Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
tag: orion-2_2_49_5…
Commits on Mar 6, 2012
  1. @miketappro

    ORI-540 osd-zfs: do not access SA on quota objects

    Alex Zhuravlev authored miketappro committed
    as they are of DMU_OT_NONE type and do not contain SA
    
    Signed-off-by: Alex Zhuravlev <bzzz@whamcloud.com>
    Change-Id: I5722a2604de32fdb7bb45933a6d521021cfeea11
    Reviewed-on: http://review.whamcloud.com/2220
    Reviewed-by: Johann Lombardi <johann@whamcloud.com>
    Tested-by: Hudson
Commits on Mar 5, 2012
  1. @miketappro

    ORI-386 ofd: reject connection from old MDS

    Alex Zhuravlev authored miketappro committed
    as MDS-OST protocol has been changed by ORION, we do not allow
    old MDS to be used with new OST (ofd-based)
    
    Change-Id: I416b9dc3246c585d85be9135603eacae9b0ef2fe
    Signed-off-by: Alex Zhuravlev <bzzz@whamcloud.com>
    Reviewed-on: http://review.whamcloud.com/2108
    Tested-by: Hudson
Commits on Mar 2, 2012
  1. @miketappro

    ORI-208 llog: recognize old (pre-osd) logid

    Alex Zhuravlev authored miketappro committed
    and convert them into IGIFs to enable forward (not backward)
    compatibility with 2.2.
    
    Change-Id: I88e045fe24a392c47392b9b2d8b16a2aee0a91f3
    Signed-off-by: Alex Zhuravlev <bzzz@whamcloud.com>
    Reviewed-on: http://review.whamcloud.com/2119
    Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
    Tested-by: Hudson
  2. @miketappro

    ORI-500 ofd: Fix return code handlings in ofd_write_attr_set()

    Li Wei authored miketappro committed
    Niu Yawei told me that he had also seen the status "32" in OST_WRITE RPC
    replies during his local testing and traced it back to a return code
    handling bug in ofd_write_attr_set().  We figured out that the "32" must
    be the filter_fid EA size returned by dt_xattr_get().  Given the time
    this started to appear and the RPC status "32", I believe it is also the
    root cause behind the iozone failures.  This patch fixes the return
    handlings and makes ofd_object_ff_check() to return only zero or
    negative---EA sizes are useless for upper layers.
    
    Change-Id: I270a9ee04f29f78949a3dc831226b581c6714a95
    Signed-off-by: Li Wei <liwei@whamcloud.com>
    Reviewed-by: Niu Yawei <niu@whamcloud.com>
    Reviewed-by: Johann Lombardi <johann@whamcloud.com>
  3. @miketappro

    ORI-509 llog: Ensure names are deleted along with logs

    Li Wei authored miketappro committed
    When opening an existing log by name, llog_osd_open() doesn't fill the
    name into the log handle.  If the log handle is then fed into
    llog_osd_destroy() (e.g., in the case of llog_erase()), the name will
    not be deleted from the index, causing the back end file system to be
    corrupted.  This patch fixes this problem by always saving names in log
    handles when opening by name.  Also, a minor fix of the CONFIGS
    directory mode is included.
    
    Change-Id: Ie7e77fd2d8daf9cd8db84c4f8f2c0c93ca2fb380
    Signed-off-by: Li Wei <liwei@whamcloud.com>
    Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
    Reviewed-by: Alex Zhuravlev <bzzz@whamcloud.com>
  4. @miketappro

    ORI-390 grant: acquire grant space for precreate

    Johann Lombardi authored miketappro committed
    This patch changes the precreate path to acquire grant space.
    This avoids running into the situation where object precreation
    consumes space already granted to clients. Space reserved for object
    creation is hold by the self export.
    An initial amount of space is granted at the end of recovery and more
    space is proactively reserved after each creation request handling.
    Grant checks are also not enforced during recovery and orphan cleanup.
    The patch also fills ddp_inodespace/ddp_grant_frag for ZFS OSD.
    
    Change-Id: I7cc68357dbf44ec0f78a0cafe66c8d0e297160f2
    Signed-off-by: Johann Lombardi <johann@whamcloud.com>
    Reviewed-on: http://review.whamcloud.com/1673
  5. @miketappro

    ORI-508 grant: release grant when destroying export

    Johann Lombardi authored miketappro committed
    Unlike obdfilter, the OFD code only releases grant when disconnecting
    the export and not when destroying it. The problem is that there can
    still be requests in flight under processing at the time
    ofd_obd_disconnect() is invoked and we should still release any
    additional grant space allocated to the evicted export.
    
    This patch modifies the OFD layer to behave like obdfilter.
    
    Signed-off-by: Johann Lombardi <johann@whamcloud.com>
    Change-Id: I2b747eed8ce74ce9e3e46f5ac0dcccb8d3eee494
    Reviewed-on: http://review.whamcloud.com/2177
  6. @miketappro

    ORI-473 grant: sync backend filesystem less aggressively

    Johann Lombardi authored miketappro committed
    The grant code should not sync the backend filesystem to force block
    allocation when there is no need to acquire grant space on the
    server side (i.e. all pages of the OST_WRITE have the
    OBD_BRW_FROM_GRANT flat set).
    
    Signed-off-by: Johann Lombardi <johann@whamcloud.com>
    Change-Id: Iab573d27bd26f05b59b677389a01fda4fb6dbce5
    Reviewed-on: http://review.whamcloud.com/2027
  7. @miketappro

    ORI-208 osp: do not set OBD_MD_FLOBJCOUNT for old llog records

    Alex Zhuravlev authored miketappro committed
    if osp meet a record generated by 2.2 with zero count field,
    do not set OBD_MD_FLOBJCOUNT - OFD will recognize this as
    count=1
    
    Change-Id: Ib9a643ceb5161f2a3bdb094a78d82ad2e6b3ca31
    Signed-off-by: Alex Zhuravlev <bzzz@whamcloud.com>
    Reviewed-on: http://review.whamcloud.com/2120
    Reviewed-by: Mike Pershin <tappro@whamcloud.com>
    Tested-by: Hudson
  8. @prakashsurya @miketappro

    ORI-375 tunefs: Populate mop with device name sooner

    prakashsurya authored miketappro committed
    The tunefs.lustre command was not populating it's mkfs_opts structure
    with the device name prior to calling the osd_is_lustre. Thus, when
    osd_is_lustre function was called it had no chance of properly detecting
    whether a given device was formatted with mkfs.lustre. This change fixes
    issue by strscpy-ing the command's last argument into the mkfs_opts
    structure prior to calling osd_is_lustre.
    
    Signed-off-by: Prakash Surya <surya1@llnl.gov>
    Change-Id: Idc8967bc2297e3a6ed9ebfc7613ec2f5dac73f43
    Reviewed-by: Niu Yawei <niu@whamcloud.com>
    Reviewed-by: Li Wei <liwei@whamcloud.com>
  9. @prakashsurya @miketappro

    ORI-375 tunefs: Call parse_opts after osd_read_ldd

    prakashsurya authored miketappro committed
    The parse_opts function must be called after the osd_read_ldd function
    is called. This is to ensure that the values read from disk are
    overwritten by the values passes as options to the command. Otherwise,
    the on disk options would always be used, regardless of what is passed
    in to the command.
    
    Signed-off-by: Prakash Surya <surya1@llnl.gov>
    Change-Id: I2f4c236849b82b6e112cc2f558e46ba6b92e6eb9
    Reviewed-by: Niu Yawei <niu@whamcloud.com>
    Reviewed-by: Alex Zhuravlev <bzzz@whamcloud.com>
  10. @miketappro

    ORI-254 ofd: Set filter_fid EA for OST_{SETATTR,PUNCH}

    Li Wei authored miketappro committed
    Current OFD doesn't set filter_fid EAs for OST_SETATTR and OST_PUNCH
    requests, like what OBDFilter does.  This patch adds the missing logic
    to OFD.  Now filter_fid EAs are set whenever absent, disregarding
    SUID/SGID bits.
    
    This patch also fixes the SUID/SGID handling code for OST_SETATTR and
    OST_PUNCH requests so that SUID/SGID bits are cleared wheneven UID/GIDs
    have been initialized.  To reduce duplicate code, that is implemented by
    introducing a helper function, ofd_attr_handle_ugid().
    
    Change-Id: I67246d9c2cf55e94160b794b0fcad613925dbaea
    Signed-off-by: Li Wei <liwei@whamcloud.com>
    Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
    Reviewed-by: Alex Zhuravlev <bzzz@whamcloud.com>
  11. @miketappro

    ORI-493 grant: always release pending space

    Johann Lombardi authored miketappro committed
    If ofd_preprw_write() fails, ofd_grant_commit() isn't invoked whereas
    ofd_grant_prepare_write() has been called already. As a result, the
    "pending" space might never be released.
    This patch makes sure that ofd_grant_commit() is always calls after
    ofd_grant_prepare_write().
    
    Signed-off-by: Johann Lombardi <johann@whamcloud.com>
    Change-Id: Icc112ff979df9d4b10b4fb6efadd00f86a448cb2
    Reviewed-on: http://review.whamcloud.com/2093
    Tested-by: Hudson
    Reviewed-by: Alex Zhuravlev <bzzz@whamcloud.com>
  12. @miketappro

    ORI-439 osd-zfs: fix code duplication in mount_utils_zfs.c

    Johann Lombardi authored miketappro committed
    Use zfs_fini() to close the zfs library when zfs_init() failed.
    
    Signed-off-by: Johann Lombardi <johann@whamcloud.com>
    Change-Id: Ifcfb98e0222a41da0b233bad14fc9824099722c7
    Reviewed-on: http://review.whamcloud.com/1806
  13. @miketappro

    ORI-360 osd-zfs: place xattrs in SA if possible

    miketappro authored
    Extended attributes are stored in SA. SA are placed in dnode but
    if size if not enough the additional block is used. Therefore
    the maximum size is 128K for SA xattrs.
    
    Current code is trying to work with SA but old ea-in-dnode way is
    not yet removed and is used as fallback for very big xattrs.
    
    Signed-off-by: Mikhail Pershin <tappro@whamcloud.com>
    Change-Id: I59580389a93b82b0e41920ceae6f8542eaacbe50
    Reviewed-on: http://review.whamcloud.com/1992
    Tested-by: Hudson
    Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
  14. @miketappro

    ORI-360 osd-zfs: cache SA values in osd_object

    miketappro authored
    - cache object attributes in osd_object to avoid SA lookup every time
    - use lu_env to store stack values
    - protect object attributes with read/write lock
    - replace old udmu ASSERT with Lustre LASSERT
    - remove unused old udmu code
    - add error handling and reporting in code
    - all osd and __osd functions returns negative error codes
    - use 'rc' instead of 'error' to keep common style
    
    Signed-off-by: Mikhail Pershin <tappro@whamcloud.com>
    Change-Id: I01c2b443199d12676e081289fcc9789c1e807a16
    Reviewed-on: http://review.whamcloud.com/2006
    Reviewed-by: Alex Zhuravlev <bzzz@whamcloud.com>
    Tested-by: Hudson
  15. @miketappro

    ORI-227 tests: Fix an OFD parameter reference in lfsck.sh

    Li Wei authored miketappro committed
    Currently, OFD is still known as "obdfilter" to lctl(8).  Before we move
    to "ofd", test scripts should keep using "obdfilter" to refer to OFD
    parameters.  This patch fixes such an incorrect reference in lfsck.sh.
    
    Change-Id: Ibe2b49c19f66f2db78b8631ecc5ce518055fc660
    Signed-off-by: Li Wei <liwei@whamcloud.com>
    Reviewed-on: http://review.whamcloud.com/2039
    Tested-by: Hudson
    Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
    Tested-by: Maloo <whamcloud.maloo@gmail.com>
  16. @miketappro

    ORI-360 osd-zfs: replace znode_phys_t with SA

    miketappro authored
    Update osd-zfs code to store/fetch attributes from SA instead of
    znode_phys_t. This patch doesn't place extended attributes in SA yet
    
    Signed-off-by: Mikhail Pershin <tappro@whamcloud.com>
    Change-Id: I50a5f655080a3eda05b11056e358c514bab5f507
    Reviewed-on: http://review.whamcloud.com/1991
    Tested-by: Hudson
    Tested-by: Maloo <whamcloud.maloo@gmail.com>
    Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
    Reviewed-by: Alex Zhuravlev <bzzz@whamcloud.com>
  17. @miketappro

    ORI-453 tests: unlink all files when mdsrate times out

    Johann Lombardi authored miketappro committed
    When mdsrate cannot create the requested number of files in the
    allocated timeframe (i.e. the --time option), filenames are not
    necessarily consecutive and can have an index higher than the
    number of files that was actually created. Here is an example:
    
    0: iters 0 nfiles 100000 time 1 begin 0 end 99998 dirthreads 2.
    1: iters 0 nfiles 100000 time 1 begin 1 end 99999 dirthreads 2.
    0: client-4.lab.whamcloud.com starting at Mon Jan  9 08:26:07 2012
    1: client-12.lab.whamcloud.com starting at Mon Jan  9 08:26:07 2012
    0: created 1884 files, last file 'f3766'.
    Rate: 2470.42 eff 2470.59 aggr 1235.29 avg client creates/sec
         (total: 2 threads 2440 creates 1 dirs 2 threads/dir 0.99 secs)
    0: client-4.lab.whamcloud.com finished at Mon Jan  9 08:26:08 2012
    1: created 556 files, last file 'f1111'.
    1: client-12.lab.whamcloud.com finished at Mon Jan  9 08:26:08 2012
    
    2440 files have been created and the last file name was f1111 for
    rank 1 and f3766 for rank 0. Running mdsrate --unlink --nfiles 2440
    in this case would discard all files with an index >= 2440.
    
    With this patch, each mdsrate task will complete properly and report
    the right number of executed operation as soon as one file is missing
    in the per-task sequence. This behavior can be changed with the
    --ignore and --recreate option which are already available.
    This way, mdsrate --unlink --nfiles 100000 will work fine in the
    example above since it will unlink all the existing files, without
    doing tons of useless lookups.
    
    Signed-off-by: Johann Lombardi <johann@whamcloud.com>
    Change-Id: If520ba5e3f638e17132fee60d8acbe17b5f1cf74
    Reviewed-on: http://review.whamcloud.com/1932
  18. @miketappro

    ORI-478 tests: Fix "fs2" device references in conf-sanity

    Li Wei authored miketappro committed
    Regardless of FSTYPE, Autotest sets fs2<facet>_DEVs to block devices.
    This matches the usage of <FACET>DEV<n> (e.g., MDSDEV1) but conflicts
    with what some conf-sanity tests, which expect fs2<facet>_DEVs to be ZFS
    file system names and fs2<facet>_VDEVs to contain block devices, when
    FSTYPE is "zfs".  This patch fixes the tests so that "fs2" devices are
    specified in a uniform manner.
    
    Change-Id: I5b436f4fe93f6c2be0ece99d1051edeaefa2a574
    Signed-off-by: Li Wei <liwei@whamcloud.com>
    Reviewed-by: Johann Lombardi <johann@whamcloud.com>
  19. @miketappro

    ORI-416 tests: Don't look for ZFS utilities on clients

    Li Wei authored miketappro committed
    Test Framework looks for ZFS utilities on clients, which are unlikely to
    have ZFS installed.  This patch removes the "which" checks, as in most
    cases ZFS is installed in standard paths.  A clean fix depends on
    LU-862.
    
    Change-Id: I5df32f7b47b9b440406bde2a1cf85cfd1bbfb869
    Signed-off-by: Li Wei <liwei@whamcloud.com>
    Reviewed-by: Chris Gearing <chris@whamcloud.com>
  20. @miketappro

    ORI-377 osd-zfs: code reorganization to save stack

    Alex Zhuravlev authored miketappro committed
    few functions from udmu have been integrated into osd-zfs methods
    
    Change-Id: I29d23ef8cb1d69b0c0990a9fa7c3dc89b23ee45a
    Signed-off-by: Alex Zhuravlev <bzzz@whamcloud.com>
    Reviewed-on: http://review.whamcloud.com/1713
    Reviewed-by: Johann Lombardi <johann@whamcloud.com>
    Tested-by: Johann Lombardi <johann@whamcloud.com>
  21. @miketappro

    ORI-377 debug: helpers for RETURN/GOTO to save stack

    Alex Zhuravlev authored miketappro committed
    RETURN/GOTO prints return code 3 times, but usually RETURN/GOTO
    are disabled, so libcfs_log_return() and libcfs_log_goto() are
    introduced which accept return code once.
    
    This is supposed to save 8-16 bytes on stack in some cases.
    
    Change-Id: I6e8571585ffb11baea741040278c658b30914232
    Signed-off-by: Alex Zhuravlev <bzzz@whamcloud.com>
    Reviewed-on: http://review.whamcloud.com/1712
    Reviewed-by: Johann Lombardi <johann@whamcloud.com>
    Tested-by: Johann Lombardi <johann@whamcloud.com>
  22. @miketappro

    ORI-377 debug: DEBUG_CAPA() to call _debug_capa() helper

    Alex Zhuravlev authored miketappro committed
    this way we save few bytes on stack in few functions as the
    functions themself don't need to reserve space for all
    arguments to CDEBUG()
    
    Change-Id: Ic6df711cfc8e96cc733029d1a952b0411f3e63c3
    Signed-off-by: Alex Zhuravlev <bzzz@whamcloud.com>
    Reviewed-on: http://review.whamcloud.com/1711
    Reviewed-by: Johann Lombardi <johann@whamcloud.com>
    Tested-by: Johann Lombardi <johann@whamcloud.com>
  23. @miketappro

    ORI-377 debug: reduce stack usage in logging infrastructure

    Alex Zhuravlev authored miketappro committed
    The purpose of this patch is to help the debug logging infrastructure
    to save stack usage. This is done through the following changes:
     - libcfs_debug_vmsg2() to accept libcfs_debug_msg_data structure
       to replace SUBSYSTEM, __FILE__, __FUNCTION__, __LINE__ and cdls
       on the stack
     - libcfs_debug_vmsg1() for regular users like CDEBUG() which don't
       need 2 format strings
     - CDEBUG, RETURN, EXIT, GOTO, CHECK_STACK to init
       libcfs_debug_msg_data
    
    This patch (along with the other ones submitted in ORI-377) reduces
    stack consumption by 1328 bytes.
    
    Change-Id: Id663188f18adbe5059010929fc0b40927e714771
    Signed-off-by: Alex Zhuravlev <bzzz@whamcloud.com>
    Signed-off-by: Johann Lombardi <johann@whamcloud.com>
    Reviewed-on: http://review.whamcloud.com/1709
  24. @miketappro

    ORI-448 tests: Allocate sequences on MOUNT2 clients in replay_barrier()

    Li Wei authored miketappro committed
    In order to avoid sequence file updates after a target is made
    read-only, replay_barrier() creates a file in "MOUNT" on every client
    node.  This is not enough because the "MOUNT2" clients may also trigger
    sequence file updates.  This patch makes sure the trick is applied to
    "MOUNT2" clients as well.
    
    Change-Id: I832689a2d2ca205d927bc5d0a15ab14fceb3bf80
    Signed-off-by: Li Wei <liwei@whamcloud.com>
    Reviewed-by: Mikhail Pershin <tappro@whamcloud.com>
    Reviewed-on: http://review.whamcloud.com/1939
  25. @miketappro

    ORI-471 osd-zfs: don't print grant-related message for MDT

    Johann Lombardi authored miketappro committed
    Messages due to bad overhead estimate related to grant reservation
    should only be printed for OST, not for MDT.
    
    Signed-off-by: Johann Lombardi <johann@whamcloud.com>
    Change-Id: I703290cdd8ee742b2e571653da3971d12892496a
    Reviewed-on: http://review.whamcloud.com/1927
  26. @miketappro

    ORI-391 test: skip quota tests

    Niu Yawei authored miketappro committed
    Skip quota tests when e2fsprogs doesn't support quota.
    
    Signed-off-by: Niu Yawei <niu@whamcloud.com>
    Change-Id: Ia0887df894387a0c99933231a52e92cfa76147d5
    Reviewed-on: http://review.whamcloud.com/1937
    Reviewed-by: Li Wei <liwei@whamcloud.com>
    Tested-by: Hudson
    Reviewed-by: Johann Lombardi <johann@whamcloud.com>
  27. @miketappro

    LU-354 Change dev_set_rdonly() check to warning

    Andreas Dilger authored miketappro committed
    The dev_set_rdonly() symbol is only required to fence I/O
    from the underlying device for testing.  This patch is not
    strictly required to build a lustre release.
    
    In addition, this change updates the dt_ro callback to
    potentially return an error code.
    
    Change-Id: I35b0a120087845e2a57c9df6c3e1de5b9e4b6ff7
    Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
    Reviewed-on: http://review.whamcloud.com/1476
    Tested-by: Hudson
    Tested-by: Maloo <whamcloud.maloo@gmail.com>
    Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
  28. @miketappro

    ORI-451 osd: finish site initialization

    Alex Zhuravlev authored miketappro committed
    this is how the site gets onto the global list of active sites.
    then shrinker for lu_object can scan them to purge objects from
    the cache.
    
    Change-Id: Id6090bf1cf5e2e91ece6e2c424e92338a777b753
    Signed-off-by: Alex Zhuravlev <bzzz@whamcloud.com>
    Reviewed-on: http://review.whamcloud.com/1920
    Reviewed-by: Johann Lombardi <johann@whamcloud.com>
    Tested-by: Johann Lombardi <johann@whamcloud.com>
  29. @miketappro

    ORI-450 ofd: fix bug in ofd_health_check.

    miketappro authored
    OFD thread info must be initialized there with ofd_info_init().
    
    Signed-off-by: Mikhail Pershin <tappro@whamcloud.com>
    Change-Id: If4c24b6511fbc7722aed92f306dfcff14a8b77ab
    Reviewed-on: http://review.whamcloud.com/1913
    Tested-by: Hudson
    Tested-by: Maloo <whamcloud.maloo@gmail.com>
    Reviewed-by: Prakash Surya <surya1@llnl.gov>
    Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
  30. @miketappro

    ORI-367 build: Update lbuild to prepare and build lustre for zfs

    Chris Gearing authored miketappro committed
    lbuild has been updated to fetch the spl/zfs source from
    the git repository, build the source against the lustre
    kernel and then provide the appropriate information to the
    lustre build to include the spl/zfs modules appropriately
    
    The default location for the git repo is git@github.com:zfsonlinux
    this can be overridden by exporting SPLZFSGITREPO to point somewhere
    else.
    
    The results of the build at not cached and so a new ticket has been
    raised to add the functionality, this is captured in ORI-426.
    
    Change-Id: I6b65272bb9d9b7992ee01c6e0736de506e332668
    Signed-off-by: Chris Gearing <chris@whamcloud.com>
    Reviewed-on: http://review.whamcloud.com/1732
    Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
    Reviewed-by: Brian J. Murrell <brian@whamcloud.com>
    Reviewed-by: Johann Lombardi <johann@whamcloud.com>
    Tested-by: Hudson
    Tested-by: Maloo <whamcloud.maloo@gmail.com>
  31. @miketappro

    LU-474 build: document the build release versions

    Andreas Dilger authored miketappro committed
    Update ancient document to better describe the build versions
    and Git tags that are being used today.
    
    This also re-uses the Gerrit Change-Id to quiet an error that
    spuriously was committed to lustre-dev with a Change-Id that
    is already used on master.  By landing and committing this
    patch to lustre-dev it will quiet the false dependency that
    is printed for every patch on master.
    
    Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
    Change-Id: I804f10bf486745ddd3b23b89e959dfd585589ac0
    Reviewed-on: http://review.whamcloud.com/1777
    Tested-by: Hudson
  32. @miketappro

    ORI-227 tests: fix lfsck.sh for OFD/OSD

    Andreas Dilger authored miketappro committed
    Change the /proc paths that lfsck.sh to use the new OFD/OSD paths
    on the orion branch.  Otherwise, lfsck.sh aborts early because it
    cannot find the OST filesystem.
    
    Disable lfsck.sh for non-ldiskfs filesystems, since it won't work
    there until we have proper in-kernel lfsck support.
    
    Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
    Change-Id: I085fa489816261857fb843d0dea3ee5e790630ee
    Reviewed-on: http://review.whamcloud.com/1755
    Tested-by: Hudson
  33. @miketappro

    ORI-297 build: use unique orion build from master

    Andreas Dilger authored miketappro committed
    Make the build version for the orion branch different from that used
    on master, to avoid confusion with the built and installed RPMs.
    
    This is starting the orion development version at 2.2.49.50 (i.e. pre-
    pre-2.3) so that when we start the actual 2.3 development versions at
    2.2.50 the orion builds will not conflict with those versions.
    
    Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
    Change-Id: Ic3f3c63fe32935d407c15e540ac4f7ca1a594ea7
    Reviewed-on: http://review.whamcloud.com/1721
    Tested-by: Hudson
Something went wrong with that request. Please try again.