[Gluster-devel] "gluster vol start" is failing when glusterfs is compiled with debug enable .

Wed Jul 22 12:42:10 UTC 2015

On Wed, 2015-07-22 at 12:19 +0530, Anand Nekkunti wrote:
> 
> On 07/22/2015 11:47 AM, Raghavendra Bhat wrote:
> > On 07/22/2015 09:50 AM, Atin Mukherjee wrote:
> > > 
> > > On 07/22/2015 12:50 AM, Anand Nekkunti wrote:
> > > > Hi All
> > > >     "gluster vol start" is failing when glusterfs is compiled 
> > > > with 
> > > > debug
> > > > enable .
> > > > Link: :https://bugzilla.redhat.com/show_bug.cgi?id=1245331
> > > > 
> > > > *brick start is failing with fallowing error:*
> > > > 2015-07-21 19:01:59.408729] I [MSGID: 100030] 
> > > > [glusterfsd.c:2296:main]
> > > > 0-/usr/local/sbin/glusterfsd: Started running 
> > > > /usr/local/sbin/glusterfsd
> > > > version 3.8dev (args: /usr/local/sbin/glusterfsd -s 192.168.0.4
> > > > --volfile-id VOL.192.168.0.4.tmp-BRICK1 -p
> > > > /var/lib/glusterd/vols/VOL/run/192.168.0.4-tmp-BRICK1.pid -S
> > > > /var/run/gluster/0a4faf3d8d782840484629176ecf307a.socket -
> > > > -brick-name
> > > > /tmp/BRICK1 -l /var/log/glusterfs/bricks/tmp-BRICK1.log -
> > > > -xlator-option
> > > > *-posix.glusterd-uuid=4ec09b0c-6043-40f0-bc1a-5cc312d49a78 -
> > > > -brick-port
> > > > 49152 --xlator-option VOL-server.listen-port=49152)
> > > > [2015-07-21 19:02:00.075574] I [MSGID: 101190]
> > > > [event-epoll.c:627:event_dispatch_epoll_worker] 0-epoll: 
> > > > Started thread
> > > > with index 1
> > > > [2015-07-21 19:02:00.078905] W [MSGID: 101095]
> > > > [xlator.c:189:xlator_dynload] 0-xlator: 
> > > > /usr/local/lib/libgfdb.so.0:
> > > > undefined symbol: gf_sql_str2sync_t
> > > > [2015-07-21 19:02:00.078947] E [MSGID: 101002] 
> > > > [graph.y:211:volume_type]
> > > > 0-parser: Volume 'VOL-changetimerecorder', line 16: type
> > > > 'features/changetimerecorder' is not valid or not found on this 
> > > > machine
> > > > [2015-07-21 19:02:00.079020] E [MSGID: 101019] 
> > > > [graph.y:319:volume_end]
> > > > 0-parser: "type" not specified for volume VOL
> > > > -changetimerecorder
> > > > [2015-07-21 19:02:00.079150] E [MSGID: 100026]
> > > > [glusterfsd.c:2151:glusterfs_process_volfp] 0-: failed to 
> > > > construct the
> > > > graph
> > > > [2015-07-21 19:02:00.079399] W 
> > > > [glusterfsd.c:1214:cleanup_and_exit]
> > > > (-->/usr/local/sbin/glusterfsd(mgmt_getspec_cbk+0x343) 
> > > > [0x40df64]
> > > > -->/usr/local/sbin/glusterfsd(glusterfs_process_volfp+0x1a2) 
> > > > [0x409b58]
> > > > -->/usr/local/sbin/glusterfsd(cleanup_and_exit+0x77) [0x407a6f] 
> > > > ) 0-:
> > > > received signum (0), shutting down
> > > I am not able to hit this though.
> > 
> > This seems to be the case of inline functions being considerd as 
> > undefined symblols. There has been a discussion about it in the 
> > mailing list.
> > 
> > https://www.gluster.org/pipermail/gluster-devel/2015
> > -June/045942.html
> it seems issue with inline functions , but why it is happening in 
> only 
> in debug build ?

AFAIK, this is the reason why these undefined errors are not seen with 
--enable-debug option.

When --enable-debug is specified, as per the configure.ac script we
doesn't include the optimization level -O2. Instead we add -O0. See the
following snippet.

-------------------------------------------
if test "x$enable_debug" = "xyes"; then
         BUILD_DEBUG=yes
         CFLAGS="${CFLAGS} -g -O0 -DDEBUG"
else                                                                   

         BUILD_DEBUG=no
         CFLAGS="${CFLAGS} -g -O2"
fi
-------------------------------------------

with -O0 optimization level, the compiler doesn't do inlining, so it
requires extern definitions of the inline functions. It's the problem
described at https://gcc.gnu.org/gcc-5/porting_to.html regarding
porting changes with GCC v5.x.

>From gcc man page

-----------------------------------------------------------------------
-O0 Reduce compilation time and make debugging produce the expected
results. This is the default.

-O2 Optimize even more. GCC performs nearly all supported optimizations
that do not involve a space-speed tradeoff.  As compared to -O, this
option increases both compilation time and the performance of the
generated code.
-----------------------------------------------------------------------

So now we have two options:

[1] Use -fgnu89-inline. [Refer the other thread for more details (
https://www.gluster.org/pipermail/gluster-devel/2015-June/045942.html)]
[2] Fix the code to work with C99 inline semantics by either changing
all inline to static inline or provide external definitions for each
inline functions in .h file
[3] Or fix the the optimization flags in configure.ac accordingly.

Thanks,
Anoop C S.

> > 
> > Regards,
> > Raghavendra Bhat
> > 
> > > > 
> > > > Thanks&Regards
> > > > Anand.N
> > > > 
> > > > 
> > > > 
> > > > _______________________________________________
> > > > Gluster-devel mailing list
> > > > Gluster-devel at gluster.org
> > > > http://www.gluster.org/mailman/listinfo/gluster-devel
> > > > 
> > 
> > _______________________________________________
> > Gluster-devel mailing list
> > Gluster-devel at gluster.org
> > http://www.gluster.org/mailman/listinfo/gluster-devel
> 
> _______________________________________________
> Gluster-devel mailing list
> Gluster-devel at gluster.org
> http://www.gluster.org/mailman/listinfo/gluster-devel