cmd/zed/zed.d/statechange-slot_off.sh

   1 #!/bin/sh
   2 # shellcheck disable=SC3014,SC2154,SC2086,SC2034
   3 #
   4 # Turn off disk's enclosure slot if it becomes FAULTED.
   5 #
   6 # Bad SCSI disks can often "disappear and reappear" causing all sorts of chaos
   7 # as they flip between FAULTED and ONLINE.  If
   8 # ZED_POWER_OFF_ENCLOUSRE_SLOT_ON_FAULT is set in zed.rc, and the disk gets
   9 # FAULTED, then power down the slot via sysfs:
  10 #
  11 # /sys/class/enclosure/<enclosure>/<slot>/power_status
  12 #
  13 # We assume the user will be responsible for turning the slot back on again.
  14 #
  15 # Note that this script requires that your enclosure be supported by the
  16 # Linux SCSI Enclosure services (SES) driver.  The script will do nothing
  17 # if you have no enclosure, or if your enclosure isn't supported.
  18 #
  19 # Exit codes:
  20 #   0: slot successfully powered off
  21 #   1: enclosure not available
  22 #   2: ZED_POWER_OFF_ENCLOUSRE_SLOT_ON_FAULT disabled
  23 #   3: vdev was not FAULTED
  24 #   4: The enclosure sysfs path passed from ZFS does not exist
  25 #   5: Enclosure slot didn't actually turn off after we told it to
  26
  27 [ -f "${ZED_ZEDLET_DIR}/zed.rc" ] && . "${ZED_ZEDLET_DIR}/zed.rc"
  28 . "${ZED_ZEDLET_DIR}/zed-functions.sh"
  29
  30 if [ ! -d /sys/class/enclosure ] ; then
  31         # No JBOD enclosure or NVMe slots
  32         exit 1
  33 fi
  34
  35 if [ "${ZED_POWER_OFF_ENCLOUSRE_SLOT_ON_FAULT}" != "1" ] ; then
  36         exit 2
  37 fi
  38
  39 if [ "$ZEVENT_VDEV_STATE_STR" != "FAULTED" ] ; then
  40         exit 3
  41 fi
  42
  43 if [ ! -f "$ZEVENT_VDEV_ENC_SYSFS_PATH/power_status" ] ; then
  44         exit 4
  45 fi
  46
  47 # Turn off the slot and wait for sysfs to report that the slot is off.
  48 # It can take ~400ms on some enclosures and multiple retries may be needed.
  49 for i in $(seq 1 20) ; do
  50         echo "off" | tee "$ZEVENT_VDEV_ENC_SYSFS_PATH/power_status"
  51
  52         for j in $(seq 1 5) ; do
  53                 if [ "$(cat $ZEVENT_VDEV_ENC_SYSFS_PATH/power_status)" == "off" ] ; then
  54                         break 2
  55                 fi
  56                 sleep 0.1
  57         done
  58 done
  59
  60 if [ "$(cat $ZEVENT_VDEV_ENC_SYSFS_PATH/power_status)" != "off" ] ; then
  61         exit 5
  62 fi
  63
  64 zed_log_msg "powered down slot $ZEVENT_VDEV_ENC_SYSFS_PATH for $ZEVENT_VDEV_PATH"