Permalink
Commits on Dec 6, 2018
Commits on Dec 4, 2018
Commits on Nov 30, 2018
Commits on Nov 27, 2018
  1. Minor change to rebuilder to allow to build against threaded erasureU…

    gransom committed Nov 27, 2018
    …tils. However, the rebuilder still requires testing before being run in production.
Commits on Nov 1, 2018
  1. compile-time fix for non-sockets builds.

    jti-lanl committed Nov 1, 2018
    For non-sockets builds, we had the wrong sense of a compile-time test.
    This would've caused pftool to spew warnings about reading the
    ~/.awsAuth file, but was hidden by the same thing that required
    erasureUtils 54bea71ba066.  Unhidden by that, and now fixed by this.
Commits on Oct 31, 2018
  1. Detect build-mismatches with libne (regarding uDAL sockets) at config…

    jti-lanl committed Oct 31, 2018
    …ure-time.
    
    Note that if your MARFSCONFIGRC has one or more RDMA repos, it will
    fail to load (at run-time) unless you have built both libne and marfs
    to support RDMA-sockets-based uDALs.
Commits on Oct 25, 2018
Commits on Oct 18, 2018
  1. Add the new TimingData to MarFS_FileHandle (see erasureUtils 198f6b99…

    jti-lanl committed Oct 18, 2018
    …56b9)
    
    This avoids the need to malloc a static blob, in order to hold ad hoc
    timing-data collected inside libne, during low-level operations.  The
    timing-data optionally maintained in ne_handle can not survive the
    free of the ne_handle in ne_close(), which is also potentially
    something we want to time.  Thus, we had previously malloc'ed a blob,
    to hold a subset of this data, but we in turn depended on pftool to
    free it, which leaks memory in the case of MarFS fuse.
    
    erasureUtils now allows us to supercede the TimingData in ne_handle by
    passing in a ptr to our own TimingData.  Thus, no need to malloc,
    because our TimningData lives beyond ne_close().
Commits on Oct 15, 2018
  1. Updated marfs_cctest.cfg, to try to improve pedagogical value.

    jti-lanl committed Oct 15, 2018
    Some old (untested) remnants of Scality/EMC interaction, plus Multi-Component @ NFS and RDMA.
  2. Merge branch 'rdma'

    jti-lanl committed Oct 15, 2018
Commits on Oct 10, 2018
  1. timing-statistics support in mc_open() shouldn't allocate storage unl…

    jti-lanl committed Oct 10, 2018
    …ess data will be collected.
Commits on Oct 5, 2018
Commits on Oct 1, 2018
  1. Tweaked comment.

    jti-lanl committed Oct 1, 2018
Commits on Sep 17, 2018
  1. Install 'rebuilder' and 'mc_path_convert' apps into the build destina…

    jti-lanl committed Sep 17, 2018
    …tion.
    
    I don't see a way to get autotools to nicely build some applications
    with a different compiler (e.g. rebuilder needs mpicc), which would
    also take care of putting everything into the build destination.
    Slightly refactored @gransom's approach, while I tweaking the
    build-destination for 'rebuilder'.
Commits on Sep 13, 2018
Commits on Sep 12, 2018
Commits on Sep 11, 2018
  1. Merge branch 'master' into rdma_merge_master

    jti-lanl committed Sep 11, 2018
    Conflicts:
    	fuse/src/marfs_ops.c
  2. forgot to add path_convert.c

    leicao88124 authored and gransom committed Jul 16, 2018
    (cherry picked from commit c827cc4)
  3. Revert "Added in a modified version of Lei's path_convert code"

    gransom committed Sep 11, 2018
    This reverts commit 5c68352.
    I will be pulling in Lei's path_convert code via merge rather than copy-paste.
Commits on Sep 10, 2018
Commits on Sep 8, 2018
  1. missing semi

    jti-lanl committed Sep 8, 2018