If its going to be a pain then we need to support writing the older class name in the FFT (trailer) and then use our internal mapping.
On Wed, Dec 7, 2016 at 3:22 PM, Anoop John <[email protected]> wrote: > I read the comments in the issue 16189.. This concern was raised > there already.. This is what I replied there. > "Ya we may need to make such a need. User on 1.x has to first upgrade > (rolling upgrade supported) to latest version in 1.x and then to 2.0. > This was discussed some other place also. I believe in the mail thread > where we discussed abt rolling upgrade support in 2.0" > > Not able to find the original mail thread which discussed abt rolling > upgrade support in 2.0 Pls correct if took some thing wrong way.. > Ya I agree that this way of indirection might be inconvenience for > users. > > > -Anoop- > > On Wed, Dec 7, 2016 at 3:13 PM, Ted Yu <[email protected]> wrote: > > bq. write old name only and within code we will have to map it with new > > name. > > > > This is a viable approach. > > Even if we document upgrading to 1.2.3+, some users may not perform this > > step before upgrading to 2.0 > > > > On Tue, Dec 6, 2016 at 10:04 PM, Anoop John <[email protected]> > wrote: > > > >> Ya that bug raised by Enis was very valid.. So this means when > >> rolling upgrade happens, if some of the other RSs with version <1.2.3 > >> , which is not having this fix, this issue might come up! > >> How to address then? Do we need to enforce a 1.2.3+ upgrade 1st and > >> then only 2.0 rolling upgrade? > >> > >> Or else we will need to fix in 2.0.. We write the new Comparator > >> class name in FFT (trunk code) To fix, we might have to write old > >> name only and within code we will have to map it with new name. It > >> will be ugly! It can be fixed only in 3.0 may be.. But that can make > >> the rolling upgrade story easier for users.. Just saying the > >> possibilities.. > >> > >> Thanks Ted to bring it up again. > >> > >> -Anoop- > >> > >> On Wed, Dec 7, 2016 at 10:04 AM, ramkrishna vasudevan > >> <[email protected]> wrote: > >> > I think when Enis reported the issue ( > >> > https://issues.apache.org/jira/browse/HBASE-16189), in rolling > upgrade > >> > regions could move around. So a region created in 2.0 moved to a > server > >> > with 1.x. > >> > > >> > Regards > >> > Ram > >> > > >> > > >> > On Wed, Dec 7, 2016 at 1:27 AM, Stack <[email protected]> wrote: > >> > > >> >> On Tue, Dec 6, 2016 at 10:19 AM, Ted Yu <[email protected]> wrote: > >> >> > >> >> > Looking at http://hbase.apache.org/book.html#executing.the.0.96. > >> upgrade > >> >> , > >> >> > there is step of running "bin/hbase upgrade -check" > >> >> > > >> >> > How about adding a sample hfile which contains > >> >> > CellComparator$MetaCellComparator > >> >> > so that the upgrade check can read and verify ? > >> >> > > >> >> > > >> >> > >> >> We don't support downgrade. Never have. > >> >> St.Ack > >> >> > >> >> > >> >> > >> >> > On Tue, Dec 6, 2016 at 8:50 AM, Ted Yu <[email protected]> > wrote: > >> >> > > >> >> > > The build I used yesterday didn't include HBASE-16189 > >> >> > > <https://issues.apache.org/jira/browse/HBASE-16189> > >> >> > > > >> >> > > Once it is included, the cluster can be downgraded fine. > >> >> > > > >> >> > > I wonder how users would know that their existing deployment has > >> >> > > HBASE-16189 <https://issues.apache.org/jira/browse/HBASE-16189> > >> before > >> >> > > upgrading to 2.0 release. > >> >> > > > >> >> > > On Tue, Dec 6, 2016 at 2:29 AM, ramkrishna vasudevan < > >> >> > > [email protected]> wrote: > >> >> > > > >> >> > >> @Ted > >> >> > >> Does your version have this fix > >> >> > >> https://issues.apache.org/jira/browse/HBASE-16189 > >> >> > >> > >> >> > >> Regards > >> >> > >> Ram > >> >> > >> > >> >> > >> On Tue, Dec 6, 2016 at 3:56 PM, Ted Yu <[email protected]> > >> wrote: > >> >> > >> > >> >> > >> > Is the assumption that hbase:meta would not split ? > >> >> > >> > > >> >> > >> > In other thread, Francis Liu was proposing supporting > splittable > >> >> > >> > hbase:meta in 2.0 release. > >> >> > >> > > >> >> > >> > Cheers > >> >> > >> > > >> >> > >> > > On Dec 6, 2016, at 2:20 AM, 张铎 <[email protected]> > wrote: > >> >> > >> > > > >> >> > >> > > Could this be solved by hosting meta only on master? > >> >> > >> > > > >> >> > >> > > BTW, MetaCellComparator is introduced in HBASE-10800. > >> >> > >> > > > >> >> > >> > > Thanks. > >> >> > >> > > > >> >> > >> > > 2016-12-06 17:44 GMT+08:00 Ted Yu <[email protected]>: > >> >> > >> > > > >> >> > >> > >> Hi, > >> >> > >> > >> When I restarted a cluster with 1.1 , I found that > hbase:meta > >> >> > region > >> >> > >> > >> (written to by the previously deployed 2.0) couldn't be > >> opened: > >> >> > >> > >> > >> >> > >> > >> Caused by: java.io.IOException: > >> >> > >> > >> org.apache.hadoop.hbase.io.hfile.CorruptHFileException: > >> Problem > >> >> > >> reading > >> >> > >> > >> HFile Trailer from file hdfs://yz1.xx.com:8020/apps/ > >> >> > >> hbase/data/data/ > >> >> > >> > >> hbase/meta/1588230740/info/599fc8a37311414e876803312009a9 > 86 > >> >> > >> > >> at > >> >> > >> > >> org.apache.hadoop.hbase.regionserver.HStore. > openStoreFiles( > >> >> > >> HStore.java: > >> >> > >> > >> 579) > >> >> > >> > >> at > >> >> > >> > >> org.apache.hadoop.hbase.regionserver.HStore. > loadStoreFiles( > >> >> > >> HStore.java: > >> >> > >> > >> 534) > >> >> > >> > >> at > >> >> > >> > >> org.apache.hadoop.hbase.regionserver.HStore.<init>( > >> >> > HStore.java:275) > >> >> > >> > >> at > >> >> > >> > >> org.apache.hadoop.hbase.regionserver.HRegion. > instantiateHSto > >> >> > >> re(HRegion. > >> >> > >> > >> java:5150) > >> >> > >> > >> at > >> >> > >> > >> org.apache.hadoop.hbase.regionserver.HRegion$1.call( > HRegion. > >> >> > >> java:912) > >> >> > >> > >> at > >> >> > >> > >> org.apache.hadoop.hbase.regionserver.HRegion$1.call( > HRegion. > >> >> > >> java:909) > >> >> > >> > >> at java.util.concurrent.FutureTask.run(FutureTask. > >> >> java:266) > >> >> > >> > >> at > >> >> > >> > >> java.util.concurrent.Executors$RunnableAdapter.call( > >> >> > >> Executors.java:511) > >> >> > >> > >> at java.util.concurrent.FutureTask.run(FutureTask. > >> >> java:266) > >> >> > >> > >> ... 3 more > >> >> > >> > >> Caused by: org.apache.hadoop.hbase.io. > >> >> hfile.CorruptHFileException: > >> >> > >> > Problem > >> >> > >> > >> reading HFile Trailer from file hdfs:// > >> >> > >> > >> yz1.xx.com:8020/apps/hbase/data/data/hbase/ > meta/1588230740/ > >> >> > >> > >> info/599fc8a37311414e876803312009a986 > >> >> > >> > >> at > >> >> > >> > >> org.apache.hadoop.hbase.io.hfile.HFile.pickReaderVersion( > >> >> > >> > HFile.java:483) > >> >> > >> > >> at > >> >> > >> > >> org.apache.hadoop.hbase.io.hfile.HFile.createReader( > >> >> > HFile.java:511) > >> >> > >> > >> at > >> >> > >> > >> org.apache.hadoop.hbase.regionserver.StoreFile$Reader. > >> >> > >> > >> <init>(StoreFile.java:1128) > >> >> > >> > >> at > >> >> > >> > >> org.apache.hadoop.hbase.regionserver.StoreFileInfo. > >> >> > >> > >> open(StoreFileInfo.java:267) > >> >> > >> > >> at > >> >> > >> > >> org.apache.hadoop.hbase.regionserver.StoreFile.open( > StoreFil > >> >> > >> e.java:409) > >> >> > >> > >> at > >> >> > >> > >> org.apache.hadoop.hbase.regionserver.StoreFile. > >> >> > >> > >> createReader(StoreFile.java:517) > >> >> > >> > >> at > >> >> > >> > >> org.apache.hadoop.hbase.regionserver.HStore. > createStoreFileA > >> >> > >> ndReader( > >> >> > >> > >> HStore.java:687) > >> >> > >> > >> at > >> >> > >> > >> org.apache.hadoop.hbase.regionserver.HStore.access$000( > >> >> > >> HStore.java:130) > >> >> > >> > >> at > >> >> > >> > >> org.apache.hadoop.hbase.regionserver.HStore$1.call( > >> >> > HStore.java:554) > >> >> > >> > >> at > >> >> > >> > >> org.apache.hadoop.hbase.regionserver.HStore$1.call( > >> >> > HStore.java:551) > >> >> > >> > >> ... 6 more > >> >> > >> > >> Caused by: java.io.IOException: java.lang. > >> >> ClassNotFoundException: > >> >> > >> > >> org.apache.hadoop.hbase.CellComparator$MetaCellComparator > >> >> > >> > >> at > >> >> > >> > >> org.apache.hadoop.hbase.io.hfile.FixedFileTrailer. > getCompara > >> >> > >> torClass( > >> >> > >> > >> FixedFileTrailer.java:581) > >> >> > >> > >> at > >> >> > >> > >> org.apache.hadoop.hbase.io.hfile.FixedFileTrailer. > >> >> > deserializeFromPB( > >> >> > >> > >> FixedFileTrailer.java:300) > >> >> > >> > >> at > >> >> > >> > >> org.apache.hadoop.hbase.io.hfile.FixedFileTrailer. > >> >> > >> > >> deserialize(FixedFileTrailer.java:242) > >> >> > >> > >> at > >> >> > >> > >> org.apache.hadoop.hbase.io.hfile.FixedFileTrailer. > >> >> readFromStream( > >> >> > >> > >> FixedFileTrailer.java:407) > >> >> > >> > >> at > >> >> > >> > >> org.apache.hadoop.hbase.io.hfile.HFile.pickReaderVersion( > >> >> > >> > HFile.java:468) > >> >> > >> > >> ... 15 more > >> >> > >> > >> Caused by: java.lang.ClassNotFoundException: > >> >> > >> > >> org.apache.hadoop.hbase.CellComparator$MetaCellComparator > >> >> > >> > >> at java.net.URLClassLoader. > >> findClass(URLClassLoader.java: > >> >> > 381) > >> >> > >> > >> at java.lang.ClassLoader. > loadClass(ClassLoader.java: > >> 424) > >> >> > >> > >> at sun.misc.Launcher$AppClassLoader.loadClass( > >> >> > Launcher.java: > >> >> > >> 331) > >> >> > >> > >> at java.lang.ClassLoader. > loadClass(ClassLoader.java: > >> 357) > >> >> > >> > >> at java.lang.Class.forName0(Native Method) > >> >> > >> > >> at java.lang.Class.forName(Class.java:264) > >> >> > >> > >> at > >> >> > >> > >> org.apache.hadoop.hbase.io.hfile.FixedFileTrailer. > getCompara > >> >> > >> torClass( > >> >> > >> > >> FixedFileTrailer.java:579) > >> >> > >> > >> > >> >> > >> > >> When user does rolling upgrade from 1.1 to 2.0, the above > may > >> >> cause > >> >> > >> > problem > >> >> > >> > >> if hbase:meta region is updated by server running 2.0 but > >> later > >> >> > >> > assigned to > >> >> > >> > >> a region server which still runs 1.1 (due to crash of the > >> server > >> >> > >> running > >> >> > >> > >> 2.0, e.g.) > >> >> > >> > >> > >> >> > >> > >> I want to get community feedback on the severity of this > >> issue. > >> >> > >> > >> > >> >> > >> > >> Thanks > >> >> > >> > >> > >> >> > >> > > >> >> > >> > >> >> > > > >> >> > > > >> >> > > >> >> > >> >
