1315ee00fSMartin Matuska#!/bin/sh 22ad756a6SMartin Matuska# shellcheck disable=SC3014,SC2154,SC2086,SC2034 3315ee00fSMartin Matuska# 4315ee00fSMartin Matuska# Turn off disk's enclosure slot if it becomes FAULTED. 5315ee00fSMartin Matuska# 6315ee00fSMartin Matuska# Bad SCSI disks can often "disappear and reappear" causing all sorts of chaos 7315ee00fSMartin Matuska# as they flip between FAULTED and ONLINE. If 8*5fb307d2SMartin Matuska# ZED_POWER_OFF_ENCLOSURE_SLOT_ON_FAULT is set in zed.rc, and the disk gets 9315ee00fSMartin Matuska# FAULTED, then power down the slot via sysfs: 10315ee00fSMartin Matuska# 11315ee00fSMartin Matuska# /sys/class/enclosure/<enclosure>/<slot>/power_status 12315ee00fSMartin Matuska# 13315ee00fSMartin Matuska# We assume the user will be responsible for turning the slot back on again. 14315ee00fSMartin Matuska# 15315ee00fSMartin Matuska# Note that this script requires that your enclosure be supported by the 16315ee00fSMartin Matuska# Linux SCSI Enclosure services (SES) driver. The script will do nothing 17315ee00fSMartin Matuska# if you have no enclosure, or if your enclosure isn't supported. 18315ee00fSMartin Matuska# 19315ee00fSMartin Matuska# Exit codes: 20315ee00fSMartin Matuska# 0: slot successfully powered off 21315ee00fSMartin Matuska# 1: enclosure not available 22*5fb307d2SMartin Matuska# 2: ZED_POWER_OFF_ENCLOSURE_SLOT_ON_FAULT disabled 23315ee00fSMartin Matuska# 3: vdev was not FAULTED 24315ee00fSMartin Matuska# 4: The enclosure sysfs path passed from ZFS does not exist 25315ee00fSMartin Matuska# 5: Enclosure slot didn't actually turn off after we told it to 26315ee00fSMartin Matuska 27315ee00fSMartin Matuska[ -f "${ZED_ZEDLET_DIR}/zed.rc" ] && . "${ZED_ZEDLET_DIR}/zed.rc" 28315ee00fSMartin Matuska. "${ZED_ZEDLET_DIR}/zed-functions.sh" 29315ee00fSMartin Matuska 30315ee00fSMartin Matuskaif [ ! -d /sys/class/enclosure ] ; then 31315ee00fSMartin Matuska # No JBOD enclosure or NVMe slots 32315ee00fSMartin Matuska exit 1 33315ee00fSMartin Matuskafi 34315ee00fSMartin Matuska 35*5fb307d2SMartin Matuskaif [ "${ZED_POWER_OFF_ENCLOSURE_SLOT_ON_FAULT}" != "1" ] ; then 36315ee00fSMartin Matuska exit 2 37315ee00fSMartin Matuskafi 38315ee00fSMartin Matuska 39315ee00fSMartin Matuskaif [ "$ZEVENT_VDEV_STATE_STR" != "FAULTED" ] ; then 40315ee00fSMartin Matuska exit 3 41315ee00fSMartin Matuskafi 42315ee00fSMartin Matuska 43315ee00fSMartin Matuskaif [ ! -f "$ZEVENT_VDEV_ENC_SYSFS_PATH/power_status" ] ; then 44315ee00fSMartin Matuska exit 4 45315ee00fSMartin Matuskafi 46315ee00fSMartin Matuska 472ad756a6SMartin Matuska# Turn off the slot and wait for sysfs to report that the slot is off. 482ad756a6SMartin Matuska# It can take ~400ms on some enclosures and multiple retries may be needed. 492ad756a6SMartin Matuskafor i in $(seq 1 20) ; do 50315ee00fSMartin Matuska echo "off" | tee "$ZEVENT_VDEV_ENC_SYSFS_PATH/power_status" 51315ee00fSMartin Matuska 522ad756a6SMartin Matuska for j in $(seq 1 5) ; do 53315ee00fSMartin Matuska if [ "$(cat $ZEVENT_VDEV_ENC_SYSFS_PATH/power_status)" == "off" ] ; then 542ad756a6SMartin Matuska break 2 55315ee00fSMartin Matuska fi 56315ee00fSMartin Matuska sleep 0.1 57315ee00fSMartin Matuska done 582ad756a6SMartin Matuskadone 59315ee00fSMartin Matuska 60315ee00fSMartin Matuskaif [ "$(cat $ZEVENT_VDEV_ENC_SYSFS_PATH/power_status)" != "off" ] ; then 61315ee00fSMartin Matuska exit 5 62315ee00fSMartin Matuskafi 63315ee00fSMartin Matuska 64315ee00fSMartin Matuskazed_log_msg "powered down slot $ZEVENT_VDEV_ENC_SYSFS_PATH for $ZEVENT_VDEV_PATH" 65