ceph/doc/cephfs/best-practices.rst

   1
   2 CephFS best practices
   3 =====================
   4
   5 This guide provides recommendations for best results when deploying CephFS.
   6
   7 For the actual configuration guide for CephFS, please see the instructions
   8 at :doc:`/cephfs/index`.
   9
  10 Which Ceph version?
  11 -------------------
  12
  13 Use at least the Jewel (v10.2.0) release of Ceph.  This is the first
  14 release to include stable CephFS code and fsck/repair tools.  Make sure
  15 you are using the latest point release to get bug fixes.
  16
  17 Note that Ceph releases do not include a kernel, this is versioned
  18 and released separately.  See below for guidance of choosing an
  19 appropriate kernel version if you are using the kernel client
  20 for CephFS.
  21
  22 Most stable configuration
  23 -------------------------
  24
  25 Some features in CephFS are still experimental.  See
  26 :doc:`/cephfs/experimental-features` for guidance on these.
  27
  28 For the best chance of a happy healthy filesystem, use a **single active MDS**
  29 and **do not use snapshots**.  Both of these are the default.
  30
  31 Note that creating multiple MDS daemons is fine, as these will simply be
  32 used as standbys.  However, for best stability you should avoid
  33 adjusting ``max_mds`` upwards, as this would cause multiple MDS
  34 daemons to be active at once.
  35
  36 Which client?
  37 -------------
  38
  39 The FUSE client is the most accessible and the easiest to upgrade to the
  40 version of Ceph used by the storage cluster, while the kernel client will
  41 often give better performance.
  42
  43 The clients do not always provide equivalent functionality, for example
  44 the fuse client supports client-enforced quotas while the kernel client
  45 does not.
  46
  47 When encountering bugs or performance issues, it is often instructive to
  48 try using the other client, in order to find out whether the bug was
  49 client-specific or not (and then to let the developers know).
  50
  51 Which kernel version?
  52 ---------------------
  53
  54 Because the kernel client is distributed as part of the linux kernel (not
  55 as part of packaged ceph releases),
  56 you will need to consider which kernel version to use on your client nodes.
  57 Older kernels are known to include buggy ceph clients, and may not support
  58 features that more recent Ceph clusters support.
  59
  60 Remember that the "latest" kernel in a stable linux distribution is likely
  61 to be years behind the latest upstream linux kernel where Ceph development
  62 takes place (including bug fixes).
  63
  64 As a rough guide, as of Ceph 10.x (Jewel), you should be using a least a
  65 4.x kernel.  If you absolutely have to use an older kernel, you should use
  66 the fuse client instead of the kernel client.
  67
  68 This advice does not apply if you are using a linux distribution that
  69 includes CephFS support, as in this case the distributor will be responsible
  70 for backporting fixes to their stable kernel: check with your vendor.
  71
  72 Reporting issues
  73 ----------------
  74
  75 If you have identified a specific issue, please report it with as much
  76 information as possible.  Especially important information:
  77
  78 * Ceph versions installed on client and server
  79 * Whether you are using the kernel or fuse client
  80 * If you are using the kernel client, what kernel version?
  81 * How many clients are in play, doing what kind of workload?
  82 * If a system is 'stuck', is that affecting all clients or just one?
  83 * Any ceph health messages
  84 * Any backtraces in the ceph logs from crashes
  85
  86 If you are satisfied that you have found a bug, please file it on
  87 `the tracker <http://tracker.ceph.com>`_.  For more general queries please write
  88 to the `ceph-users mailing list <http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com/>`_.