Physical Disks/Devices

From OSNEXUS Online Documentation Site
(Redirected from Physical Disk Scan)
Jump to navigation Jump to search


The Physical Disks section presents every block device QuantaStor has discovered on each system in the grid -- SAS, SATA, NVMe and SSD drives, hardware RAID logical drives, and SAN LUNs attached over Fibre Channel or iSCSI -- together with their health, capacity, encryption capability and current consumer. This page covers the disk grid and its columns, and every operation you can run against a physical disk: scanning, identifying, formatting and secure-erasing, hot-spare marking, performance testing, multipath configuration, and importing disks that arrive with data already on them.

Section Purpose
The Physical Disks section Read the disk grid, its columns, and the per-disk dashboard.
Disk health, SMART and temperature Understand the Status and SMART State columns and what drives them.
Scanning for physical disks Discover disks that were hot-plugged or newly presented from a SAN.
Identifying a disk in the enclosure Light the drive LED so you can find the disk in the chassis.
Formatting disks Clear partitions, labels and encryption settings so a disk can be reused.
Secure erase (Shred Data) modes Choose a data-sanitization procedure for decommissioning.
Universal hot-spares Reserve a disk as an automatic replacement for any pool on the system.
Disk performance testing Measure sequential read throughput per disk to find outliers.
Multipath configuration Enable multi-path access for a SAN vendor/product.
Creating pass-thru Storage Volumes Present a raw disk to a host without a Storage Pool.
Copying LUNs and disks from 3rd-party SANs Migrate a block device from another SAN onto QuantaStor.
Importing disks from an Open-ZFS pool configuration Adopt disks that already carry an OpenZFS pool.
Self-encrypting drives (SED) Read SED capability and lock state, and recover a locked drive.
Selecting disks from the CLI Use the disk-list filter syntax instead of naming devices one at a time.

The Physical Disks section

The Physical Disks section of the Storage Management tab. The tree lists every discovered device per system; the grid below shows the disks for the system currently selected.
Navigation: Storage Management → Physical Disks (tree section)

Expand Physical Disks in the left-hand tree to see one node per storage system in the grid, each listing its devices. Select a system or a disk and the grid at the bottom of the center pane lists the disks for that system; the panel header names the system, as in Physical Disks (qs-node-110). Selecting a single disk switches the dashboard above the grid to a Physical Disk Dashboard with IOPS, throughput and latency charts for that device.

Linux device names such as sda, sdb and sdc come from the kernel: the sd prefix means SCSI disk and the trailing letters identify the disk within the system. Those names are not stable across reboots or recabling, so QuantaStor labels each disk with both the kernel name and its persistent by-id path -- for example sdc (scsi-36000c29b6e80f9647dc1b672470094a5) -- and always addresses the device internally by the persistent path under /dev/disk/by-id/. Use the serial number or SCSI ID, not the kernel name, when you need to be certain which drive you are acting on.

Two categories of device appear in the tree but are deliberately omitted from the grid, because neither is available for provisioning:

  • the boot/system disk, and any other disk carrying a mounted filesystem or swap;
  • the individual member paths of a multi-path device -- only the multipath parent is listed.

Disk grid columns

These columns are shown by default.

Column Meaning
Name Kernel device name plus the persistent by-id device name.
Size Raw capacity of the device.
Block Size Logical block size reported by the device, typically 512B or 4KB.
Status Object state. The values you normally see on a disk are Normal, Warning, Error, Missing and Offline, set from the disk's SMART assessment where one is available.
Used By What currently consumes the disk. A Storage Pool, Ceph OSD, Ceph Journal Group or pass-thru Storage Volume is named directly. Bracketed values are states rather than objects: [Available] (free for use), [Contains Partitions] (unused but not blank -- format it before use), [Hot-Spare], [Journal Device (WAL/MDB)], [RESERVED] (an I/O fencing reservation from outside this grid), or [<pool name>] in brackets when the disk is I/O fenced for a pool it is not currently a member of.
Correlated Device The upstream object this disk is a view of: a Storage Volume on another QuantaStor system, a hardware disk or RAID unit behind a controller, or an External System Media device from an NVMe-oF system. N/A means the device is directly attached and has no upstream correlation. Disks with no correlation cannot have their enclosure LED lit -- see Identifying a disk in the enclosure.
SED Capable Yes when the drive reports both lock and secure-erase capability. See Self-encrypting drives (SED).
SED Status Blank when the drive is not locked, Lock Enabled when locking is armed, Post-Power-Off LOCKED state when the drive is locked, or both.
SSD Life Left Remaining endurance as a percentage, for solid-state and NVMe devices. N/A on rotating media and on any SSD that does not report an endurance indicator.
SMART State The health string collected by the SMART manager. See Disk health, SMART and temperature.
Type Bus and media type: SATA, SCSI, SAS, NVME, SSD, SSHD, PMEM, AoE, Multiple or Other. SATA, SCSI, SAS and AoE gain an -SSD suffix when the device reports itself non-rotational, so a SAS SSD reads SAS-SSD.
Vendor SCSI vendor identifier.
Product ID SCSI product identifier.
Serial Number Device serial number. This is the value to record when returning a drive under warranty.

Click the arrow at the right of any column header for Sort Ascending, Sort Descending, a Columns submenu that shows and hides columns, and -- on the filterable columns -- a Filters submenu. These columns are hidden by default and worth turning on: Storage System, Drive Temperature, Read Errors, Write Errors, Non-Medium Errors, Is SSD?, Compression, SCSI ID, RAID, RAID Segment, Pool Device, OSD Name, Is Journal Device?, Universal Spare?, Source System, Source Volume, I/O Fenced to System, I/O Fenced to Pool, I/O Fencing Key, Last Shred, Last Shred Mode, Shred Signature, Read Seq and Last Performance Test. The Name, Type, Vendor, Product ID, Serial Number, Correlated Device and Is SSD? columns are the filterable ones. Columns can also be dragged into a different order.

Turn on Storage System when you are working across a grid: the grid pane is otherwise scoped to whichever system is selected, and the header tells you which one that is.

The equivalent CLI commands are qs disk-list for the full inventory, qs disk-get --disk=<name-or-id> for one device in detail, and qs disk-search --search-filter=<key:value> for a subset. The search filter matches object field names and accepts the modifiers = (equal), ~ (not), > (greater than), < (less than) and * (like): 

qs disk-get --disk=sdc
qs disk-search --search-filter="vendorId:VMware"
qs disk-search --search-filter="*name:sdc"

Physical disk actions

Right-clicking a disk in the grid lists every action available for that device. Performance testing and hot-spare marking are only on this menu, not on the toolbar.

Select a disk and the Physical Disk toolbar group offers Identify, Format, Create Pass-thru, Scan, Multipath Config and Copy. Right-clicking the disk -- in the grid or in the tree -- adds three operations that have no toolbar button: Create Storage Pool..., Disk Performance Test... and Mark As Hot-Spare... (or Unmark as Hot-Spare... when the disk is already a spare). Properties... opens the full object detail for the device.

The context menu is sensitive to what the disk is doing. Mark As Hot-Spare... is hidden when the disk is already a spare, is a member of a Storage Pool, or carries a mounted filesystem, and Unmark as Hot-Spare... appears only for a disk that is already marked. A disk serving as a Ceph journal device offers neither Create Storage Pool... nor Create Pass-thru Storage Volume(s)....

Disk health, SMART and temperature

QuantaStor polls SMART data with smartctl and sg_logs, and with nvme smart-log on NVMe devices, and folds the result into the SMART State column and the disk's Status. The first poll for a given serial number is a full collection; later polls are a lighter health-and-log-page check, so the error counters below reflect the full collection rather than the most recent pass. The SMART manager runs every two hours; a newly discovered disk is queued for collection as soon as it is found rather than waiting for the next pass. To force a collection immediately, run qs system-signal-manager --manager-class=smart.

SMART State is not a simple pass/fail flag. Where the drive reports a health line QuantaStor stores it verbatim -- usually OK -- but when a collected counter crosses a threshold the column is replaced with a sentence naming the problem, and the disk's Status moves to Warning or Error. The thresholds live in /opt/osnexus/quantastor/conf/qs_smartmanager.conf and ship as:

Setting Default Effect when exceeded
err_count_threshold_uncorrected_reads 1 Warning: uncorrected read errors detected.
err_count_threshold_uncorrected_writes 1 Warning: uncorrected write errors detected.
err_count_threshold_non_medium 1000 Warning naming the non-medium error count. Non-medium errors point at cabling, expander or connectivity trouble rather than the media itself, which is why the message says so explicitly.
err_count_threshold_media_defects 1000 Warning: grown defect list is large.
err_count_threshold_ssd_life_left_pct 15 Warning: remaining SSD endurance is at or below this percentage.

A health alert for a non-optimal disk is raised at most once a week per disk, so a marginal drive does not flood the alert log.

Two limits are worth knowing before you read the column as gospel:

  • Plain SATA/ATA disks commonly show [N/A]. QuantaStor derives the health string from the SCSI SMART Health Status line; the ATA equivalent (SMART overall-health self-assessment test result: PASSED) sets the disk's Status but leaves the string empty, which renders as [N/A]. An [N/A] SMART State on a SATA drive is not a fault.
  • SMART collection is skipped for some devices by design, including device-mapper paths and drives whose vendor string is VMware or OSNEXUS, so virtual disks in a lab or evaluation deployment never populate the column.

Drive temperature is collected in the same pass and is shown in the hidden Drive Temperature column in both Celsius and Fahrenheit. Over-temperature alerting is driven from the enclosure/controller side rather than from this column, with a 60 C default threshold; see Hardware Controllers & Enclosures.

Scanning for physical disks

Scan for Physical Disks. Select one or more systems; leave Rescan Multipath Configuration checked unless you are only looking for a directly attached disk.
Navigation: Storage Management → Physical Disks → select a Physical Disk → Physical Disk → Scan (toolbar)

Use Scan after adding a disk to the chassis, after presenting a new LUN from a SAN, or after combining disks into a new unit on a hardware RAID controller. The dialog has two controls:

  • Select Storage System(s) -- a multi-select grid of the systems in the grid. Opening the dialog from a disk preselects that disk's system; opening it with no disk selected selects every system. Only systems in a Normal or Warning state are listed, since an unreachable system cannot be scanned. One scan is issued per selected system.
  • Rescan Multipath Configuration -- checked by default. It does three extra things: it re-applies the system's multipath auto-configuration if that is enabled, it flushes unused multipath maps and restarts multipathd, and it removes stale disk records whose device has gone away. Clear it when you only want a plain device rediscovery and do not want multipathd restarted.

The scan itself rescans iSCSI sessions, writes a scan request to every SCSI host under /sys/class/scsi_host/, re-triggers udev, and then reconciles what the kernel reports against the QuantaStor database. The SCSI bus rescan is governed by device_rescan_support in the [device] section of /etc/quantastor.conf, which ships enabled.

You rarely need to scan manually. QuantaStor installs udev rules that call back into the service whenever a SCSI, SATA or NVMe block device is added or removed, so hot-plugged drives normally appear on their own within seconds. Repeated events for the same device are throttled: once three scans have run for one device inside a minute, further events for it are suppressed, so a flapping device does not spin the service. Independently of that, the disk manager reconciles its inventory once an hour. Neither the udev path nor the hourly pass rescans SCSI buses the way a manual scan does, which is why a LUN newly presented from a SAN, and a new hardware RAID unit, still call for an explicit scan.

Disks behind a hardware RAID controller need the controller rescanned first, and the new disks combined into a RAID unit, before they appear here; see Hardware Controllers & Enclosures.

The CLI equivalent is qs disk-scan [--storage-system=<system>], which performs the plain scan; the extra multipath work that the Rescan Multipath Configuration checkbox triggers has no documented CLI argument, so run that from the web interface. 

qs disk-scan
qs disk-scan --storage-system=qs-node-110

Identifying a disk in the enclosure

Identify Physical Disk. The duration slider accepts 10 to 2000 seconds.
Navigation: Storage Management → Physical Disks → select a Physical Disk → Physical Disk → Identify (toolbar)

Identifying a disk lights or flashes its drive LED so you can find it in the rack before you pull it. The dialog has three fields:

  • Physical Disk -- the device to identify, preselected from your grid or tree selection.
  • Storage System -- read-only, showing which system owns the disk.
  • Duration -- how long to keep the beacon active, as a number field and a slider. Valid range is 10 to 2000 seconds.

How the LED is driven depends on what the disk is. QuantaStor picks one of three mechanisms, and the choice matters because they are not equally capable:

  1. A hardware RAID logical drive (the disk maps to a RAID unit) is identified through the controller's own unit-identify operation, which lights every member drive of the unit.
  2. A disk correlated to an enclosure slot is identified through SCSI Enclosure Services -- sg_ses --set=ident against the enclosure's SES device for that slot -- falling back to the RAID or HBA vendor utility where SES is unavailable. This is the real locate LED, and on a multi-path enclosure the request is sent down every path to the slot so the beacon lights regardless of which expander answers. NVMe drives use ledctl, or the Supermicro management utility on Supermicro chassis.
  3. A disk with no enclosure or controller correlation has no addressable LED, so QuantaStor pulses the drive's activity light instead, by issuing repeated 32 KB direct reads from the start of the device in an on/off pattern. This is the fallback path, and it is the only one where the Correlated Device column reads N/A.

That third case is why a disk can be identified and yet appear not to blink: an activity-light pulse is much harder to see than a locate LED, and it does nothing on a device that has no visible activity indicator. NVMe devices never use the activity pulse.

Several conditions stop identification outright, and all of them report a clear error rather than failing silently:

  • Virtual machines. Automatic fault beaconing is disabled entirely on a VM, since there is no hardware to light.
  • Service startup. The Disk Blink Manager refuses identify requests until its first hardware discovery pass has completed. Retry after a minute.
  • Recent pool failover. LED checks are skipped for a period after a Storage Pool failover so that beacon traffic does not compete with recovery.
  • Unsupported adapters. Some controller families expose no identify operation at all and report This operation is not supported with this vendor/model of hardware. An HPE H241 in HBA mode reports a specific not supported in HBA mode alert.
  • Missing drives cannot be lit as drives; QuantaStor lights the empty enclosure slot instead, which is what you want when you are looking for the bay to populate.

QuantaStor also lights drive LEDs on its own for disks in a Warning, Error or Offline state, re-checking every 120 seconds. That fault beaconing only works on SAS HBA and LSI storcli class adapters; on other controller families the LED is lit only when you ask for it explicitly. An indefinite manual beacon on the same disk is throttled to three requests per 24 hours -- specify a duration instead if you need to re-light a drive repeatedly. Beacon state is not persisted, so a service restart clears every manual beacon.

Automatic fault beaconing can be switched off per system without affecting manual identify, with qs system-modify --storage-system=<system> --service-option=disable-disk-beacon-management; prefix the value with a tilde to remove it from the block list again. Alternatively set diskblink_status_check_interval under a [diskblink_manager] section of /etc/quantastor.conf -- a value of 0 disables the manager at service start, and any value below 20 seconds is raised to 20. There is no web interface control for either.

The CLI equivalent is qs disk-identify --disk=<name-or-id> [--duration=<seconds>] [--pattern=<pattern>] [--blink-type=on|off]. The pattern argument applies only to the activity-pulse fallback and is built from four characters: p a 400 ms pulse, P an 800 ms pulse, d a 400 ms pause and D an 800 ms pause. The pattern repeats until the duration expires; the default is ppD, and the web interface always uses that default. A duration of 0 on an LED-capable disk leaves the beacon on until you turn it off with --blink-type=off; on the activity-pulse path a duration of 0 means 30 seconds. An identify task can be cancelled from the Tasks pane. 

qs disk-identify --disk=sdc --duration=60
qs disk-identify --disk=sdc --pattern=pppD --duration=120
qs disk-identify --disk=sdc --blink-type=off

For hardware disks, RAID units and enclosure slots that are not presented as physical disks, use the separate identify operations in the Controllers & Enclosures section, documented in Hardware Controllers & Enclosures.

Formatting disks

Format Physical Disk. The counters beside the search box report how many disks are available, how many match the filter, and how many are selected.
Navigation: Storage Management → Physical Disks → select a Physical Disk → Physical Disk → Format (toolbar)

Disks that arrive with partitions, filesystem signatures, an old ZFS label, an MD RAID superblock or LUKS encryption metadata on them are refused for pool creation and other provisioning operations. Formatting clears all of that so the disk reads as blank.

Formatting destroys all data on the selected disks. The dialog asks for confirmation before it proceeds.

The dialog has these controls:

  • Storage System -- which system's disks to list. Format is a single-system operation; every selected disk must belong to the same system.
  • Select the Disks to Format -- a multi-select grid. Alongside the standard columns it shows Last Shred, Correlated Device and Serial Number, so you can confirm you have the right physical drive and see whether it has been sanitized before. Type into the box beside Search to narrow the list and click Reset to clear the filter and show all disks. The Total, Found and Selected counters beside it report how many disks are available, how many match the current filter, and how many you have checked -- read Selected before clicking OK.
  • Security Options → Shred Data -- unchecked by default. Checking it enables the mode list beside it; see Secure erase (Shred Data) modes. Left unchecked, the disk is quick-formatted only.
  • Security Options → Release I/O Fence -- unchecked by default. Releases the SCSI-3 persistent reservations on the disk. A device that is I/O fenced has been reserved for write access by one host; every other host has read access only, so a format cannot proceed until the reservation is released. QuantaStor uses fencing to protect pool members in a high-availability configuration, so a disk that was pulled from an HA pool commonly still carries a reservation. Without this option a fenced disk is refused.
  • Force -- unchecked by default. Required to release a reservation that still belongs to a storage system and pool QuantaStor knows about, and it skips the pre-flight sweep for disks in use as Ceph OSDs. Its tooltip reads The force option is required when one or more disks are in use.

A quick format removes device-mapper dependencies, deletes partitions 1 through 12 and re-reads the partition table, zeroes the MD RAID superblock, writes zeros over the first 32 MB of the device, and clears any ZFS label. On an encrypted device it also closes the LUKS mapping and removes the device's entries from /etc/crypttab, which is what the dialog means by clears any encryption settings. On an NVMe device the equivalent step is an nvme format of the namespace. Afterwards QuantaStor refreshes the multipath configuration, rescans for disks, and clears the disk's pool, OSD, journal and pass-thru associations in the database.

Formatting is refused outright, with no force override, when the disk is:

  • an underlying path of a multi-path parent device -- format the parent instead;
  • carrying a mounted filesystem, which is how the boot/system disk is protected;
  • a member of a Storage Pool;
  • a member of a Ceph Journal Group;
  • in use as a Ceph OSD;
  • in use as a pass-thru Storage Volume.

Remove the disk from its consumer first -- delete the pool, remove the OSD, delete the pass-thru volume -- and then format it.

The CLI equivalent is qs disk-format --disk-list=<disks> [--shred-mode=<mode>] [--release-io-fence=true] [--flags=force].

--disk-list accepts filter tokens, and a filter can match far more disks than you intended. There is no preview mode on qs disk-format itself, so resolve the filter first with qs disk-list --filter=<same tokens>, which takes the same syntax and only reads. See Selecting disks from the CLI. 

qs disk-list --filter="[local],[unused]"
qs disk-format --disk-list=sdc
qs disk-format --disk-list=sdc,sdd,sde --shred-mode=dod

Secure erase (Shred Data) modes

Checking Shred Data in the Format dialog runs a data-sanitization procedure after the quick format. Shredding is a long-running operation -- the dialog's tooltip warns it can take hours to complete -- and its progress is reported per disk in the Tasks pane. Up to 60 disks are shredded concurrently.

The overwrite procedures are executed by the /usr/bin/scrub utility, one scrub process per disk and up to 60 disks at a time. The drive-native procedures bypass overwriting entirely and instead ask the drive to discard or replace its own encryption key, which is why they finish in seconds rather than hours.

Mode as shown in the dialog CLI value Pattern What it does
Auto auto   Picks a procedure per disk from what the drive supports. The default selection.
4-pass DoD 5220.22-M section 8-306 procedure dod dod Overwrite passes per the DoD 5220.22-M sequence.
4-pass NNSA Policy Letter NAP-14.1-C procedure nnsa nnsa Overwrite passes per the NNSA NAP-14.1-C sequence.
US Army AR380-19 method usarmy usarmy Overwrite passes per the US Army AR380-19 sequence.
Fill-zeros single-pass procedure zeros fillzero A single pass of zeros. The fastest overwrite, and enough where the requirement is only that the previous data not be readable.
Random-data single-pass procedure random random A single pass of random data.
SED Instant Erase sed-instant-erase   Cryptographic erase on a self-encrypting drive: the drive discards its media encryption key, rendering all data unrecoverable without writing to the media. Requires the drive's PSID -- see Self-encrypting drives (SED).
NVME Secure nvme_secure   Namespace format on an NVMe device.
SSD Secure ssd_secure   Reserved for non-SED, non-NVMe SSDs. Not implemented in QuantaStor 6.9.0 -- selecting it, or letting Auto select it, fails with Quick secure formatting of non-SED non-nvme SSD devices not supported.

(Note: the pass counts in the mode names are the dialog's own labels and do not all match the shipped utility. The scrub version installed on QuantaStor 6.9.0 reports the DoD, NNSA and US Army patterns as 3-pass sequences. Run scrub --help on the appliance for the authoritative pattern list and pass counts.)

Auto resolves in this order, first match winning: a self-encrypting drive gets SED Instant Erase; an NVMe device gets NVME Secure; any other solid-state device gets SSD Secure; a rotating disk gets Random-data single-pass procedure. Because SSD Secure is not implemented, Auto fails on a plain SATA or SAS SSD that is not self-encrypting. Choose Fill-zeros or Random-data explicitly for those drives.

A mode other than Auto requires a homogeneous selection: QuantaStor sorts the selected disks into rotating, SSD, NVMe and SED groups and rejects the operation if they do not all fall in the same group. Select one media type at a time, or use Auto.

After a successful overwrite QuantaStor records the completion timestamp, the mode used, and a shred signature on the disk record; these appear in the hidden Last Shred, Last Shred Mode and Shred Signature columns and are preserved across rediscovery, so they survive a rescan and a reboot. Completion is also written to the audit log at /var/log/qs/qs_audit.log as a verification record naming the device and its final progress -- that log entry is the artifact to keep for a decommissioning audit, since QuantaStor does not generate a separate erasure certificate. Note that SED Instant Erase and NVME Secure complete through a different code path and do not populate the Last Shred columns.

Universal hot-spares

Mark Universal Storage Pool Hot-Spare. The information panel confirms the vendor, product, serial number and size of the disk you selected.
Navigation: Storage Management → Physical Disks → Physical Disk (select + right-click) → Mark As Hot-Spare

A universal hot-spare is a disk reserved on a system so that any Storage Pool on that system can consume it automatically when a pool member fails. It is not tied to one pool, which is the difference between it and a pool-specific spare added from the Storage Pools section.

The dialog has a Storage System selector, a Physical Disk selector, and a read-only Physical Disk Information panel showing the selected disk's vendor ID, product ID, serial number, size and owning system. Check those against the drive you intend to reserve before clicking OK -- the disk is taken out of the available pool of disks. Unmark as Hot-Spare... on the same menu reverses it.

Only a disk that is not already a spare, not a member of a Storage Pool and not carrying a mounted filesystem can be marked, which is why the menu item disappears for disks that fail those tests. The hidden Universal Spare? column reports which disks are marked.

The CLI equivalents are qs disk-global-spare-add --disk-list=<disks>, qs disk-global-spare-remove --disk-list=<disks> and qs disk-spare-list.

Disk performance testing

Physical Disk Performance Test. Results land in the Read Seq and Last Performance Test columns of the grid within the dialog.
Navigation: Storage Management → Physical Disks → Physical Disk (select + right-click) → Disk Performance Test

The performance test reads sequentially from each selected disk and reports average throughput, which is the quickest way to find a drive that is markedly slower than its peers before you build a pool on it. It is a read-only test and does not modify the disks.

  • Mode -- how the selected disks are scheduled. Serial (the default) tests one disk at a time, which gives each drive the whole bus and produces the cleanest per-drive number. Parallel tests all of them at once in batches, which measures aggregate capability and will expose an HBA or expander bottleneck. Group by VDEV tests each pool VDEV's disks in parallel, one VDEV at a time; Group by Enclosure does the same per enclosure. The grouped modes exist to isolate where a bottleneck sits -- a VDEV or an enclosure that is slow as a group when its individual drives are fast points at the shared path, not the media.
  • Block Size -- the sequential read size, from 512 bytes to 1 GB, default 1 MB. Larger blocks give a better estimate for backup and archive workloads.
  • Block Count -- number of blocks to read, 1 to 1,000,000, default 1000. Block size multiplied by block count is the total data read per disk, so the defaults read 1 GB from each drive.
  • Select the Disks for Performance Testing -- the same multi-select grid, search box and Reset button as the Format dialog. Disks in use are selectable here, since the test only reads.

Click Run Test to start; results appear in the Read Seq and Last Performance Test columns as each disk finishes. Close dismisses the dialog. The same columns are available in the main disk grid, so a set of results can be compared later without re-running the test.

The CLI equivalent is qs disk-performance-test --disk-list=<disks> [--block-size=<size>] [--block-count=<n>] [--perf-mode=serial|parallel|group-by-vdev|group-by-enclosure]: 

qs disk-performance-test --disk-list=sdc,sdd --block-size=1M --block-count=1000

Multipath configuration

Multi-path Device Configurator, listing the SCSI vendor and product identifiers configured for multi-path access.
Navigation: Storage Management → Physical Disks → select a Physical Disk → Physical Disk → Multipath Config (toolbar)

The Multi-path Device Configurator controls which SAS, Fibre Channel and iSCSI devices QuantaStor presents as multi-path devices, keyed by SCSI vendor and product identifier. QuantaStor ships entries for the common SAN vendors -- 3PARdata, DELL, DGC, DataCore, EMC, HITACHI, HP and others -- and rewrites the blacklist_exceptions block of /etc/multipath.conf from this list.

The dialog offers a Storage System selector, the SCSI Device Multipath Vendor/Product Identifiers grid, a combo box with Add to add a vendor/product tag, Scan to probe attached devices for multipath-capable vendor/product pairs, Remove Selected to delete checked entries, and Clear Changes to discard edits before you commit them with OK. A tag needs both a vendor and a product value. Either field may end in a single * wildcard -- HITACHI:OPEN-.* and DGC:.* are shipped entries -- but a wildcard has to be the last character, only one is allowed per field, and the vendor may not be a bare *.

Multi-path setup as a whole -- cabling, path policies, and how QuantaStor presents the parent device -- is covered in Multipath Configuration. The CLI equivalents are qs disk-multipath-config-list, qs disk-multipath-config-scan, qs disk-multipath-config-add --device-tag-list=<vendor:product> and qs disk-multipath-config-remove --device-tag-list=<vendor:product>.

Remember that the individual paths of a multi-path device are hidden from the disk grid; only the parent is listed, and provisioning and formatting operate on the parent.

Creating pass-thru Storage Volumes

Navigation: Storage Management → Physical Disks → select a Physical Disk → Physical Disk → Create Pass-thru (toolbar)

Create Pass-thru presents a raw physical disk to a host as a Storage Volume without putting a Storage Pool in front of it. Use it where the client wants the drive itself -- a database or application that manages its own redundancy, or a benchmark that must not see a pool's caching. The disk's Used By column then names the pass-thru volume, and the disk cannot be formatted until that volume is deleted.

Copying LUNs and disks from 3rd-party SANs

QuantaStor can copy a Fibre Channel or iSCSI attached block device from a third-party SAN directly onto a QuantaStor system, which is the usual way to migrate off an array that is being retired. Select the disk and use Copy from the Physical Disk toolbar group.

Navigation: Storage Management → Physical Disks → select a Physical Disk → Physical Disk → Copy (toolbar)

The source can be a physical disk or a Storage Volume, and the target can be a physical disk, a Storage Volume in a pool, or a file in a Network Share -- writing to a share lets you capture the LUN as a raw, qcow, VHD or VMDK image rather than restoring it onto a device immediately. The full procedure, including how to attach the third-party LUN in the first place, is documented in Physical Disk Data Migration.

The CLI equivalent is qs disk-copy --source-id=<id> --target-id=<id> --source-type=disk|volume --target-type=disk|share|volume.

Importing disks from an Open-ZFS pool configuration

Disks that already carry an OpenZFS pool created outside QuantaStor -- on FreeBSD, illumos, or plain Linux ZFS -- are imported as a pool rather than formatted. Do not format them: formatting clears the ZFS label and the pool with it.

Import the pool from the Storage Pools section, and use the console-level procedure in Console Level OpenZFS Storage Pool Importing for pools that the web interface will not adopt directly.

Self-encrypting drives (SED)

The SED Capable column reads Yes when a drive reports both locking and secure-erase capability over the TCG Opal, Opal 2, Ruby or Enterprise interfaces; QuantaStor detects this with sedutil-cli and re-checks once an hour. SED Status is blank on an unlocked drive, Lock Enabled once locking has been armed, and Post-Power-Off LOCKED state when the drive is locked -- both appear together on a drive that is armed and locked. A locked drive is also reported as encrypted with a cipher of SED.

Instant cryptographic erase of a SED, whether selected explicitly as SED Instant Erase or chosen by Auto, requires the drive's PSID -- the Physical Secure ID printed on the drive label. QuantaStor reads PSIDs from /var/opt/osnexus/quantastor/conf/qs_physical_disks_psids.conf, one SERIAL PSID pair per line; a shred fails if the file or the drive's entry is missing.

A SED that is locked with a key QuantaStor no longer holds can be recovered only by reverting it to factory state with its PSID, which destroys all data on it. That is qs disk-sed-hard-reset --disk-list=<disks> --psid-passphrase=<psid>, a CLI-only operation with no web interface equivalent. Treat the PSID as a credential: it is the one value that unconditionally erases the drive.

Selecting disks from the CLI

Several commands -- qs disk-format, qs disk-performance-test, qs disk-global-spare-add, qs disk-sed-hard-reset and the pool creation commands -- take a --disk-list argument. Besides a comma-delimited list of disk names, IDs, system names or pool names, it accepts bracketed filter tokens, combined with AND:

Token Selects
[unused], [any], [*] Disks not used by a pool, OSD, journal or mounted filesystem. All three are synonyms.
[used] The complement: disks in use by any of those.
[local] Only disks on the system the command is issued against.
[multi] Allow disks from more than one system to resolve.
[ssd] / [hdd] / [nvme] Solid-state / rotating / NVMe devices.
[spares] / [nospares] Include only, or exclude, hot-spares.
[paths] / [nopaths] Only the member paths of multi-path devices, or only non-path devices.
[mpath:true] / [mpath:false] Disks that do, or do not, report multipath capability.
[gt:SIZE], [lt:SIZE], [eq:SIZE] Size comparisons, for example [gt:10TB]. Sizes are matched with a 200 MiB tolerance, and the suffixes are decimal -- [eq:10GB] does not match a 10 GiB drive, so write [eq:10GiB] when you mean binary units.
[vendor:VVVV], [product:PPPP], [disktype:N] Match the SCSI vendor, SCSI product, or numeric disk type.
[storageSystemId:ID] Disks on one specific storage system.
[tag:VALUE] or [tag:KEY=VALUE] Disks tagged with qs tag-set --object-id=<disk-uuid> --tag=tag --value=<VALUE>. Tagging is the reliable way to pre-select an arbitrary set of drives -- a shelf, a batch, a replacement group -- and then act on exactly that set; clear a tag by setting its value to a hyphen.
[N] Auto-select N disks from what the other filters matched, preferring the largest same-size group and dropping the smallest disks first. Fails rather than under-selecting if fewer than N match.

Three syntax rules catch people out:

  • The brackets are part of the token. --disk-list=[unused] works; --disk-list=unused is read as a disk name and fails. Quote the whole argument in a shell so the brackets are not glob-expanded.
  • Every token is a separate comma-separated list item. Write --disk-list="qs-node-110,[unused],[lt:20GB]", not "qs-node-110[unused][lt:20GB]".
  • [storageSystemId:] takes the system's UUID, not its name -- use a bare system name as a list item instead if you want to scope by name.

qs disk-list --filter=<tokens> takes the same syntax and only reads, which makes it the way to check what a filter resolves to before handing the same tokens to a destructive command. Add --normalize=true to collapse duplicate views of the same device -- child paths and the copies each grid member reports -- to one row per disk. 

qs disk-list --filter="qs-node-110,[unused],[eq:10GiB]"
qs disk-list --normalize=true
qs disk-format --disk-list="[local],[unused],[eq:10GiB]"
qs disk-performance-test --disk-list="[local],[ssd]" --perf-mode=parallel

(Note: the --disk-list help text in QuantaStor 6.9.0 documents a [all] token to select all disks. That token is not recognized -- use [*] or [any], which select unused disks.)

Related pages

Verified against QuantaStor 6.9.0.

Retrieved from "https://wiki.osnexus.com/index.php?title=Physical_Disks/Devices&oldid=27518#Physical_disk_actions"