Switched to jdk1.8 and the issue disappeared. Thanks!
However, someone mentioned in the hbase mailing list that the limit of 256 open files may be too low.
Do you agree? my system is the result of (update after update) an original r151006, maybe I still have an old value?
Thanks Olaf On 13/11/2016 18:37, Peter Tribble wrote:
On Sun, Nov 13, 2016 at 4:49 PM, Olaf Marzocchi <[email protected] <mailto:[email protected]>> wrote: Hello, I was trying to run hbase 1.2.4 (prerequisite for OpenTSDB) and I found that, contrary to my previous experience with hbase on OS X, it does not run out of the box (+ basic config). Instead, the java process only eats all the CPU. I tried to get a stack trace as suggested in the hbase mailing list (see the discussion at the end of this email), but jstack does not seems to work. This is the output (as also pasted in the link below): $ /usr/java/bin/jstack -m 16177 Attaching to process ID 16177, please wait... Exception in thread "main" java.lang.reflect.InvocationTargetException at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method) at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:57) at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43) at java.lang.reflect.Method.invoke(Method.java:606) at sun.tools.jstack.JStack.runJStackTool(JStack.java:136) at sun.tools.jstack.JStack.main(JStack.java:102) Caused by: java.lang.UnsatisfiedLinkError: no saproc in java.library.path at java.lang.ClassLoader.loadLibrary(ClassLoader.java:1889) at java.lang.Runtime.loadLibrary0(Runtime.java:849) at java.lang.System.loadLibrary(System.java:1088) at sun.jvm.hotspot.debugger.proc.ProcDebuggerLocal.<clinit>(ProcDebuggerLocal.java:660) at sun.jvm.hotspot.bugspot.BugSpotAgent.setupDebuggerSolaris(BugSpotAgent.java:634) at sun.jvm.hotspot.bugspot.BugSpotAgent.setupDebugger(BugSpotAgent.java:513) at sun.jvm.hotspot.bugspot.BugSpotAgent.go(BugSpotAgent.java:492) at sun.jvm.hotspot.bugspot.BugSpotAgent.attach(BugSpotAgent.java:331) at sun.jvm.hotspot.tools.Tool.start(Tool.java:163) at sun.jvm.hotspot.tools.JStack.main(JStack.java:86) ... 6 more Both java and jdk packages are installed from main OmniOS repo and compiled for r151020, the OS version I'm running right now You're using the openjdk 1.7.0_101 that comes with omnios? You don't need hbase to trigger this. I just tested jstack with an arbitrary pid, and it fails the same way. I'm not sure why, libsaproc.so is installed where I would expect it to be, but jstack looks for it somewhere else. However, the one thing I would say is that for any real work, don't use the jdk that omnios supplies. It's old (java 7) and 32-bit only. At $DAYJOB, we use the Oracle jdk8 and it works fine. (Although my own builds of openjdk7 and openjdk8 are also fine, but it's easier to rely on someone else for maintaining something like java.) I would appreciate help in the solution of this problem so that the guys from the hbase mailing list can pinpoint the issue about hbase under OmniOS. Of course, if anyone has experience with hbase on OmniOS and has seen similar issues, or is successfully running previous versions, feel free to let me know. I haven't checked any previous version, I only tested the current stable. Thanks Olaf -------- Forwarded Message -------- Subject: Re: "Quickstart" on OmniOS/illumos not working, java eats CPU Date: Sun, 13 Nov 2016 08:28:42 -0800 From: Ted Yu <[email protected] <mailto:[email protected]>> Reply-To: [email protected] <mailto:[email protected]> To: [email protected] <mailto:[email protected]> <[email protected] <mailto:[email protected]>> bq. Caused by: java.lang.UnsatisfiedLinkError: no saproc in java.library.path Was the jstack of version 1.7.0_101 ? I used google and found: http://umchee.blogspot.com/p/i-recently-needed-to-investigate.html <http://umchee.blogspot.com/p/i-recently-needed-to-investigate.html> but it was about Java 1.5 You can use this command to find the commandline for hbase process (replace 16059 with the new process Id): ps aux | grep 16059 Then use the jstack from same path as java to capture stack trace: Thanks On Sun, Nov 13, 2016 at 8:16 AM, Olaf Marzocchi <[email protected] <mailto:[email protected]>> wrote: I'm not 100% sure about the stack trace (not a developer myself), but I searched online and used jstack: http://pastebin.com/AT6pScBv Additional info: after turning on DEBUG, the java process did not immediately jump to 100% cpu as it used to do, it ramped up and it took 3 minutes to get there. It may however be a coincidence. Hope this can help pinpointing the issue. Olaf On 13/11/2016 01:18, Ted Yu wrote: Can you turn on DEBUG logging ? In log4j.properties : log4j.logger.org.apache.hadoop.hbase=DEBUG Since this is reproducible, can you capture one or two stack traces before killing the server ? Thanks On Sat, Nov 12, 2016 at 3:09 PM, Olaf Marzocchi <[email protected] <mailto:[email protected]>> wrote: Hello, I tested HBase with OpenTSDB some time ago on OS X locally and the quickstart guide worked perfectly, merged with http://opentsdb.net/setup-hbase.html <http://opentsdb.net/setup-hbase.html> where necessary. I tried yesterday to do the same on my OmniOS/illumos server and I encountered issues with HBase. First of all I set "JAVA_HOME=/usr/java" in "conf/hbase-env.sh". $ /usr/java/bin/java -version openjdk version "1.7.0_101" OpenJDK Runtime Environment (build 1.7.0_101-b00) OpenJDK Server VM (build 24.101-b00, mixed mode) I edited also "hbase-site.xml": <configuration> <property> <name>hbase.rootdir</name> <value>file:///export/home/olaf/hbase</value> </property> <property> <name>hbase.zookeeper.property.dataDir</name> <value>/export/home/olaf/zookeeper</value> </property> </configuration> I then launched "bin/start-hbase.sh" and it quits apparently normally. A directory "/export/home/olaf/zookeeper" appeared, but no /export/home/olaf/hbase. $ ls -l /export/home/olaf/ drwxrwxr-x 3 olaf olaf 3 Nov 12 01:16 zookeeper I also checked with prstat ("top") and I see that java is taking a full CPU. After some minutes it's still crunching something. If I launch "./bin/hbase shell" it never gets to "hbase(main):001:0>". At this point the only way is to kill -9 both java processes: the stop script doesn't work, it outputs dots forever. I uploaded the two log files (.log and .out) to Pastebin: http://pastebin.com/aFMXmWYU I'm not sure where to begin with the debugging, since I did no customisation at all and I don't see any error logged. I would be grateful for some help. Thanks Olaf _______________________________________________ OmniOS-discuss mailing list [email protected] <mailto:[email protected]> http://lists.omniti.com/mailman/listinfo/omnios-discuss <http://lists.omniti.com/mailman/listinfo/omnios-discuss> -- -Peter Tribble http://www.petertribble.co.uk/ - http://ptribble.blogspot.com/
_______________________________________________ OmniOS-discuss mailing list [email protected] http://lists.omniti.com/mailman/listinfo/omnios-discuss
