Description of problem: ceph-fuse is being built without being linked to tcmalloc. The default malloc is very bad; a ceph-users report on centos7 says he easily exceeds 10GB of VSS per process, and it is correlated with other issues that are still being pinned down. (Past experience has shown us that using gmalloc also results in much worse throughput/IOPs in general.) Version-Release number of selected component (if applicable): Downstream releases based on upstream Jewel, as well as upstream releases. See http://tracker.ceph.com/issues/16655 How reproducible: It's a build issue. Steps to Reproduce: 1. Build ceph-fuse 2. Observe it is not linked with tcmalloc Expected results: ceph-fuse is linked with tcmalloc Additional info:
QE team: you can check this link by running "ldd /usr/bin/ceph-fuse | grep tcmalloc"
patch has been pushed into ceph-2-rhel-patches
As indicated by Ken, I have run the following command on RHEL and ubuntu, -----ldd /usr/bin/ceph-fuse | grep tcmalloc and here is the output ------libtcmalloc.so.4 => /lib64/libtcmalloc.so.4 (0x00007fce1a298000) so moving the bug to verifed stage.
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://rhn.redhat.com/errata/RHBA-2016-1755.html