there is a memleak in glusterd. This is the valgrind report indicating the leak. ==18038== LEAK SUMMARY: ==18038== definitely lost: 100,623 bytes in 525 blocks ==18038== indirectly lost: 0 bytes in 0 blocks ==18038== possibly lost: 17,456,525 bytes in 1,691 blocks ==18038== still reachable: 3,372,663 bytes in 127 blocks ==18038== suppressed: 0 bytes in 0 blocks ==18038== Reachable blocks (those to which a pointer was found) are not shown. ==18038== To see them, rerun with: --leak-check=full --show-reachable=yes ==18038== the volgrind log file is attached
Created attachment 386
I repeated the tests and these are the new valgrind logs indicating the leak. I cleaned up the mount point after the tests (posix compliance, dbench 100, areual /etc) and then unmounted it.
Created attachment 387
PATCH: http://patches.gluster.com/patch/5739 in master (Remove spurious inode_ref call on parent dir in fuse_create_cbk)
*** Bug 2114 has been marked as a duplicate of this bug. ***
some of the leaks are due to fini not being invoked. Once that is fixed, we can revisit those leaks
Repeated the same tests (posix_compliance, arequal of /etc, dbench 100). This is the output of the valgrind. CLIENT: ==6991== LEAK SUMMARY: ==6991== definitely lost: 180 bytes in 2 blocks ==6991== indirectly lost: 152 bytes in 1 blocks ==6991== possibly lost: 20,111,663 bytes in 116 blocks ==6991== still reachable: 3,366,649 bytes in 118 blocks ==6991== suppressed: 0 bytes in 0 blocks SERVER: ==6962== LEAK SUMMARY: ==6962== definitely lost: 221 bytes in 3 blocks ==6962== indirectly lost: 0 bytes in 0 blocks ==6962== possibly lost: 2,857,856 bytes in 53 blocks ==6962== still reachable: 8,995,178 bytes in 185 blocks ==6962== suppressed: 0 bytes in 0 blocks