Re: Issues running Minicluster in o.a.a.start.Main
I am working on a fix for this. On Tue, Oct 8, 2013 at 10:18 PM, Corey Nolet cjno...@gmail.com wrote: I had started my bisect at the first commit at which it was introduced, though it looks like it still took me on a similar path as where it took you- needless to say, after the 9 or so steps that it allowed me to take, the minicuster command was still broken. On Tue, Oct 8, 2013 at 10:12 PM, Keith Turner ke...@deenlo.com wrote: On Tue, Oct 8, 2013 at 10:06 PM, Corey Nolet cjno...@gmail.com wrote: With risk of making this more complicated- I just noticed that the first commit posted was still broken- though it didn't lock up like the version currently in master, it appeared to run but threw the ClassNotFound exception in the Main.err log. I'm still poking at this as well. I am going to take a look at the first place it was introduced in master and see if it works there. On Tue, Oct 8, 2013 at 10:03 PM, Keith Turner ke...@deenlo.com wrote: I tried using git bisect (for the second time) to tackle this issue and ran into an interesting issue. git bisect took me from master to 1.4. This was unexpected at first, but I understand why git is doing this. I would like to avoid this and only consider commits in master. $ grep -A 5 org.apache.accumulo pom.xml | head -5 groupIdorg.apache.accumulo/groupId artifactIdaccumulo-project/artifactId version1.6.0-SNAPSHOT/version packagingpom/packaging nameApache Accumulo Project/name $ git bisect start $ git bisect bad $ git bisect good 6965a8aaa2f53ec796a3487c1639affe0dfc6bfa Bisecting: 540 revisions left to test after this (roughly 9 steps) [c9469405c3d1aab3784ed5f290df4acaf8568489] ACCUMULO-1168 $ grep -A 5 org.apache.accumulo pom.xml | head -5 groupIdorg.apache.accumulo/groupId artifactIdaccumulo/artifactId packagingpom/packaging version1.4.3/version nameaccumulo/name I found the a post[1] on stack overflow that mentioned using git rev-list --bisect --first-parent so I tried the following command and saw git segfault. I blame this on our switch from svn. $ git rev-list --bisect --first-parent 6965a8aaa2f53ec796a3487c1639affe0dfc6bfa..HEAD Segmentation fault (core dumped) If I drop the --first-parent option then it shows me the same commit that git bisect does. $git rev-list --bisect 6965a8aaa2f53ec796a3487c1639affe0dfc6bfa..HEAD c9469405c3d1aab3784ed5f290df4acaf8568489 I am still poking at this. If anyone has advice I would like to hear it. [1]: http://stackoverflow.com/questions/5638211/how-do-you-get-git-bisect-to-ignore-merged-branches On Tue, Oct 8, 2013 at 12:02 AM, Corey Nolet cjno...@gmail.com wrote: Keith, You are right- I mistyped. I meant Main.err not Master.err. I just verified this feature worked during the time of this commit: 6965a8aaa2f53ec796a3487c1639affe0dfc6bfa. On Mon, Oct 7, 2013 at 11:39 PM, Keith Turner ke...@deenlo.com wrote: I just tried running accumulo miniscluster and saw the same thing. But in Main.err, not Master.err are you sure you saw this in Master.err? Has this ever worked? By default the accumulo scripts construct a very minimal classpath w/ accumulo-start.jar, log4j-1.2.15.jar, and the conf dir. If you modify the MAC exec method to print the classpath it uses to start a java process, then you can see this. MAC makes the assumption that everything it needs is on the Java classpath, which is true when its run from Maven or Eclipse. However when its run from the accumulo scripts, this is not true. Also, for some reason MAC starts zookeeper using Accumulo start main. I have no idea why its doing this. Even if it was not doing this, I think would fail in different way (i.e. instead of not finding VFS it would not find zookeeper class). Are you familiar with accumulo start module? On Mon, Oct 7, 2013 at 7:36 AM, Corey Nolet cjno...@gmail.com wrote: The MiniAccumuloRunner class that's wired up to o.o.a.start.Main. I was specifically wondering if anyone else is experiencing issues running 'accumulo minicluster' as both the proxy with useMini=true and the minicluster command seem broken for me. I'm building from remote HEAD in master. On Oct 6, 2013 11:32 AM, John Vines jvi...@gmail.com wrote: How are you running minicluster? On Sun, Oct 6, 2013 at 8:36 AM, Corey Nolet cjno...@gmail.com wrote: I'm having issues running the minicluster both in the
Re: Issues running Minicluster in o.a.a.start.Main
On Tue, Oct 8, 2013 at 9:45 PM, Corey Nolet cjno...@gmail.com wrote: Keith, It appears that the proxy's useMini and the minicluster command broke somewhere between 1edccf6b30541841bb08329317c6289aca8c8d73 ACCUMULO-1707 applying Steve's patch and 98d7a9efc6d07e71d3803b3830bf9dc9ce8dec9d ACCUMULO-1558 made import table fail when files do not exist There was some weird merging happening during that sequence of commits. At some point it looks like the master pom was set to version 1.5.1. If a bug was fixed in this sequence of commits and the fix inherently broke the commands because they didn't belong in start.Main to begin with, I'm thinking it would probably be best to take the useMini property out of the proxy and the minicluster command out of start.Main. If a bug was introduced, I'd like to get a ticket up for it. Yeah, go ahead and open a ticket. On Tue, Oct 8, 2013 at 12:02 AM, Corey Nolet cjno...@gmail.com wrote: Keith, You are right- I mistyped. I meant Main.err not Master.err. I just verified this feature worked during the time of this commit: 6965a8aaa2f53ec796a3487c1639affe0dfc6bfa. On Mon, Oct 7, 2013 at 11:39 PM, Keith Turner ke...@deenlo.com wrote: I just tried running accumulo miniscluster and saw the same thing. But in Main.err, not Master.err are you sure you saw this in Master.err? Has this ever worked? By default the accumulo scripts construct a very minimal classpath w/ accumulo-start.jar, log4j-1.2.15.jar, and the conf dir. If you modify the MAC exec method to print the classpath it uses to start a java process, then you can see this. MAC makes the assumption that everything it needs is on the Java classpath, which is true when its run from Maven or Eclipse. However when its run from the accumulo scripts, this is not true. Also, for some reason MAC starts zookeeper using Accumulo start main. I have no idea why its doing this. Even if it was not doing this, I think would fail in different way (i.e. instead of not finding VFS it would not find zookeeper class). Are you familiar with accumulo start module? On Mon, Oct 7, 2013 at 7:36 AM, Corey Nolet cjno...@gmail.com wrote: The MiniAccumuloRunner class that's wired up to o.o.a.start.Main. I was specifically wondering if anyone else is experiencing issues running 'accumulo minicluster' as both the proxy with useMini=true and the minicluster command seem broken for me. I'm building from remote HEAD in master. On Oct 6, 2013 11:32 AM, John Vines jvi...@gmail.com wrote: How are you running minicluster? On Sun, Oct 6, 2013 at 8:36 AM, Corey Nolet cjno...@gmail.com wrote: I'm having issues running the minicluster both in the 'accumulo proxy -p proxy.properties' and via 'accumulo minicluster'. It looks like the Zookeeper process is not starting and the MAC is going into an infinite loop waiting for it to start. I checked the Master.err logs for the minicluster command and I see the following: Uncaught exception: java.lang.NoClassDefFoundError: org/apache/commons/vfs2/impl/VFSClassLoader java.lang.NoClassDefFoundError: org/apache/commons/vfs2/impl/VFSClassLoader at java.lang.Class.getDeclaredMethods0(Native Method) at java.lang.Class.privateGetDeclaredMethods(Class.java:2521) at java.lang.Class.getMethod0(Class.java:2764) at java.lang.Class.getMethod(Class.java:1653) at org.apache.accumulo.start.Main.main(Main.java:42) Caused by: java.lang.ClassNotFoundException: org.apache.commons.vfs2.impl.VFSClassLoader at java.net.URLClassLoader$1.run(URLClassLoader.java:366) at java.net.URLClassLoader$1.run(URLClassLoader.java:355) at java.security.AccessController.doPrivileged(Native Method) at java.net.URLClassLoader.findClass(URLClassLoader.java:354) at java.lang.ClassLoader.loadClass(ClassLoader.java:424) at sun.misc.Launcher$AppClassLoader.loadClass(Launcher.java:308) at java.lang.ClassLoader.loadClass(ClassLoader.java:357) ... 5 more the commons-vfs2.jar is in Accumulo's lib directory. I'm using Hadoop 1.2.1. -- Cheers ~John
Re: Issues running Minicluster in o.a.a.start.Main
Looking at the commit history in master using gitk/gitx shows duplicated history. It looks like January through July 2013 occur at least twice in the commit history (with 2012 in between them). I'm suspecting this may have been why bisect wasn't working for me On Wed, Oct 9, 2013 at 1:25 PM, Keith Turner ke...@deenlo.com wrote: On Tue, Oct 8, 2013 at 9:45 PM, Corey Nolet cjno...@gmail.com wrote: Keith, It appears that the proxy's useMini and the minicluster command broke somewhere between 1edccf6b30541841bb08329317c6289aca8c8d73 ACCUMULO-1707 applying Steve's patch and 98d7a9efc6d07e71d3803b3830bf9dc9ce8dec9d ACCUMULO-1558 made import table fail when files do not exist There was some weird merging happening during that sequence of commits. At some point it looks like the master pom was set to version 1.5.1. If a bug was fixed in this sequence of commits and the fix inherently broke the commands because they didn't belong in start.Main to begin with, I'm thinking it would probably be best to take the useMini property out of the proxy and the minicluster command out of start.Main. If a bug was introduced, I'd like to get a ticket up for it. Yeah, go ahead and open a ticket. On Tue, Oct 8, 2013 at 12:02 AM, Corey Nolet cjno...@gmail.com wrote: Keith, You are right- I mistyped. I meant Main.err not Master.err. I just verified this feature worked during the time of this commit: 6965a8aaa2f53ec796a3487c1639affe0dfc6bfa. On Mon, Oct 7, 2013 at 11:39 PM, Keith Turner ke...@deenlo.com wrote: I just tried running accumulo miniscluster and saw the same thing. But in Main.err, not Master.err are you sure you saw this in Master.err? Has this ever worked? By default the accumulo scripts construct a very minimal classpath w/ accumulo-start.jar, log4j-1.2.15.jar, and the conf dir. If you modify the MAC exec method to print the classpath it uses to start a java process, then you can see this. MAC makes the assumption that everything it needs is on the Java classpath, which is true when its run from Maven or Eclipse. However when its run from the accumulo scripts, this is not true. Also, for some reason MAC starts zookeeper using Accumulo start main. I have no idea why its doing this. Even if it was not doing this, I think would fail in different way (i.e. instead of not finding VFS it would not find zookeeper class). Are you familiar with accumulo start module? On Mon, Oct 7, 2013 at 7:36 AM, Corey Nolet cjno...@gmail.com wrote: The MiniAccumuloRunner class that's wired up to o.o.a.start.Main. I was specifically wondering if anyone else is experiencing issues running 'accumulo minicluster' as both the proxy with useMini=true and the minicluster command seem broken for me. I'm building from remote HEAD in master. On Oct 6, 2013 11:32 AM, John Vines jvi...@gmail.com wrote: How are you running minicluster? On Sun, Oct 6, 2013 at 8:36 AM, Corey Nolet cjno...@gmail.com wrote: I'm having issues running the minicluster both in the 'accumulo proxy -p proxy.properties' and via 'accumulo minicluster'. It looks like the Zookeeper process is not starting and the MAC is going into an infinite loop waiting for it to start. I checked the Master.err logs for the minicluster command and I see the following: Uncaught exception: java.lang.NoClassDefFoundError: org/apache/commons/vfs2/impl/VFSClassLoader java.lang.NoClassDefFoundError: org/apache/commons/vfs2/impl/VFSClassLoader at java.lang.Class.getDeclaredMethods0(Native Method) at java.lang.Class.privateGetDeclaredMethods(Class.java:2521) at java.lang.Class.getMethod0(Class.java:2764) at java.lang.Class.getMethod(Class.java:1653) at org.apache.accumulo.start.Main.main(Main.java:42) Caused by: java.lang.ClassNotFoundException: org.apache.commons.vfs2.impl.VFSClassLoader at java.net.URLClassLoader$1.run(URLClassLoader.java:366) at java.net.URLClassLoader$1.run(URLClassLoader.java:355) at java.security.AccessController.doPrivileged(Native Method) at java.net.URLClassLoader.findClass(URLClassLoader.java:354) at java.lang.ClassLoader.loadClass(ClassLoader.java:424) at sun.misc.Launcher$AppClassLoader.loadClass(Launcher.java:308) at java.lang.ClassLoader.loadClass(ClassLoader.java:357) ... 5 more the commons-vfs2.jar is in Accumulo's lib directory. I'm using Hadoop 1.2.1. -- Cheers ~John
Re: Issues running Minicluster in o.a.a.start.Main
I have accumulo minicluster working now. I changed the way MAC determines what classpath to use. I pushed a commit [1] to a branch on github[2] if you want to take a look. I am still experimenting with the change. I ran into some other bugs in master while working on this and fixed those. I also reopened ACCUMULO-1368 rather than opening a new ticket. [1] https://github.com/keith-turner/accumulo/commit/1b9fbbc82a49e4939179f2f7b0b013517836b7f0 [2] https://github.com/keith-turner/accumulo/tree/ACCUMULO-1368 On Wed, Oct 9, 2013 at 8:52 PM, Corey Nolet cjno...@gmail.com wrote: Looking at the commit history in master using gitk/gitx shows duplicated history. It looks like January through July 2013 occur at least twice in the commit history (with 2012 in between them). I'm suspecting this may have been why bisect wasn't working for me On Wed, Oct 9, 2013 at 1:25 PM, Keith Turner ke...@deenlo.com wrote: On Tue, Oct 8, 2013 at 9:45 PM, Corey Nolet cjno...@gmail.com wrote: Keith, It appears that the proxy's useMini and the minicluster command broke somewhere between 1edccf6b30541841bb08329317c6289aca8c8d73 ACCUMULO-1707 applying Steve's patch and 98d7a9efc6d07e71d3803b3830bf9dc9ce8dec9d ACCUMULO-1558 made import table fail when files do not exist There was some weird merging happening during that sequence of commits. At some point it looks like the master pom was set to version 1.5.1. If a bug was fixed in this sequence of commits and the fix inherently broke the commands because they didn't belong in start.Main to begin with, I'm thinking it would probably be best to take the useMini property out of the proxy and the minicluster command out of start.Main. If a bug was introduced, I'd like to get a ticket up for it. Yeah, go ahead and open a ticket. On Tue, Oct 8, 2013 at 12:02 AM, Corey Nolet cjno...@gmail.com wrote: Keith, You are right- I mistyped. I meant Main.err not Master.err. I just verified this feature worked during the time of this commit: 6965a8aaa2f53ec796a3487c1639affe0dfc6bfa. On Mon, Oct 7, 2013 at 11:39 PM, Keith Turner ke...@deenlo.com wrote: I just tried running accumulo miniscluster and saw the same thing. But in Main.err, not Master.err are you sure you saw this in Master.err? Has this ever worked? By default the accumulo scripts construct a very minimal classpath w/ accumulo-start.jar, log4j-1.2.15.jar, and the conf dir. If you modify the MAC exec method to print the classpath it uses to start a java process, then you can see this. MAC makes the assumption that everything it needs is on the Java classpath, which is true when its run from Maven or Eclipse. However when its run from the accumulo scripts, this is not true. Also, for some reason MAC starts zookeeper using Accumulo start main. I have no idea why its doing this. Even if it was not doing this, I think would fail in different way (i.e. instead of not finding VFS it would not find zookeeper class). Are you familiar with accumulo start module? On Mon, Oct 7, 2013 at 7:36 AM, Corey Nolet cjno...@gmail.com wrote: The MiniAccumuloRunner class that's wired up to o.o.a.start.Main. I was specifically wondering if anyone else is experiencing issues running 'accumulo minicluster' as both the proxy with useMini=true and the minicluster command seem broken for me. I'm building from remote HEAD in master. On Oct 6, 2013 11:32 AM, John Vines jvi...@gmail.com wrote: How are you running minicluster? On Sun, Oct 6, 2013 at 8:36 AM, Corey Nolet cjno...@gmail.com wrote: I'm having issues running the minicluster both in the 'accumulo proxy -p proxy.properties' and via 'accumulo minicluster'. It looks like the Zookeeper process is not starting and the MAC is going into an infinite loop waiting for it to start. I checked the Master.err logs for the minicluster command and I see the following: Uncaught exception: java.lang.NoClassDefFoundError: org/apache/commons/vfs2/impl/VFSClassLoader java.lang.NoClassDefFoundError: org/apache/commons/vfs2/impl/VFSClassLoader at java.lang.Class.getDeclaredMethods0(Native Method) at java.lang.Class.privateGetDeclaredMethods(Class.java:2521) at java.lang.Class.getMethod0(Class.java:2764) at java.lang.Class.getMethod(Class.java:1653) at org.apache.accumulo.start.Main.main(Main.java:42) Caused by: java.lang.ClassNotFoundException: org.apache.commons.vfs2.impl.VFSClassLoader at java.net.URLClassLoader$1.run(URLClassLoader.java:366) at
Re: Issues running Minicluster in o.a.a.start.Main
I tried using git bisect (for the second time) to tackle this issue and ran into an interesting issue. git bisect took me from master to 1.4. This was unexpected at first, but I understand why git is doing this. I would like to avoid this and only consider commits in master. $ grep -A 5 org.apache.accumulo pom.xml | head -5 groupIdorg.apache.accumulo/groupId artifactIdaccumulo-project/artifactId version1.6.0-SNAPSHOT/version packagingpom/packaging nameApache Accumulo Project/name $ git bisect start $ git bisect bad $ git bisect good 6965a8aaa2f53ec796a3487c1639affe0dfc6bfa Bisecting: 540 revisions left to test after this (roughly 9 steps) [c9469405c3d1aab3784ed5f290df4acaf8568489] ACCUMULO-1168 $ grep -A 5 org.apache.accumulo pom.xml | head -5 groupIdorg.apache.accumulo/groupId artifactIdaccumulo/artifactId packagingpom/packaging version1.4.3/version nameaccumulo/name I found the a post[1] on stack overflow that mentioned using git rev-list --bisect --first-parent so I tried the following command and saw git segfault. I blame this on our switch from svn. $ git rev-list --bisect --first-parent 6965a8aaa2f53ec796a3487c1639affe0dfc6bfa..HEAD Segmentation fault (core dumped) If I drop the --first-parent option then it shows me the same commit that git bisect does. $git rev-list --bisect 6965a8aaa2f53ec796a3487c1639affe0dfc6bfa..HEAD c9469405c3d1aab3784ed5f290df4acaf8568489 I am still poking at this. If anyone has advice I would like to hear it. [1]: http://stackoverflow.com/questions/5638211/how-do-you-get-git-bisect-to-ignore-merged-branches On Tue, Oct 8, 2013 at 12:02 AM, Corey Nolet cjno...@gmail.com wrote: Keith, You are right- I mistyped. I meant Main.err not Master.err. I just verified this feature worked during the time of this commit: 6965a8aaa2f53ec796a3487c1639affe0dfc6bfa. On Mon, Oct 7, 2013 at 11:39 PM, Keith Turner ke...@deenlo.com wrote: I just tried running accumulo miniscluster and saw the same thing. But in Main.err, not Master.err are you sure you saw this in Master.err? Has this ever worked? By default the accumulo scripts construct a very minimal classpath w/ accumulo-start.jar, log4j-1.2.15.jar, and the conf dir. If you modify the MAC exec method to print the classpath it uses to start a java process, then you can see this. MAC makes the assumption that everything it needs is on the Java classpath, which is true when its run from Maven or Eclipse. However when its run from the accumulo scripts, this is not true. Also, for some reason MAC starts zookeeper using Accumulo start main. I have no idea why its doing this. Even if it was not doing this, I think would fail in different way (i.e. instead of not finding VFS it would not find zookeeper class). Are you familiar with accumulo start module? On Mon, Oct 7, 2013 at 7:36 AM, Corey Nolet cjno...@gmail.com wrote: The MiniAccumuloRunner class that's wired up to o.o.a.start.Main. I was specifically wondering if anyone else is experiencing issues running 'accumulo minicluster' as both the proxy with useMini=true and the minicluster command seem broken for me. I'm building from remote HEAD in master. On Oct 6, 2013 11:32 AM, John Vines jvi...@gmail.com wrote: How are you running minicluster? On Sun, Oct 6, 2013 at 8:36 AM, Corey Nolet cjno...@gmail.com wrote: I'm having issues running the minicluster both in the 'accumulo proxy -p proxy.properties' and via 'accumulo minicluster'. It looks like the Zookeeper process is not starting and the MAC is going into an infinite loop waiting for it to start. I checked the Master.err logs for the minicluster command and I see the following: Uncaught exception: java.lang.NoClassDefFoundError: org/apache/commons/vfs2/impl/VFSClassLoader java.lang.NoClassDefFoundError: org/apache/commons/vfs2/impl/VFSClassLoader at java.lang.Class.getDeclaredMethods0(Native Method) at java.lang.Class.privateGetDeclaredMethods(Class.java:2521) at java.lang.Class.getMethod0(Class.java:2764) at java.lang.Class.getMethod(Class.java:1653) at org.apache.accumulo.start.Main.main(Main.java:42) Caused by: java.lang.ClassNotFoundException: org.apache.commons.vfs2.impl.VFSClassLoader at java.net.URLClassLoader$1.run(URLClassLoader.java:366) at java.net.URLClassLoader$1.run(URLClassLoader.java:355) at java.security.AccessController.doPrivileged(Native Method) at java.net.URLClassLoader.findClass(URLClassLoader.java:354) at java.lang.ClassLoader.loadClass(ClassLoader.java:424) at sun.misc.Launcher$AppClassLoader.loadClass(Launcher.java:308) at java.lang.ClassLoader.loadClass(ClassLoader.java:357) ... 5 more the commons-vfs2.jar is in Accumulo's lib directory. I'm using Hadoop 1.2.1. -- Cheers
Re: Issues running Minicluster in o.a.a.start.Main
With risk of making this more complicated- I just noticed that the first commit posted was still broken- though it didn't lock up like the version currently in master, it appeared to run but threw the ClassNotFound exception in the Main.err log. I'm still poking at this as well. On Tue, Oct 8, 2013 at 10:03 PM, Keith Turner ke...@deenlo.com wrote: I tried using git bisect (for the second time) to tackle this issue and ran into an interesting issue. git bisect took me from master to 1.4. This was unexpected at first, but I understand why git is doing this. I would like to avoid this and only consider commits in master. $ grep -A 5 org.apache.accumulo pom.xml | head -5 groupIdorg.apache.accumulo/groupId artifactIdaccumulo-project/artifactId version1.6.0-SNAPSHOT/version packagingpom/packaging nameApache Accumulo Project/name $ git bisect start $ git bisect bad $ git bisect good 6965a8aaa2f53ec796a3487c1639affe0dfc6bfa Bisecting: 540 revisions left to test after this (roughly 9 steps) [c9469405c3d1aab3784ed5f290df4acaf8568489] ACCUMULO-1168 $ grep -A 5 org.apache.accumulo pom.xml | head -5 groupIdorg.apache.accumulo/groupId artifactIdaccumulo/artifactId packagingpom/packaging version1.4.3/version nameaccumulo/name I found the a post[1] on stack overflow that mentioned using git rev-list --bisect --first-parent so I tried the following command and saw git segfault. I blame this on our switch from svn. $ git rev-list --bisect --first-parent 6965a8aaa2f53ec796a3487c1639affe0dfc6bfa..HEAD Segmentation fault (core dumped) If I drop the --first-parent option then it shows me the same commit that git bisect does. $git rev-list --bisect 6965a8aaa2f53ec796a3487c1639affe0dfc6bfa..HEAD c9469405c3d1aab3784ed5f290df4acaf8568489 I am still poking at this. If anyone has advice I would like to hear it. [1]: http://stackoverflow.com/questions/5638211/how-do-you-get-git-bisect-to-ignore-merged-branches On Tue, Oct 8, 2013 at 12:02 AM, Corey Nolet cjno...@gmail.com wrote: Keith, You are right- I mistyped. I meant Main.err not Master.err. I just verified this feature worked during the time of this commit: 6965a8aaa2f53ec796a3487c1639affe0dfc6bfa. On Mon, Oct 7, 2013 at 11:39 PM, Keith Turner ke...@deenlo.com wrote: I just tried running accumulo miniscluster and saw the same thing. But in Main.err, not Master.err are you sure you saw this in Master.err? Has this ever worked? By default the accumulo scripts construct a very minimal classpath w/ accumulo-start.jar, log4j-1.2.15.jar, and the conf dir. If you modify the MAC exec method to print the classpath it uses to start a java process, then you can see this. MAC makes the assumption that everything it needs is on the Java classpath, which is true when its run from Maven or Eclipse. However when its run from the accumulo scripts, this is not true. Also, for some reason MAC starts zookeeper using Accumulo start main. I have no idea why its doing this. Even if it was not doing this, I think would fail in different way (i.e. instead of not finding VFS it would not find zookeeper class). Are you familiar with accumulo start module? On Mon, Oct 7, 2013 at 7:36 AM, Corey Nolet cjno...@gmail.com wrote: The MiniAccumuloRunner class that's wired up to o.o.a.start.Main. I was specifically wondering if anyone else is experiencing issues running 'accumulo minicluster' as both the proxy with useMini=true and the minicluster command seem broken for me. I'm building from remote HEAD in master. On Oct 6, 2013 11:32 AM, John Vines jvi...@gmail.com wrote: How are you running minicluster? On Sun, Oct 6, 2013 at 8:36 AM, Corey Nolet cjno...@gmail.com wrote: I'm having issues running the minicluster both in the 'accumulo proxy -p proxy.properties' and via 'accumulo minicluster'. It looks like the Zookeeper process is not starting and the MAC is going into an infinite loop waiting for it to start. I checked the Master.err logs for the minicluster command and I see the following: Uncaught exception: java.lang.NoClassDefFoundError: org/apache/commons/vfs2/impl/VFSClassLoader java.lang.NoClassDefFoundError: org/apache/commons/vfs2/impl/VFSClassLoader at java.lang.Class.getDeclaredMethods0(Native Method) at java.lang.Class.privateGetDeclaredMethods(Class.java:2521) at java.lang.Class.getMethod0(Class.java:2764) at java.lang.Class.getMethod(Class.java:1653) at org.apache.accumulo.start.Main.main(Main.java:42) Caused by: java.lang.ClassNotFoundException: org.apache.commons.vfs2.impl.VFSClassLoader at java.net.URLClassLoader$1.run(URLClassLoader.java:366) at java.net.URLClassLoader$1.run(URLClassLoader.java:355) at
Re: Issues running Minicluster in o.a.a.start.Main
On Tue, Oct 8, 2013 at 10:06 PM, Corey Nolet cjno...@gmail.com wrote: With risk of making this more complicated- I just noticed that the first commit posted was still broken- though it didn't lock up like the version currently in master, it appeared to run but threw the ClassNotFound exception in the Main.err log. I'm still poking at this as well. I am going to take a look at the first place it was introduced in master and see if it works there. On Tue, Oct 8, 2013 at 10:03 PM, Keith Turner ke...@deenlo.com wrote: I tried using git bisect (for the second time) to tackle this issue and ran into an interesting issue. git bisect took me from master to 1.4. This was unexpected at first, but I understand why git is doing this. I would like to avoid this and only consider commits in master. $ grep -A 5 org.apache.accumulo pom.xml | head -5 groupIdorg.apache.accumulo/groupId artifactIdaccumulo-project/artifactId version1.6.0-SNAPSHOT/version packagingpom/packaging nameApache Accumulo Project/name $ git bisect start $ git bisect bad $ git bisect good 6965a8aaa2f53ec796a3487c1639affe0dfc6bfa Bisecting: 540 revisions left to test after this (roughly 9 steps) [c9469405c3d1aab3784ed5f290df4acaf8568489] ACCUMULO-1168 $ grep -A 5 org.apache.accumulo pom.xml | head -5 groupIdorg.apache.accumulo/groupId artifactIdaccumulo/artifactId packagingpom/packaging version1.4.3/version nameaccumulo/name I found the a post[1] on stack overflow that mentioned using git rev-list --bisect --first-parent so I tried the following command and saw git segfault. I blame this on our switch from svn. $ git rev-list --bisect --first-parent 6965a8aaa2f53ec796a3487c1639affe0dfc6bfa..HEAD Segmentation fault (core dumped) If I drop the --first-parent option then it shows me the same commit that git bisect does. $git rev-list --bisect 6965a8aaa2f53ec796a3487c1639affe0dfc6bfa..HEAD c9469405c3d1aab3784ed5f290df4acaf8568489 I am still poking at this. If anyone has advice I would like to hear it. [1]: http://stackoverflow.com/questions/5638211/how-do-you-get-git-bisect-to-ignore-merged-branches On Tue, Oct 8, 2013 at 12:02 AM, Corey Nolet cjno...@gmail.com wrote: Keith, You are right- I mistyped. I meant Main.err not Master.err. I just verified this feature worked during the time of this commit: 6965a8aaa2f53ec796a3487c1639affe0dfc6bfa. On Mon, Oct 7, 2013 at 11:39 PM, Keith Turner ke...@deenlo.com wrote: I just tried running accumulo miniscluster and saw the same thing. But in Main.err, not Master.err are you sure you saw this in Master.err? Has this ever worked? By default the accumulo scripts construct a very minimal classpath w/ accumulo-start.jar, log4j-1.2.15.jar, and the conf dir. If you modify the MAC exec method to print the classpath it uses to start a java process, then you can see this. MAC makes the assumption that everything it needs is on the Java classpath, which is true when its run from Maven or Eclipse. However when its run from the accumulo scripts, this is not true. Also, for some reason MAC starts zookeeper using Accumulo start main. I have no idea why its doing this. Even if it was not doing this, I think would fail in different way (i.e. instead of not finding VFS it would not find zookeeper class). Are you familiar with accumulo start module? On Mon, Oct 7, 2013 at 7:36 AM, Corey Nolet cjno...@gmail.com wrote: The MiniAccumuloRunner class that's wired up to o.o.a.start.Main. I was specifically wondering if anyone else is experiencing issues running 'accumulo minicluster' as both the proxy with useMini=true and the minicluster command seem broken for me. I'm building from remote HEAD in master. On Oct 6, 2013 11:32 AM, John Vines jvi...@gmail.com wrote: How are you running minicluster? On Sun, Oct 6, 2013 at 8:36 AM, Corey Nolet cjno...@gmail.com wrote: I'm having issues running the minicluster both in the 'accumulo proxy -p proxy.properties' and via 'accumulo minicluster'. It looks like the Zookeeper process is not starting and the MAC is going into an infinite loop waiting for it to start. I checked the Master.err logs for the minicluster command and I see the following: Uncaught exception: java.lang.NoClassDefFoundError: org/apache/commons/vfs2/impl/VFSClassLoader java.lang.NoClassDefFoundError: org/apache/commons/vfs2/impl/VFSClassLoader at java.lang.Class.getDeclaredMethods0(Native Method) at java.lang.Class.privateGetDeclaredMethods(Class.java:2521) at java.lang.Class.getMethod0(Class.java:2764) at
Re: Issues running Minicluster in o.a.a.start.Main
I had started my bisect at the first commit at which it was introduced, though it looks like it still took me on a similar path as where it took you- needless to say, after the 9 or so steps that it allowed me to take, the minicuster command was still broken. On Tue, Oct 8, 2013 at 10:12 PM, Keith Turner ke...@deenlo.com wrote: On Tue, Oct 8, 2013 at 10:06 PM, Corey Nolet cjno...@gmail.com wrote: With risk of making this more complicated- I just noticed that the first commit posted was still broken- though it didn't lock up like the version currently in master, it appeared to run but threw the ClassNotFound exception in the Main.err log. I'm still poking at this as well. I am going to take a look at the first place it was introduced in master and see if it works there. On Tue, Oct 8, 2013 at 10:03 PM, Keith Turner ke...@deenlo.com wrote: I tried using git bisect (for the second time) to tackle this issue and ran into an interesting issue. git bisect took me from master to 1.4. This was unexpected at first, but I understand why git is doing this. I would like to avoid this and only consider commits in master. $ grep -A 5 org.apache.accumulo pom.xml | head -5 groupIdorg.apache.accumulo/groupId artifactIdaccumulo-project/artifactId version1.6.0-SNAPSHOT/version packagingpom/packaging nameApache Accumulo Project/name $ git bisect start $ git bisect bad $ git bisect good 6965a8aaa2f53ec796a3487c1639affe0dfc6bfa Bisecting: 540 revisions left to test after this (roughly 9 steps) [c9469405c3d1aab3784ed5f290df4acaf8568489] ACCUMULO-1168 $ grep -A 5 org.apache.accumulo pom.xml | head -5 groupIdorg.apache.accumulo/groupId artifactIdaccumulo/artifactId packagingpom/packaging version1.4.3/version nameaccumulo/name I found the a post[1] on stack overflow that mentioned using git rev-list --bisect --first-parent so I tried the following command and saw git segfault. I blame this on our switch from svn. $ git rev-list --bisect --first-parent 6965a8aaa2f53ec796a3487c1639affe0dfc6bfa..HEAD Segmentation fault (core dumped) If I drop the --first-parent option then it shows me the same commit that git bisect does. $git rev-list --bisect 6965a8aaa2f53ec796a3487c1639affe0dfc6bfa..HEAD c9469405c3d1aab3784ed5f290df4acaf8568489 I am still poking at this. If anyone has advice I would like to hear it. [1]: http://stackoverflow.com/questions/5638211/how-do-you-get-git-bisect-to-ignore-merged-branches On Tue, Oct 8, 2013 at 12:02 AM, Corey Nolet cjno...@gmail.com wrote: Keith, You are right- I mistyped. I meant Main.err not Master.err. I just verified this feature worked during the time of this commit: 6965a8aaa2f53ec796a3487c1639affe0dfc6bfa. On Mon, Oct 7, 2013 at 11:39 PM, Keith Turner ke...@deenlo.com wrote: I just tried running accumulo miniscluster and saw the same thing. But in Main.err, not Master.err are you sure you saw this in Master.err? Has this ever worked? By default the accumulo scripts construct a very minimal classpath w/ accumulo-start.jar, log4j-1.2.15.jar, and the conf dir. If you modify the MAC exec method to print the classpath it uses to start a java process, then you can see this. MAC makes the assumption that everything it needs is on the Java classpath, which is true when its run from Maven or Eclipse. However when its run from the accumulo scripts, this is not true. Also, for some reason MAC starts zookeeper using Accumulo start main. I have no idea why its doing this. Even if it was not doing this, I think would fail in different way (i.e. instead of not finding VFS it would not find zookeeper class). Are you familiar with accumulo start module? On Mon, Oct 7, 2013 at 7:36 AM, Corey Nolet cjno...@gmail.com wrote: The MiniAccumuloRunner class that's wired up to o.o.a.start.Main. I was specifically wondering if anyone else is experiencing issues running 'accumulo minicluster' as both the proxy with useMini=true and the minicluster command seem broken for me. I'm building from remote HEAD in master. On Oct 6, 2013 11:32 AM, John Vines jvi...@gmail.com wrote: How are you running minicluster? On Sun, Oct 6, 2013 at 8:36 AM, Corey Nolet cjno...@gmail.com wrote: I'm having issues running the minicluster both in the 'accumulo proxy -p proxy.properties' and via 'accumulo minicluster'. It looks like the Zookeeper process is not starting and the MAC is going into an infinite loop waiting for it to start. I checked the Master.err
Re: Issues running Minicluster in o.a.a.start.Main
The MiniAccumuloRunner class that's wired up to o.o.a.start.Main. I was specifically wondering if anyone else is experiencing issues running 'accumulo minicluster' as both the proxy with useMini=true and the minicluster command seem broken for me. I'm building from remote HEAD in master. On Oct 6, 2013 11:32 AM, John Vines jvi...@gmail.com wrote: How are you running minicluster? On Sun, Oct 6, 2013 at 8:36 AM, Corey Nolet cjno...@gmail.com wrote: I'm having issues running the minicluster both in the 'accumulo proxy -p proxy.properties' and via 'accumulo minicluster'. It looks like the Zookeeper process is not starting and the MAC is going into an infinite loop waiting for it to start. I checked the Master.err logs for the minicluster command and I see the following: Uncaught exception: java.lang.NoClassDefFoundError: org/apache/commons/vfs2/impl/VFSClassLoader java.lang.NoClassDefFoundError: org/apache/commons/vfs2/impl/VFSClassLoader at java.lang.Class.getDeclaredMethods0(Native Method) at java.lang.Class.privateGetDeclaredMethods(Class.java:2521) at java.lang.Class.getMethod0(Class.java:2764) at java.lang.Class.getMethod(Class.java:1653) at org.apache.accumulo.start.Main.main(Main.java:42) Caused by: java.lang.ClassNotFoundException: org.apache.commons.vfs2.impl.VFSClassLoader at java.net.URLClassLoader$1.run(URLClassLoader.java:366) at java.net.URLClassLoader$1.run(URLClassLoader.java:355) at java.security.AccessController.doPrivileged(Native Method) at java.net.URLClassLoader.findClass(URLClassLoader.java:354) at java.lang.ClassLoader.loadClass(ClassLoader.java:424) at sun.misc.Launcher$AppClassLoader.loadClass(Launcher.java:308) at java.lang.ClassLoader.loadClass(ClassLoader.java:357) ... 5 more the commons-vfs2.jar is in Accumulo's lib directory. I'm using Hadoop 1.2.1. -- Cheers ~John
Re: Issues running Minicluster in o.a.a.start.Main
I just tried running accumulo miniscluster and saw the same thing. But in Main.err, not Master.err are you sure you saw this in Master.err? Has this ever worked? By default the accumulo scripts construct a very minimal classpath w/ accumulo-start.jar, log4j-1.2.15.jar, and the conf dir. If you modify the MAC exec method to print the classpath it uses to start a java process, then you can see this. MAC makes the assumption that everything it needs is on the Java classpath, which is true when its run from Maven or Eclipse. However when its run from the accumulo scripts, this is not true. Also, for some reason MAC starts zookeeper using Accumulo start main. I have no idea why its doing this. Even if it was not doing this, I think would fail in different way (i.e. instead of not finding VFS it would not find zookeeper class). Are you familiar with accumulo start module? On Mon, Oct 7, 2013 at 7:36 AM, Corey Nolet cjno...@gmail.com wrote: The MiniAccumuloRunner class that's wired up to o.o.a.start.Main. I was specifically wondering if anyone else is experiencing issues running 'accumulo minicluster' as both the proxy with useMini=true and the minicluster command seem broken for me. I'm building from remote HEAD in master. On Oct 6, 2013 11:32 AM, John Vines jvi...@gmail.com wrote: How are you running minicluster? On Sun, Oct 6, 2013 at 8:36 AM, Corey Nolet cjno...@gmail.com wrote: I'm having issues running the minicluster both in the 'accumulo proxy -p proxy.properties' and via 'accumulo minicluster'. It looks like the Zookeeper process is not starting and the MAC is going into an infinite loop waiting for it to start. I checked the Master.err logs for the minicluster command and I see the following: Uncaught exception: java.lang.NoClassDefFoundError: org/apache/commons/vfs2/impl/VFSClassLoader java.lang.NoClassDefFoundError: org/apache/commons/vfs2/impl/VFSClassLoader at java.lang.Class.getDeclaredMethods0(Native Method) at java.lang.Class.privateGetDeclaredMethods(Class.java:2521) at java.lang.Class.getMethod0(Class.java:2764) at java.lang.Class.getMethod(Class.java:1653) at org.apache.accumulo.start.Main.main(Main.java:42) Caused by: java.lang.ClassNotFoundException: org.apache.commons.vfs2.impl.VFSClassLoader at java.net.URLClassLoader$1.run(URLClassLoader.java:366) at java.net.URLClassLoader$1.run(URLClassLoader.java:355) at java.security.AccessController.doPrivileged(Native Method) at java.net.URLClassLoader.findClass(URLClassLoader.java:354) at java.lang.ClassLoader.loadClass(ClassLoader.java:424) at sun.misc.Launcher$AppClassLoader.loadClass(Launcher.java:308) at java.lang.ClassLoader.loadClass(ClassLoader.java:357) ... 5 more the commons-vfs2.jar is in Accumulo's lib directory. I'm using Hadoop 1.2.1. -- Cheers ~John
Re: Issues running Minicluster in o.a.a.start.Main
Keith, You are right- I mistyped. I meant Main.err not Master.err. I just verified this feature worked during the time of this commit: 6965a8aaa2f53ec796a3487c1639affe0dfc6bfa. On Mon, Oct 7, 2013 at 11:39 PM, Keith Turner ke...@deenlo.com wrote: I just tried running accumulo miniscluster and saw the same thing. But in Main.err, not Master.err are you sure you saw this in Master.err? Has this ever worked? By default the accumulo scripts construct a very minimal classpath w/ accumulo-start.jar, log4j-1.2.15.jar, and the conf dir. If you modify the MAC exec method to print the classpath it uses to start a java process, then you can see this. MAC makes the assumption that everything it needs is on the Java classpath, which is true when its run from Maven or Eclipse. However when its run from the accumulo scripts, this is not true. Also, for some reason MAC starts zookeeper using Accumulo start main. I have no idea why its doing this. Even if it was not doing this, I think would fail in different way (i.e. instead of not finding VFS it would not find zookeeper class). Are you familiar with accumulo start module? On Mon, Oct 7, 2013 at 7:36 AM, Corey Nolet cjno...@gmail.com wrote: The MiniAccumuloRunner class that's wired up to o.o.a.start.Main. I was specifically wondering if anyone else is experiencing issues running 'accumulo minicluster' as both the proxy with useMini=true and the minicluster command seem broken for me. I'm building from remote HEAD in master. On Oct 6, 2013 11:32 AM, John Vines jvi...@gmail.com wrote: How are you running minicluster? On Sun, Oct 6, 2013 at 8:36 AM, Corey Nolet cjno...@gmail.com wrote: I'm having issues running the minicluster both in the 'accumulo proxy -p proxy.properties' and via 'accumulo minicluster'. It looks like the Zookeeper process is not starting and the MAC is going into an infinite loop waiting for it to start. I checked the Master.err logs for the minicluster command and I see the following: Uncaught exception: java.lang.NoClassDefFoundError: org/apache/commons/vfs2/impl/VFSClassLoader java.lang.NoClassDefFoundError: org/apache/commons/vfs2/impl/VFSClassLoader at java.lang.Class.getDeclaredMethods0(Native Method) at java.lang.Class.privateGetDeclaredMethods(Class.java:2521) at java.lang.Class.getMethod0(Class.java:2764) at java.lang.Class.getMethod(Class.java:1653) at org.apache.accumulo.start.Main.main(Main.java:42) Caused by: java.lang.ClassNotFoundException: org.apache.commons.vfs2.impl.VFSClassLoader at java.net.URLClassLoader$1.run(URLClassLoader.java:366) at java.net.URLClassLoader$1.run(URLClassLoader.java:355) at java.security.AccessController.doPrivileged(Native Method) at java.net.URLClassLoader.findClass(URLClassLoader.java:354) at java.lang.ClassLoader.loadClass(ClassLoader.java:424) at sun.misc.Launcher$AppClassLoader.loadClass(Launcher.java:308) at java.lang.ClassLoader.loadClass(ClassLoader.java:357) ... 5 more the commons-vfs2.jar is in Accumulo's lib directory. I'm using Hadoop 1.2.1. -- Cheers ~John
Re: Issues running Minicluster in o.a.a.start.Main
How are you running minicluster? On Sun, Oct 6, 2013 at 8:36 AM, Corey Nolet cjno...@gmail.com wrote: I'm having issues running the minicluster both in the 'accumulo proxy -p proxy.properties' and via 'accumulo minicluster'. It looks like the Zookeeper process is not starting and the MAC is going into an infinite loop waiting for it to start. I checked the Master.err logs for the minicluster command and I see the following: Uncaught exception: java.lang.NoClassDefFoundError: org/apache/commons/vfs2/impl/VFSClassLoader java.lang.NoClassDefFoundError: org/apache/commons/vfs2/impl/VFSClassLoader at java.lang.Class.getDeclaredMethods0(Native Method) at java.lang.Class.privateGetDeclaredMethods(Class.java:2521) at java.lang.Class.getMethod0(Class.java:2764) at java.lang.Class.getMethod(Class.java:1653) at org.apache.accumulo.start.Main.main(Main.java:42) Caused by: java.lang.ClassNotFoundException: org.apache.commons.vfs2.impl.VFSClassLoader at java.net.URLClassLoader$1.run(URLClassLoader.java:366) at java.net.URLClassLoader$1.run(URLClassLoader.java:355) at java.security.AccessController.doPrivileged(Native Method) at java.net.URLClassLoader.findClass(URLClassLoader.java:354) at java.lang.ClassLoader.loadClass(ClassLoader.java:424) at sun.misc.Launcher$AppClassLoader.loadClass(Launcher.java:308) at java.lang.ClassLoader.loadClass(ClassLoader.java:357) ... 5 more the commons-vfs2.jar is in Accumulo's lib directory. I'm using Hadoop 1.2.1. -- Cheers ~John