Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

dev@ on BCC to prevent additional replies Please start a new thread for this so it gets the appropriate visibility. Beta 1 had been out for a while. On Feb 26, 2018 10:58 AM, "Jean-Marc Spaggiari" wrote: > Hum. "Broke" my cluster cluster again... > > 2018-02-26

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

Hum. "Broke" my cluster cluster again... 2018-02-26 13:54:44,053 WARN [ProcExecWrkr-14] assignment.RegionTransitionProcedure: Retryable error trying to transition: pid=409, ppid=344, state=RUNNABLE:REGION_TRANSITION_DISPATCH; UnassignProcedure table=page_crc,

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

Thanks JMS. S On Thu, Jan 11, 2018 at 9:36 AM, Jean-Marc Spaggiari < jean-m...@spaggiari.org> wrote: > Opened HBASE-19767 > and HBASE-19768. > Regarding the issue to create the log writer, it fails even if the DN is > already declared dead on

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

Opened HBASE-19767 and HBASE-19768. Regarding the issue to create the log writer, it fails even if the DN is already declared dead on the NN side... 2018-01-10 21:37 GMT-05:00 Apekshit Sharma : > On Wed, Jan 10, 2018 at 11:25

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

On Wed, Jan 10, 2018 at 11:25 AM, Zach York wrote: > What is the expectation for flaky tests? I was going to post some test > failures, but saw that they were included in the excludes for flaky tests. > > I understand we might be okay with having flaky tests for

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

What is the expectation for flaky tests? I was going to post some test failures, but saw that they were included in the excludes for flaky tests. I understand we might be okay with having flaky tests for this beta-1 (and obviously for dev), but I would assume that we want consistent test results

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

Put up a JIRA and dump this stuff in JMS. Sounds like we need a bit more test coverage at least. Thanks sir. St.Ack On Wed, Jan 10, 2018 at 2:52 AM, Jean-Marc Spaggiari < jean-m...@spaggiari.org> wrote: > The DN was dead since December 31st... I really hope the DN figured that > :-/ > > I will

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

The DN was dead since December 31st... I really hope the DN figured that :-/ I will retry with making sure that the NN is aware the local DN is dead, and see. I let you know. Thanks, JMS 2018-01-10 5:50 GMT-05:00 张铎(Duo Zhang) : > The problem maybe that the DN is dead,

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

The problem maybe that the DN is dead, but NN does not know and keep telling RS that you should try to connect to it. And for the new AsyncFSWAL, we need to connect to all the 3 DNs successfully before writing actual data to it, so the RS sucks... This maybe a problem. 2018-01-10 18:40

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

You're correct. It was dead. I thought HBase will be able to survive that. Same the DN dies after the RS has started, RS will fail closing nicely :( 2018-01-10 5:38 GMT-05:00 张铎(Duo Zhang) : > Connection refuse? Have you checked the status of the datanode on node8? > >

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

Connection refuse? Have you checked the status of the datanode on node8? 2018-01-10 18:31 GMT+08:00 Jean-Marc Spaggiari : > I know, this one sunk, but still running it on my cluster, so here is a new > issue I just got > > Any idea what this can be? I see this only a

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

Oh, interesting! If the local DN is dead, HBase can not start... I will have expected it to just used HDFS and any other node... That's why my HBase was not able to start. Same, if the DN dies, HBase will not be able to stop. Should we not be able to survive one DN failure? JM 2018-01-10 5:31

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

I know, this one sunk, but still running it on my cluster, so here is a new issue I just got Any idea what this can be? I see this only a one of my nodes... 2018-01-10 05:22:55,786 WARN [regionserver/node8.com/192.168.23.2:16020] wal.AsyncFSWAL: create wal log writer hdfs://

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

I just vetoed the RC because TestMemstoreLABWithoutPool always fails for me. It was the same with the last RC too. My Java is Oracle Java 8u144 running on x64 Linux (Ubuntu xenial). Let me know if you need me to provide the test output. On Tue, Jan 9, 2018 at 9:31 AM, Stack

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

On Sun, Jan 7, 2018 at 3:14 AM, Jean-Marc Spaggiari wrote: > Ok, thanks Stack. I will keep it running all day long until I get a > successful one. Is that useful that I report all the failed? Or just a wast > of time? Here is the last failed: > > [INFO] Results: >

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

Hope we still have time to get the procedure based replication peer modification(HBASE-19397) in before cutting branch-2.0... :( 2018-01-07 21:51 GMT+08:00 Stack : > On Sun, Jan 7, 2018 at 12:55 AM, Apekshit Sharma > wrote: > > > bq. Don't you think we have

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

On Sun, Jan 7, 2018 at 12:55 AM, Apekshit Sharma wrote: > bq. Don't you think we have enough branches already mighty Appy? > Yeah we do...sigh. > > > idk about that. But don't we need a *patch* branch branch-2.0 (just like > branch-1.4) where we "make backwards-compatible bug

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

Excellent! Thanks again! Starting again with the tests... JMS 2018-01-07 8:04 GMT-05:00 张铎(Duo Zhang) : > The last '-fn' option in the mvn command does that magic for you. > > 2018-01-07 19:03 GMT+08:00 Jean-Marc Spaggiari : > > > So that's the

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

The last '-fn' option in the mvn command does that magic for you. 2018-01-07 19:03 GMT+08:00 Jean-Marc Spaggiari : > So that's the way! Super. Thanks 张铎. Last, is there a way to keep going > with the remaining tests even if we get a failure on a test? > > JMS > >

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

So that's the way! Super. Thanks 张铎. Last, is there a way to keep going with the remaining tests even if we get a failure on a test? JMS 2018-01-07 5:56 GMT-05:00 张铎(Duo Zhang) : > You can try to copy the command line from the pre commit job where we will > bypass the

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

You can try to copy the command line from the pre commit job where we will bypass the flakey tests... This is the command I use to run UTs mvn -PrunAllTests

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

Ok, thanks Stack. I will keep it running all day long until I get a successful one. Is that useful that I report all the failed? Or just a wast of time? Here is the last failed: [INFO] Results: [INFO] [ERROR] Failures: [ERROR] TestFromClientSide.testCheckAndDeleteWithCompareOp:4982 expected:

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

bq. Don't you think we have enough branches already mighty Appy? Yeah we do...sigh. idk about that. But don't we need a *patch* branch branch-2.0 (just like branch-1.4) where we "make backwards-compatible bug fixes" and a *minor* branch branch-2 where we "add functionality in a

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

It is not you. There are a bunch of flies we need to fix. This latter is for sure flakey. Let me take a look. Thanks, JMS. S On Jan 6, 2018 5:57 PM, "Jean-Marc Spaggiari" wrote: I might not doing the right magic to get that run If someone is able to get all the

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

I might not doing the right magic to get that run If someone is able to get all the tests pass, can you please share the command you run? Thanks, JMS [INFO] Results: [INFO] [ERROR] Failures: [ERROR] TestFromClientSide.testCheckAndDeleteWithCompareOp:4982 expected: but was: [ERROR]

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

Deleted the class to get all the tests running. Was running on the RC1 from the tar. I know get those one failing. [ERROR] Failures: [ERROR] TestFavoredStochasticLoadBalancer.test2FavoredNodesDead:352 Balancer did not run [ERROR]

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

Looks like you didn't include HBASE-19666 which would be in the next RC. On Sat, Jan 6, 2018 at 10:52 AM, Jean-Marc Spaggiari < jean-m...@spaggiari.org> wrote: > Trying with a different command line (mvn test -P runAllTests > -Dsurefire.secondPartThreadCount=12

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

I can reproduce the issue locally. I think it has to do with the java version being used - IIRC this is related to the version of java used, but we can discuss in more detail on the JIRA. https://issues.apache.org/jira/browse/HBASE-19721 Thanks, JMS! On Sat, Jan 6, 2018 at 6:42 AM, Jean-Marc

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

How you guys are able to get the tests running? For me it keeps failing on TestReversedScannerCallable. I tried many times, always fails in the same place. I'm running on a 4GB tmpfs. Details are below. Am I doing something wrong? JM ./dev-support/hbasetests.sh runAllTests [INFO] Running

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

On Jan 5, 2018 4:44 PM, "Apekshit Sharma" wrote: bq. Care needs to be exercised backporting. Bug fixes only please. If in doubt, ping me, the RM, please. Thanks. In that case, shouldn't we branch out branch-2.0? We can then do normal backports to branch-2 and only bug fixes to

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

bq. Care needs to be exercised backporting. Bug fixes only please. If in doubt, ping me, the RM, please. Thanks. In that case, shouldn't we branch out branch-2.0? We can then do normal backports to branch-2 and only bug fixes to branch-2.0. On Fri, Jan 5, 2018 at 9:48 AM, Andrew Purtell

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

TestMemstoreLABWithoutPool is a flake, not a consistent fail. On Fri, Jan 5, 2018 at 7:18 AM, Stack wrote: > On Thu, Jan 4, 2018 at 2:24 PM, Andrew Purtell > wrote: > > > This one is probably my fault: > > > > TestDefaultCompactSelection > > > >

[RESULT][VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

Lets fail this RC because it has a broken unit test (Fixed subsequently in HBASE-19666). As per Andrew, unit tests should be passing by beta-time. JMS damaged his cluster starting w/ an ill-configured SNAPPY. I opened HBASE-19701 for beta-2 to look into better messaging around his failure state

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

On Thu, Jan 4, 2018 at 2:24 PM, Andrew Purtell wrote: > This one is probably my fault: > > TestDefaultCompactSelection > > HBASE-19406 > > Balazs fixed it above, HBASE-19666 > It can easily be reverted. The failure of interest > is

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

On Thu, Jan 4, 2018 at 12:39 PM, Jean-Marc Spaggiari < jean-m...@spaggiari.org> wrote: > If I re-run from the original cluster, now that I have snappy enabled, it > works. But if it helps I can easily remove snappy libs, transfer from > source, re-run and capture all the logs. It's an easy step.

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

This one is probably my fault: TestDefaultCompactSelection HBASE-19406 It can easily be reverted. The failure of interest is TestMemstoreLABWithoutPool.testLABChunkQueueWithMultipleMSLABs. On Thu, Jan 4, 2018 at 12:22 PM, Andrew Purtell wrote: > Should all unit tests

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

Should all unit tests pass on a beta? I think so, at least if the failures are 100% repeatable. -0 Checked sums and signatures: ok RAT check: ok Built from source: ok (8u144) Ran unit tests: some failures (8u144) [ERROR]

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

If I re-run from the original cluster, now that I have snappy enabled, it works. But if it helps I can easily remove snappy libs, transfer from source, re-run and capture all the logs. It's an easy step. Just confirm and I will do it. Apart from that, everything else seems to run correctly. I ran

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

+1 from me. S On Fri, Dec 29, 2017 at 12:15 PM, Stack wrote: > The first release candidate for HBase 2.0.0-beta-1 is up at: > > https://dist.apache.org/repos/dist/dev/hbase/hbase-2.0.0-beta-1-RC0/ > > Maven artifacts are available from a staging directory here: > >

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

On Sun, Dec 31, 2017 at 7:23 AM, Jean-Marc Spaggiari < jean-m...@spaggiari.org> wrote: > Nothing bad that I can see. Here is a region server log: > https://pastebin.com/0r76Y6ap > > Good one JMS. This log has "nothing" about why we decide to close the Region post successful open (If it was a

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

Ok. Lets see if there other issues out there. Hopefully we can find something more substantial than a failing unit test as reason for sinking an rc. Thanks balazs, S On Jan 3, 2018 8:53 AM, "Balazs Meszaros" wrote: > Only that one failed. > > Balazs > > On Wed,

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

Only that one failed. Balazs On Wed, Jan 3, 2018 at 3:48 PM, stack wrote: > Thanks balazs. Did you find other failing tests or just this one? > > S > > On Jan 3, 2018 5:33 AM, "Balazs Meszaros" > wrote: > > > My observations: > > > > -

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

Thanks balazs. Did you find other failing tests or just this one? S On Jan 3, 2018 5:33 AM, "Balazs Meszaros" wrote: > My observations: > > - signatures checksums are ok > - shell worked > - load test tool with read/write also worked > - when I built it,

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

My observations: - signatures checksums are ok - shell worked - load test tool with read/write also worked - when I built it, TestDefaultCompactSelection failed. The fix is already here: HBASE-19666 On Mon, Jan 1, 2018 at 11:30 PM, stack

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

Yes. Of course. Need your input lads. S On Jan 1, 2018 3:15 PM, "Andrew Purtell" wrote: > Seconded. I’ll be back later this week. Can try it out then? > > > > On Jan 1, 2018, at 12:13 PM, Mike Drob wrote: > > > > Is an extension here a reasonable

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

Seconded. I’ll be back later this week. Can try it out then? > On Jan 1, 2018, at 12:13 PM, Mike Drob wrote: > > Is an extension here a reasonable ask? Putting the vote up right before > what is a long New Year weekend for many folks doesn't give a lot of > opportunity for

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

Is an extension here a reasonable ask? Putting the vote up right before what is a long New Year weekend for many folks doesn't give a lot of opportunity for thorough review. Mike On Mon, Jan 1, 2018 at 1:30 PM, stack wrote: > This is great stuff jms. Thank you. Away from

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

This is great stuff jms. Thank you. Away from computer at mo but will dig in. Is it possible old files left over written with old hbase with old hfile version? Can you see on source? They should have but updated by a compaction if a long time idle, I agree. Yeah. If region assign fails, and

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

Sorry to spam the list :( Another interesting thing. Now most of my tablesare online. For few I'm getting this: Caused by: java.lang.IllegalArgumentException: Invalid HFile version: major=2, minor=1: expected at least major=2 and minor=3 at

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

Ok. With a brand new DestCP from source cluster, regions are getting assigned correctly. So sound like if they get stuck initially for any reason, then even if the reason is fixed they can not get assigned anymore again. Will keep playing. I kept the previous /hbase just in case we need something

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

Nothing bad that I can see. Here is a region server log: https://pastebin.com/0r76Y6ap Disabling the table makes the regions leave the transition mode. I'm trying to disable all tables one by one (because it get stuck after each disable) and will see if re-enabling them helps... On the master

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

> Good questions. If you disable snappy does it work? See below. I don't think it's related to snappy anymore. > If you start over fresh does it work? DistCP in progress. Will let you know in 4 hours... > It should be picking up native libs. Make an issue please jms. Thanks for giving it a

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

There is nothing further up in the master log from regionservers or on regionservers side on open? Thanks, S On Dec 31, 2017 8:37 AM, "stack" wrote: > Good questions. If you disable snappy does it work? If you start over > fresh does it work? It should be picking up

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

Good questions. If you disable snappy does it work? If you start over fresh does it work? It should be picking up native libs. Make an issue please jms. Thanks for giving it a go. S On Dec 30, 2017 11:49 PM, "Jean-Marc Spaggiari" wrote: > Hi Stack, > > I just

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

I forgot to say that I distCP the entire /hbase folder from another 1.3 HBase cluster ;) That's why there is data here. 2017-12-31 0:48 GMT-05:00 Jean-Marc Spaggiari : > Hi Stack, > > I just tried to give it a try... Wipe out all HDFS content and code, all > HBase

Re: [VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

Hi Stack, I just tried to give it a try... Wipe out all HDFS content and code, all HBase content and code, and all ZK. Re-build a brand new cluster with 7 physical worker nodes. I'm able to get HBase start, how-ever I'm not able to get my regions online. 2017-12-31 00:42:03,187 WARN

[VOTE] The first hbase-2.0.0-beta-1 Release Candidate is available

The first release candidate for HBase 2.0.0-beta-1 is up at: https://dist.apache.org/repos/dist/dev/hbase/hbase-2.0.0-beta-1-RC0/ Maven artifacts are available from a staging directory here: https://repository.apache.org/content/repositories/orgapachehbase-1188 All was signed with my key at