Re: RC for 4.3.1?
Are you going to cut from 4.3.1 from master or branch-4.3? IMO, since there's been no big feature nor breaking change, it should come from master (and branch-4.3 should be deleted until it is actually needed). -Ivan On Wed, May 13, 2015 at 9:06 PM, Sijie Guo guosi...@gmail.com wrote: yes. I will cut the new RC candidate. - Sijie On Wed, May 13, 2015 at 6:13 AM, Flavio Junqueira fpjunque...@yahoo.com wrote: +1 On Wednesday, May 13, 2015 9:57 AM, Rakesh R rake...@huawei.com wrote: Hi Sijie, BOOKKEEPER-836 has been resolved. Now, we have only BOOKKEEPER-835 is marked for this release and we have got two +1s and no -1 Shall we go ahead with the release ? Best Regards, Rakesh -Original Message- From: Flavio Junqueira [mailto:fpjunque...@yahoo.com] Sent: 21 March 2015 03:34 To: Rakesh R Cc: dev@bookkeeper.apache.org; Sijie Guo Subject: Re: RC for 4.3.1? I have uploaded the logs to BK-846. Having the timeout set to 3s might not be enough, but I'd like to understand if it is really necessary to increase rather than increasing arbitrarily. I have checked that it doesn't fail consistently. -Flavio On 20 Mar 2015, at 12:10, Rakesh R rake...@huawei.com wrote: Hi Flavio, I have just noticed one thing, it is configured 3000 milliseconds timeout. That is too small value. Can you please increase to @Test(timeout = 6) and verify the test case again. Regards, Rakesh -Original Message- From: Rakesh R Sent: 20 March 2015 15:50 To: dev@bookkeeper.apache.org Cc: Sijie Guo Subject: RE: RC for 4.3.1? Hi Flavio, testShouldGetTwoFrgamentsIfTwoBookiesFailedInSameEnsemble(org.a p ache.bookkeeper.client.TestLedgerChecker): test timed out after 3000 milliseconds I could see the following call can take some amount of time SetLedgerFragment result = getUnderReplicatedFragments(lh); I think, will get some hint if you can get the logs and do the analysis. Do you have the logs available with you. Regards, Rakesh -Original Message- From: Flavio Junqueira [mailto:fpjunque...@yahoo.com.INVALID] Sent: 20 March 2015 13:51 To: Sijie Guo Cc: dev@bookkeeper.apache.org Subject: Re: RC for 4.3.1? I've actually been able to get most of the tests to pass by adding an entry to /etc/hosts. I got only different test failure this time around: testShouldGetTwoFrgamentsIfTwoBookiesFailedInSameEnsemble(org.apache.b ookkeeper.client.TestLedgerChecker): test timed out after 3000 milliseconds On 19 Mar 2015, at 22:54, Sijie Guo guosi...@gmail.com wrote: Yup. But it seems that your vm returns IP address as hostname. I guess that might be related your vm's DNS entry in cloud environment. On Thu, Mar 19, 2015 at 3:47 PM, Flavio Junqueira fpjunque...@yahoo.com mailto:fpjunque...@yahoo.com wrote: I'm not sure this is right. When I run locally, I get this in the logs of CookieTest Host address: 127.0.0.1 Host name: localhost while in the vm I get this: Host address: 10.0.0.4 Host name: 10.0.0.4 Host name is what I get here in Bookie.java: if (conf.getUseHostNameAsBookieID()) { hostAddress = inetAddr.getAddress().getCanonicalHostName(); LOG.info(Host name: + hostAddress); } It shouldn't be returning the IP address, no? -Flavio On 19 Mar 2015, at 17:08, Sijie Guo guosi...@gmail.com mailto: guosi...@gmail.com wrote: The hostname in that host will be resolved to be IP, which the IP and hostname would be same. But the tests expect that the IP and hostname are different. We should change the tool to allow passing in any bookie id, which would make the tests more deterministic. - Sijie On Thu, Mar 19, 2015 at 3:26 AM, Flavio Junqueira fpjunque...@yahoo.com.invalid mailto:fpjunque...@yahoo.com.invalid wrote: Sijie, The problem seems to be that the public address (the one the hostname maps to) and the virtual network are different. The tests that are failing seem to expect that they are the same. Does it make sense? -Flavio On Wednesday, March 18, 2015 5:12 AM, Rakesh R rake...@huawei.com mailto:rake...@huawei.com wrote: Can we include BOOKKEEPER-834 fix also in 4.3.1, this is addressing one test case failure. -Rakesh -Original Message- From: Sijie Guo [mailto:guosi...@gmail.com mailto:guosi...@gmail.com] Sent: 18 March 2015 10:23 To: dev@bookkeeper.apache.org mailto:dev@bookkeeper.apache.org Subject: Re: RC for 4.3.1? I think RC0 is failed because of the failed tests. We need to address those tests for producing the new RC. - Sijie On Tue, Mar 17, 2015 at 3:04 PM, Flavio Junqueira fpjunque...@yahoo.com.invalid mailto:fpjunque...@yahoo.com.invalid wrote: Do we have a code freeze on branch 4.3 right now because of release 4.3.1? I'm actually not sure what's going on with the RC0 of 4.3.1. -Flavio
Re: RC for 4.3.1?
I am cutting 4.3.1 from branch 4.3. as we already cut 4.3.0, I'd think we should release 4.3.1 from branch 4.3 rather than from master. otherwise, it is a bit confused for maintaining. we could focus on 4.4 and move on from 4.3 after this. btw, the last jira for 4.3.1: https://issues.apache.org/jira/browse/BOOKKEEPER-854 could anyone review it? - Sijie On Tue, May 19, 2015 at 1:44 AM, Ivan Kelly iv...@apache.org wrote: Are you going to cut from 4.3.1 from master or branch-4.3? IMO, since there's been no big feature nor breaking change, it should come from master (and branch-4.3 should be deleted until it is actually needed). -Ivan On Wed, May 13, 2015 at 9:06 PM, Sijie Guo guosi...@gmail.com wrote: yes. I will cut the new RC candidate. - Sijie On Wed, May 13, 2015 at 6:13 AM, Flavio Junqueira fpjunque...@yahoo.com wrote: +1 On Wednesday, May 13, 2015 9:57 AM, Rakesh R rake...@huawei.com wrote: Hi Sijie, BOOKKEEPER-836 has been resolved. Now, we have only BOOKKEEPER-835 is marked for this release and we have got two +1s and no -1 Shall we go ahead with the release ? Best Regards, Rakesh -Original Message- From: Flavio Junqueira [mailto:fpjunque...@yahoo.com] Sent: 21 March 2015 03:34 To: Rakesh R Cc: dev@bookkeeper.apache.org; Sijie Guo Subject: Re: RC for 4.3.1? I have uploaded the logs to BK-846. Having the timeout set to 3s might not be enough, but I'd like to understand if it is really necessary to increase rather than increasing arbitrarily. I have checked that it doesn't fail consistently. -Flavio On 20 Mar 2015, at 12:10, Rakesh R rake...@huawei.com wrote: Hi Flavio, I have just noticed one thing, it is configured 3000 milliseconds timeout. That is too small value. Can you please increase to @Test(timeout = 6) and verify the test case again. Regards, Rakesh -Original Message- From: Rakesh R Sent: 20 March 2015 15:50 To: dev@bookkeeper.apache.org Cc: Sijie Guo Subject: RE: RC for 4.3.1? Hi Flavio, testShouldGetTwoFrgamentsIfTwoBookiesFailedInSameEnsemble(org.a p ache.bookkeeper.client.TestLedgerChecker): test timed out after 3000 milliseconds I could see the following call can take some amount of time SetLedgerFragment result = getUnderReplicatedFragments(lh); I think, will get some hint if you can get the logs and do the analysis. Do you have the logs available with you. Regards, Rakesh -Original Message- From: Flavio Junqueira [mailto:fpjunque...@yahoo.com.INVALID] Sent: 20 March 2015 13:51 To: Sijie Guo Cc: dev@bookkeeper.apache.org Subject: Re: RC for 4.3.1? I've actually been able to get most of the tests to pass by adding an entry to /etc/hosts. I got only different test failure this time around: testShouldGetTwoFrgamentsIfTwoBookiesFailedInSameEnsemble(org.apache.b ookkeeper.client.TestLedgerChecker): test timed out after 3000 milliseconds On 19 Mar 2015, at 22:54, Sijie Guo guosi...@gmail.com wrote: Yup. But it seems that your vm returns IP address as hostname. I guess that might be related your vm's DNS entry in cloud environment. On Thu, Mar 19, 2015 at 3:47 PM, Flavio Junqueira fpjunque...@yahoo.com mailto:fpjunque...@yahoo.com wrote: I'm not sure this is right. When I run locally, I get this in the logs of CookieTest Host address: 127.0.0.1 Host name: localhost while in the vm I get this: Host address: 10.0.0.4 Host name: 10.0.0.4 Host name is what I get here in Bookie.java: if (conf.getUseHostNameAsBookieID()) { hostAddress = inetAddr.getAddress().getCanonicalHostName(); LOG.info(Host name: + hostAddress); } It shouldn't be returning the IP address, no? -Flavio On 19 Mar 2015, at 17:08, Sijie Guo guosi...@gmail.com mailto: guosi...@gmail.com wrote: The hostname in that host will be resolved to be IP, which the IP and hostname would be same. But the tests expect that the IP and hostname are different. We should change the tool to allow passing in any bookie id, which would make the tests more deterministic. - Sijie On Thu, Mar 19, 2015 at 3:26 AM, Flavio Junqueira fpjunque...@yahoo.com.invalid mailto:fpjunque...@yahoo.com.invalid wrote: Sijie, The problem seems to be that the public address (the one the hostname maps to) and the virtual network are different. The tests that are failing seem to expect that they are the same. Does it make sense? -Flavio On Wednesday, March 18, 2015 5:12 AM, Rakesh R rake...@huawei.com mailto:rake...@huawei.com wrote: Can we include BOOKKEEPER-834 fix also in 4.3.1, this is addressing one test case failure. -Rakesh -Original Message- From: Sijie Guo [mailto:guosi
Re: RC for 4.3.1?
yes. I will cut the new RC candidate. - Sijie On Wed, May 13, 2015 at 6:13 AM, Flavio Junqueira fpjunque...@yahoo.com wrote: +1 On Wednesday, May 13, 2015 9:57 AM, Rakesh R rake...@huawei.com wrote: Hi Sijie, BOOKKEEPER-836 has been resolved. Now, we have only BOOKKEEPER-835 is marked for this release and we have got two +1s and no -1 Shall we go ahead with the release ? Best Regards, Rakesh -Original Message- From: Flavio Junqueira [mailto:fpjunque...@yahoo.com] Sent: 21 March 2015 03:34 To: Rakesh R Cc: dev@bookkeeper.apache.org; Sijie Guo Subject: Re: RC for 4.3.1? I have uploaded the logs to BK-846. Having the timeout set to 3s might not be enough, but I'd like to understand if it is really necessary to increase rather than increasing arbitrarily. I have checked that it doesn't fail consistently. -Flavio On 20 Mar 2015, at 12:10, Rakesh R rake...@huawei.com wrote: Hi Flavio, I have just noticed one thing, it is configured 3000 milliseconds timeout. That is too small value. Can you please increase to @Test(timeout = 6) and verify the test case again. Regards, Rakesh -Original Message- From: Rakesh R Sent: 20 March 2015 15:50 To: dev@bookkeeper.apache.org Cc: Sijie Guo Subject: RE: RC for 4.3.1? Hi Flavio, testShouldGetTwoFrgamentsIfTwoBookiesFailedInSameEnsemble(org.a p ache.bookkeeper.client.TestLedgerChecker): test timed out after 3000 milliseconds I could see the following call can take some amount of time SetLedgerFragment result = getUnderReplicatedFragments(lh); I think, will get some hint if you can get the logs and do the analysis. Do you have the logs available with you. Regards, Rakesh -Original Message- From: Flavio Junqueira [mailto:fpjunque...@yahoo.com.INVALID] Sent: 20 March 2015 13:51 To: Sijie Guo Cc: dev@bookkeeper.apache.org Subject: Re: RC for 4.3.1? I've actually been able to get most of the tests to pass by adding an entry to /etc/hosts. I got only different test failure this time around: testShouldGetTwoFrgamentsIfTwoBookiesFailedInSameEnsemble(org.apache.b ookkeeper.client.TestLedgerChecker): test timed out after 3000 milliseconds On 19 Mar 2015, at 22:54, Sijie Guo guosi...@gmail.com wrote: Yup. But it seems that your vm returns IP address as hostname. I guess that might be related your vm's DNS entry in cloud environment. On Thu, Mar 19, 2015 at 3:47 PM, Flavio Junqueira fpjunque...@yahoo.com mailto:fpjunque...@yahoo.com wrote: I'm not sure this is right. When I run locally, I get this in the logs of CookieTest Host address: 127.0.0.1 Host name: localhost while in the vm I get this: Host address: 10.0.0.4 Host name: 10.0.0.4 Host name is what I get here in Bookie.java: if (conf.getUseHostNameAsBookieID()) { hostAddress = inetAddr.getAddress().getCanonicalHostName(); LOG.info(Host name: + hostAddress); } It shouldn't be returning the IP address, no? -Flavio On 19 Mar 2015, at 17:08, Sijie Guo guosi...@gmail.com mailto: guosi...@gmail.com wrote: The hostname in that host will be resolved to be IP, which the IP and hostname would be same. But the tests expect that the IP and hostname are different. We should change the tool to allow passing in any bookie id, which would make the tests more deterministic. - Sijie On Thu, Mar 19, 2015 at 3:26 AM, Flavio Junqueira fpjunque...@yahoo.com.invalid mailto:fpjunque...@yahoo.com.invalid wrote: Sijie, The problem seems to be that the public address (the one the hostname maps to) and the virtual network are different. The tests that are failing seem to expect that they are the same. Does it make sense? -Flavio On Wednesday, March 18, 2015 5:12 AM, Rakesh R rake...@huawei.com mailto:rake...@huawei.com wrote: Can we include BOOKKEEPER-834 fix also in 4.3.1, this is addressing one test case failure. -Rakesh -Original Message- From: Sijie Guo [mailto:guosi...@gmail.com mailto:guosi...@gmail.com] Sent: 18 March 2015 10:23 To: dev@bookkeeper.apache.org mailto:dev@bookkeeper.apache.org Subject: Re: RC for 4.3.1? I think RC0 is failed because of the failed tests. We need to address those tests for producing the new RC. - Sijie On Tue, Mar 17, 2015 at 3:04 PM, Flavio Junqueira fpjunque...@yahoo.com.invalid mailto:fpjunque...@yahoo.com.invalid wrote: Do we have a code freeze on branch 4.3 right now because of release 4.3.1? I'm actually not sure what's going on with the RC0 of 4.3.1. -Flavio
Re: RC for 4.3.1?
I have uploaded the logs to BK-846. Having the timeout set to 3s might not be enough, but I'd like to understand if it is really necessary to increase rather than increasing arbitrarily. I have checked that it doesn't fail consistently. -Flavio On 20 Mar 2015, at 12:10, Rakesh R rake...@huawei.com wrote: Hi Flavio, I have just noticed one thing, it is configured 3000 milliseconds timeout. That is too small value. Can you please increase to @Test(timeout = 6) and verify the test case again. Regards, Rakesh -Original Message- From: Rakesh R Sent: 20 March 2015 15:50 To: dev@bookkeeper.apache.org Cc: Sijie Guo Subject: RE: RC for 4.3.1? Hi Flavio, testShouldGetTwoFrgamentsIfTwoBookiesFailedInSameEnsemble(org.ap ache.bookkeeper.client.TestLedgerChecker): test timed out after 3000 milliseconds I could see the following call can take some amount of time SetLedgerFragment result = getUnderReplicatedFragments(lh); I think, will get some hint if you can get the logs and do the analysis. Do you have the logs available with you. Regards, Rakesh -Original Message- From: Flavio Junqueira [mailto:fpjunque...@yahoo.com.INVALID] Sent: 20 March 2015 13:51 To: Sijie Guo Cc: dev@bookkeeper.apache.org Subject: Re: RC for 4.3.1? I've actually been able to get most of the tests to pass by adding an entry to /etc/hosts. I got only different test failure this time around: testShouldGetTwoFrgamentsIfTwoBookiesFailedInSameEnsemble(org.apache.bookkeeper.client.TestLedgerChecker): test timed out after 3000 milliseconds On 19 Mar 2015, at 22:54, Sijie Guo guosi...@gmail.com wrote: Yup. But it seems that your vm returns IP address as hostname. I guess that might be related your vm's DNS entry in cloud environment. On Thu, Mar 19, 2015 at 3:47 PM, Flavio Junqueira fpjunque...@yahoo.com mailto:fpjunque...@yahoo.com wrote: I'm not sure this is right. When I run locally, I get this in the logs of CookieTest Host address: 127.0.0.1 Host name: localhost while in the vm I get this: Host address: 10.0.0.4 Host name: 10.0.0.4 Host name is what I get here in Bookie.java: if (conf.getUseHostNameAsBookieID()) { hostAddress = inetAddr.getAddress().getCanonicalHostName(); LOG.info(Host name: + hostAddress); } It shouldn't be returning the IP address, no? -Flavio On 19 Mar 2015, at 17:08, Sijie Guo guosi...@gmail.com mailto:guosi...@gmail.com wrote: The hostname in that host will be resolved to be IP, which the IP and hostname would be same. But the tests expect that the IP and hostname are different. We should change the tool to allow passing in any bookie id, which would make the tests more deterministic. - Sijie On Thu, Mar 19, 2015 at 3:26 AM, Flavio Junqueira fpjunque...@yahoo.com.invalid mailto:fpjunque...@yahoo.com.invalid wrote: Sijie, The problem seems to be that the public address (the one the hostname maps to) and the virtual network are different. The tests that are failing seem to expect that they are the same. Does it make sense? -Flavio On Wednesday, March 18, 2015 5:12 AM, Rakesh R rake...@huawei.com mailto:rake...@huawei.com wrote: Can we include BOOKKEEPER-834 fix also in 4.3.1, this is addressing one test case failure. -Rakesh -Original Message- From: Sijie Guo [mailto:guosi...@gmail.com mailto:guosi...@gmail.com] Sent: 18 March 2015 10:23 To: dev@bookkeeper.apache.org mailto:dev@bookkeeper.apache.org Subject: Re: RC for 4.3.1? I think RC0 is failed because of the failed tests. We need to address those tests for producing the new RC. - Sijie On Tue, Mar 17, 2015 at 3:04 PM, Flavio Junqueira fpjunque...@yahoo.com.invalid mailto:fpjunque...@yahoo.com.invalid wrote: Do we have a code freeze on branch 4.3 right now because of release 4.3.1? I'm actually not sure what's going on with the RC0 of 4.3.1. -Flavio
Re: RC for 4.3.1?
I'm not sure this is right. When I run locally, I get this in the logs of CookieTest Host address: 127.0.0.1 Host name: localhost while in the vm I get this: Host address: 10.0.0.4 Host name: 10.0.0.4 Host name is what I get here in Bookie.java: if (conf.getUseHostNameAsBookieID()) { hostAddress = inetAddr.getAddress().getCanonicalHostName(); LOG.info(Host name: + hostAddress); } It shouldn't be returning the IP address, no? -Flavio On 19 Mar 2015, at 17:08, Sijie Guo guosi...@gmail.com wrote: The hostname in that host will be resolved to be IP, which the IP and hostname would be same. But the tests expect that the IP and hostname are different. We should change the tool to allow passing in any bookie id, which would make the tests more deterministic. - Sijie On Thu, Mar 19, 2015 at 3:26 AM, Flavio Junqueira fpjunque...@yahoo.com.invalid mailto:fpjunque...@yahoo.com.invalid wrote: Sijie, The problem seems to be that the public address (the one the hostname maps to) and the virtual network are different. The tests that are failing seem to expect that they are the same. Does it make sense? -Flavio On Wednesday, March 18, 2015 5:12 AM, Rakesh R rake...@huawei.com mailto:rake...@huawei.com wrote: Can we include BOOKKEEPER-834 fix also in 4.3.1, this is addressing one test case failure. -Rakesh -Original Message- From: Sijie Guo [mailto:guosi...@gmail.com mailto:guosi...@gmail.com] Sent: 18 March 2015 10:23 To: dev@bookkeeper.apache.org mailto:dev@bookkeeper.apache.org Subject: Re: RC for 4.3.1? I think RC0 is failed because of the failed tests. We need to address those tests for producing the new RC. - Sijie On Tue, Mar 17, 2015 at 3:04 PM, Flavio Junqueira fpjunque...@yahoo.com.invalid wrote: Do we have a code freeze on branch 4.3 right now because of release 4.3.1? I'm actually not sure what's going on with the RC0 of 4.3.1. -Flavio
Re: RC for 4.3.1?
Sijie, The problem seems to be that the public address (the one the hostname maps to) and the virtual network are different. The tests that are failing seem to expect that they are the same. Does it make sense? -Flavio On Wednesday, March 18, 2015 5:12 AM, Rakesh R rake...@huawei.com wrote: Can we include BOOKKEEPER-834 fix also in 4.3.1, this is addressing one test case failure. -Rakesh -Original Message- From: Sijie Guo [mailto:guosi...@gmail.com] Sent: 18 March 2015 10:23 To: dev@bookkeeper.apache.org Subject: Re: RC for 4.3.1? I think RC0 is failed because of the failed tests. We need to address those tests for producing the new RC. - Sijie On Tue, Mar 17, 2015 at 3:04 PM, Flavio Junqueira fpjunque...@yahoo.com.invalid wrote: Do we have a code freeze on branch 4.3 right now because of release 4.3.1? I'm actually not sure what's going on with the RC0 of 4.3.1. -Flavio
Re: RC for 4.3.1?
The hostname in that host will be resolved to be IP, which the IP and hostname would be same. But the tests expect that the IP and hostname are different. We should change the tool to allow passing in any bookie id, which would make the tests more deterministic. - Sijie On Thu, Mar 19, 2015 at 3:26 AM, Flavio Junqueira fpjunque...@yahoo.com.invalid wrote: Sijie, The problem seems to be that the public address (the one the hostname maps to) and the virtual network are different. The tests that are failing seem to expect that they are the same. Does it make sense? -Flavio On Wednesday, March 18, 2015 5:12 AM, Rakesh R rake...@huawei.com wrote: Can we include BOOKKEEPER-834 fix also in 4.3.1, this is addressing one test case failure. -Rakesh -Original Message- From: Sijie Guo [mailto:guosi...@gmail.com] Sent: 18 March 2015 10:23 To: dev@bookkeeper.apache.org Subject: Re: RC for 4.3.1? I think RC0 is failed because of the failed tests. We need to address those tests for producing the new RC. - Sijie On Tue, Mar 17, 2015 at 3:04 PM, Flavio Junqueira fpjunque...@yahoo.com.invalid wrote: Do we have a code freeze on branch 4.3 right now because of release 4.3.1? I'm actually not sure what's going on with the RC0 of 4.3.1. -Flavio