>>>>>>>>> >>>>>>>>>>> >>>>>>>>>>> LinkedIn * >>>>>>>>>>> https://www.linkedin.com/profile/view?id=AAEAAAAWh2gBxianrbJd6zP6AcPCCdOABUrV8Pw >>>>>>>>>>> <https://www.linkedin.com/profile/view?id=AAEAAAAWh2gBxianrbJd6zP6AcPCCdOABUrV8Pw>* >>>>>>>>>>> >>>>>>>>>>> >>>>>>>>>>> >>>>>>>>>>> http://talebzadehmich.wordpress.com >>>>>>>>>>> >>>>>>>>>>> >>>>>>>>>>> *Disclaimer:* Use it at your own risk. Any and all >>>>>>>>>>> responsibility for any loss, damage or destruction of data or any >>>>>>>>>>> other >>>>>>>>>>> property which may arise from relying on this email's technical >>>>>>>>>>> content is >>>>>>>>>>> explicitly disclaimed. The author will in no case be liable for any >>>>>>>>>>> monetary damages arising from such loss, damage or destruction. >>>>>>>>>>> >>>>>>>>>>> >>>>>>>>>>> >>>>>>>>>>> On 11 November 2016 at 16:43, Mudit Kumar <mkumar...@sapient.com >>>>>>>>>>> > wrote: >>>>>>>>>>> >>>>>>>>>>>> Is it feasible cost wise? >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> Thanks, >>>>>>>>>>>> >>>>>>>>>>>> Mudit >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> *From:* Mich Talebzadeh [mailto:mich.talebza...@gmail.com] >>>>>>>>>>>> *Sent:* Friday, November 11, 2016 2:56 PM >>>>>>>>>>>> *To:* user @spark >>>>>>>>>>>> *Subject:* Possible DR solution >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> Hi, >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> Has anyone had experience of using WanDisco >>>>>>>>>>>> <https://www.wandisco.com/> block replication to create a >>>>>>>>>>>> fault tolerant solution to DR in Hadoop? >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> The product claims that it starts replicating as soon as the >>>>>>>>>>>> first data block lands on HDFS and takes the block and sends it to >>>>>>>>>>>> DR/replicate site. The idea is that is faster than doing it through >>>>>>>>>>>> traditional HDFS copy tools which are normally batch oriented. >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> It also claims to replicate Hive metadata as well. >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> I wanted to gauge if anyone has used it or a competitor >>>>>>>>>>>> product. The claim is that they do not have competitors! >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> Thanks >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> Dr Mich Talebzadeh >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> LinkedIn >>>>>>>>>>>> *https://www.linkedin.com/profile/view?id=AAEAAAAWh2gBxianrbJd6zP6AcPCCdOABUrV8Pw >>>>>>>>>>>> <https://www.linkedin.com/profile/view?id=AAEAAAAWh2gBxianrbJd6zP6AcPCCdOABUrV8Pw>* >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> http://talebzadehmich.wordpress.com >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> *Disclaimer:* Use it at your own risk. Any and all >>>>>>>>>>>> responsibility for any loss, damage or destruction of data or any >>>>>>>>>>>> other >>>>>>>>>>>> property which may arise from relying on this email's technical >>>>>>>>>>>> content is >>>>>>>>>>>> explicitly disclaimed. The author will in no case be liable for any >>>>>>>>>>>> monetary damages arising from such loss, damage or destruction. >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>> >>>>>>>>>>> >>>>>>>>> >>>>>>> >>>>>> >>>> >>> >> >
