Riak+Innostore segfaulting on Debian Lenny

classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|

Riak+Innostore segfaulting on Debian Lenny

Michael Montano
I built Riak and Innostore from the basho repository, followed the  
instructions for configuring innostore and successfully started a  
node. bin/riak-admin test was passing and I was able to read/write to  
the store. However, riak kept crashing after about 60 seconds. Works  
fine with riak_dets_backend. Anything I should be aware of re  
innostore on debian?

from /var/log/messages (the other logs had nothing)

Feb 22 09:09:00 riak[1571]: Starting up
Feb 22 09:09:26 kernel: beam.smp[1654]: segfault at 000000000c1fe000  
rip 00002aaaaba342af rsp 000000005eaf1fd8 error 4
Feb 22 09:09:49 riak[1688]: Starting up
Feb 22 09:10:52 kernel: beam.smp[1744]: segfault at 000000000c1eb000  
rip 00002aaaaba34254 rsp 000000004d2d5fd8 error 4
Feb 22 09:18:55 riak[1905]: Starting up
Feb 22 09:19:21 kernel: beam.smp[2013]: segfault at 00000000071e7000  
rip 00002aaaaba3426c rsp 0000000055ee3fd8 error 4'

mike

_______________________________________________
riak-users mailing list
[hidden email]
http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com
Reply | Threaded
Open this post in threaded view
|

Re: Riak+Innostore segfaulting on Debian Lenny

Dave Smith
On Tue, Feb 23, 2010 at 1:58 PM, Michael Montano <[hidden email]> wrote:
I built Riak and Innostore from the basho repository, followed the instructions for configuring innostore and successfully started a node. bin/riak-admin test was passing and I was able to read/write to the store. However, riak kept crashing after about 60 seconds. Works fine with riak_dets_backend. Anything I should be aware of re innostore on debian?

Hum, weird.

64 or 32 bit debian? What version of innostore? Can you get a backtrace of the crash?

I've tested on debian lenny 64 and it's fine -- not an excuse, just an observation. :)

D.

_______________________________________________
riak-users mailing list
[hidden email]
http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com
Reply | Threaded
Open this post in threaded view
|

Re: Riak+Innostore segfaulting on Debian Lenny

Dave Smith
Michael,

I put together a script that should make it easy to extract the backtrace:


Place the script in the rel/riak directory (assuming you built from source). Run it by doing:

./coreinfo <corefile>

This will generate a gdb.txt file that you can then send back for review. 

If you aren't getting a core file, make sure to set: "ulimit -c unlimited" Of course, you'll also need gdb installed on the machine that is crashing.

Let me know what you find.

Thanks,

D.

On Wed, Feb 24, 2010 at 3:28 AM, Michael Montano <[hidden email]> wrote:
64-bit. Reproduced on a clean install. Was running fine for about 10 minutes of light writing, then segfaulted. I was able to start it up again, but kept crashing every few minutes after that. How can I get the backtrace?

On Feb 23, 2010, at 1:02 PM, David Smith wrote:

On Tue, Feb 23, 2010 at 1:58 PM, Michael Montano <[hidden email]> wrote:
I built Riak and Innostore from the basho repository, followed the instructions for configuring innostore and successfully started a node. bin/riak-admin test was passing and I was able to read/write to the store. However, riak kept crashing after about 60 seconds. Works fine with riak_dets_backend. Anything I should be aware of re innostore on debian?

Hum, weird.

64 or 32 bit debian? What version of innostore? Can you get a backtrace of the crash?

I've tested on debian lenny 64 and it's fine -- not an excuse, just an observation. :)

D.



_______________________________________________
riak-users mailing list
[hidden email]
http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com
Reply | Threaded
Open this post in threaded view
|

Re: Riak+Innostore segfaulting on Debian Lenny

Dave Smith
(Copying riak-users...)

Mike,

The crasher in innostore should now be fixed. There was a type cast bug when extracting the length of the key. The erlang code ensured that keys > 255 bytes wouldn't be a problem, but the range between [127, 254] of key lengths was still causing problems. I've added a unit test to verify this is not going to be a problem in the future.

Please grab release 6 from the innostore page to get this bug fix. Let me know if you see any further wonky behaviour. Thanks for providing that backtrace. 

D.

_______________________________________________
riak-users mailing list
[hidden email]
http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com