ceph/doc/rados/operations/devices.rst

   1 Device Management
   2 =================
   3
   4 Ceph tracks which hardware storage devices (e.g., HDDs, SSDs) are consumed by
   5 which daemons, and collects health metrics about those devices in order to
   6 provide tools to predict and/or automatically respond to hardware failure.
   7
   8 Device tracking
   9 ---------------
  10
  11 You can query which storage devices are in use with::
  12
  13   ceph device ls
  14
  15 You can also list devices by daemon or by host::
  16
  17   ceph device ls-by-daemon <daemon>
  18   ceph device ls-by-host <host>
  19
  20 For any individual device, you can query information about its
  21 location and how it is being consumed with::
  22
  23   ceph device info <devid>
  24
  25
  26 Enabling monitoring
  27 -------------------
  28
  29 Ceph can also monitor health metrics associated with your device.  For
  30 example, SATA hard disks implement a standard called SMART that
  31 provides a wide range of internal metrics about the device's usage and
  32 health, like the number of hours powered on, number of power cycles,
  33 or unrecoverable read errors.  Other device types like SAS and NVMe
  34 implement a similar set of metrics (via slightly different standards).
  35 All of these can be collected by Ceph via the ``smartctl`` tool.
  36
  37 You can enable or disable health monitoring with::
  38
  39   ceph device monitoring on
  40
  41 or::
  42
  43   ceph device monitoring off
  44
  45
  46 Scraping
  47 --------
  48
  49 If monitoring is enabled, metrics will automatically be scraped at regular intervals.  That interval can be configured with::
  50
  51   ceph config set mgr mgr/devicehealth/scrape_frequency <seconds>
  52
  53 The default is to scrape once every 24 hours.
  54
  55 You can manually trigger a scrape of all devices with::
  56
  57   ceph device scrape-health-metrics
  58
  59 A single device can be scraped with::
  60
  61   ceph device scrape-health-metrics <device-id>
  62
  63 Or a single daemon's devices can be scraped with::
  64
  65   ceph device scrape-daemon-health-metrics <who>
  66
  67 The stored health metrics for a device can be retrieved (optionally
  68 for a specific timestamp) with::
  69
  70   ceph device get-health-metrics <devid> [sample-timestamp]
  71
  72 Failure prediction
  73 ------------------
  74
  75 Ceph can predict life expectancy and device failures based on the
  76 health metrics it collects.  There are three modes:
  77
  78 * *none*: disable device failure prediction.
  79 * *local*: use a pre-trained prediction model from the ceph-mgr daemon
  80 * *cloud*: share device health and performance metrics an external
  81   cloud service run by ProphetStor, using either their free service or
  82   a paid service with more accurate predictions
  83
  84 The prediction mode can be configured with::
  85
  86   ceph config set global device_failure_prediction_mode <mode>
  87
  88 Prediction normally runs in the background on a periodic basis, so it
  89 may take some time before life expectancy values are populated.  You
  90 can see the life expectancy of all devices in output from::
  91
  92   ceph device ls
  93
  94 You can also query the metadata for a specific device with::
  95
  96   ceph device info <devid>
  97
  98 You can explicitly force prediction of a device's life expectancy with::
  99
 100   ceph device predict-life-expectancy <devid>
 101
 102 If you are not using Ceph's internal device failure prediction but
 103 have some external source of information about device failures, you
 104 can inform Ceph of a device's life expectancy with::
 105
 106   ceph device set-life-expectancy <devid> <from> [<to>]
 107
 108 Life expectancies are expressed as a time interval so that
 109 uncertainty can be expressed in the form of a wide interval. The
 110 interval end can also be left unspecified.
 111
 112 Health alerts
 113 -------------
 114
 115 The ``mgr/devicehealth/warn_threshold`` controls how soon an expected
 116 device failure must be before we generate a health warning.
 117
 118 The stored life expectancy of all devices can be checked, and any
 119 appropriate health alerts generated, with::
 120
 121   ceph device check-health
 122
 123 Automatic Mitigation
 124 --------------------
 125
 126 If the ``mgr/devicehealth/self_heal`` option is enabled (it is by
 127 default), then for devices that are expected to fail soon the module
 128 will automatically migrate data away from them by marking the devices
 129 "out".
 130
 131 The ``mgr/devicehealth/mark_out_threshold`` controls how soon an
 132 expected device failure must be before we automatically mark an osd
 133 "out".