eG Monitoring
 

Measures reported by EsxLunsTest

LUN is a Logical Unit Number. It can be used to refer to an entire physical disk, or a subset of a larger physical disk or disk volume. The physical disk or disk volume could be an entire single disk drive, a partition (subset) of a single disk drive, or disk volume from a RAID controller comprising multiple disk drives aggregated together for larger capacity and redundancy.

The EsxLunsTest reports critical usage statistics pertaining to every LUN on the ESX server host. In the VMware environment, a LUN is typically referred to using its HBA path, where HBA stands for the Host Bus Adapter. The HBA path on the other hand is expressed using the notation: vmhba<HBA ID:Target ID:LUN ID>. For example, if the HBA path is vmhba0:0:0, then the first 0 indicates the HBA ID, the second zero represents the Target ID, and the final 0 the LUN ID.

Measurement Description Measurement Unit Interpretation
Reads The rate at which read commands were issued Commands/Sec  
Writes The rate at which write commands were issued Commands/Sec  
Issued_commands The number of commands issued per second Commands/Sec  
Commands_aborted The number of commands aborted per second Commands/Sec  
Data_writes The rate at which data was written MB/Sec  
Data_reads The rate at which data was read MB/Sec  
Bus_resets The number of SCSI bus resets Resets/Sec The VMs use the SCSI protocol to communicate to disks, even over Fibre Channel to SAN Luns. SCSI Bus Resets are issued to release resources. These SCSI Bus Resets are in effect the SCSI subsystem timing out, commands being canceled, and retrying. This happens when the HBA device is overloaded, or its q-depth is exhausted. The first thing to know is which vmhba controller (C), target/path (T), and LUN (L) experienced these problems. The more VMs sharing a single Lun the more likely that resets will occur. A rule of thumb is no more than 10 VMs sharing a Lun.
Disk_read_latency The average amount of time taken for a read from the perspective of a guest operating system Secs This is the sum of kernel latency and physical device read latency. High latency is a cause for concern, as it is an indicator of contention for storage resources. If the value of this measure is high, then check the values reported by the Kernel disk read latency and Physical device read latency measures for this storage adapter. Doing so enables you to quickly determine why exactly the guest OS is experiencing latencies while reading from the disk - is it because of latencies in the VM kernel? Or is it owing to a slowdown in the physical device?
Disk_write_latency The average amount of time taken for a write from the perspective of a guest operating system Secs This is the sum of kernel latency and physical device write latency. High latency is always a cause for concern, as it is an indicator of contention for storage resources.If the value of this measure is high, then check the values reported by the Kernel disk write latency and Physical device write latency measures for this storage adapter. Doing so enables you to quickly determine why exactly the guest OS is experiencing latencies while writing to this disk - is it because of latencies in the VM kernel? Or is it owing to a slowdown in the physical device?
Disk_cmd_latency The average amount of time taken for a command to execute, from the perspective of a guest operating system Secs This is the sum of kernel latency and physical device command latency. High latency is always a cause for concern, as it is an indicator of contention for storage resources.If the value of this measure is high, then check the values reported by the Kernel disk command latency and Physical device command latency measures for this storage adapter. Doing so enables you to quickly determine why exactly the guest OS is experiencing latencies while executing commands on this disk - is it because of latencies in the VM kernel? Or is it owing to a slowdown in the physical device?
Kernel_disk_read_latency The average time spent in the ESX server's VM kernel per read Secs A high value for this measure is a cause for concern. You might hence want to investigate the reasons for the same.
Kernel_disk_write_latency The average time spent in the ESX server's VM kernel per write Secs A high value for this measure is a cause for concern. You might hence want to investigate the reasons for the same.
Kernel_disk_cmd_latency The average time spent in the ESX server's VM kernel per command Secs  
Queue_read_latency The average time spent in the ESX server VM kernel queue per read Secs A high value of this measure indicates that the VMkernel is unable to process queued read requests quickly. If the problem persists, then the queue size could increase considerably. To avoid this, swiftly determine the reasons for a slowdown at the VMkernel, and fix it.
Queue_write_latency The average time spent in the ESX server VM kernel queue per write Secs A high value of this measure indicates that the VMkernel is unable to process queued write requests quickly. If the problem persists, then the queue size could increase considerably. To avoid this, swiftly determine the reasons for a slowdown at the VMkernel, and fix it.
Queue_cmd_latency The average time spent in the ESX server VM kernel queue per command Secs A high value of this measure indicates that the VMkernel is unable to process queued commands quickly. If the problem persists, then the queue size could increase considerably. To avoid this, swiftly determine the reasons for a slowdown at the VMkernel, and fix it.
Phy_device_read_latency The average time taken to complete a read from the physical device Secs Shortage of the physical storage resources can adversely impact the performance of the ESX host and the VMs configured on it. Therefore, if the value of this measure is very high or is steadily increasing, then quickly figure out what is pulling down the performance of the physical disk, and attend to it.
Phy_device_write_latency The average time taken to complete a write to the physical device Secs
Phy_device_cmd_latency The average time taken to complete a command from the physical device Secs
Lun_state The current state of the LUN   A LUN can be in one of the following states:
  • Error
  • Ok
  • Off
  • Unknown
  • Lost communication
  • Degraded
  • Quiesced

The numeric values that correspond to each of the aforesaid tests have been discussed in the table below:

State Value
Unknown 0
Ok 1
Off 2
Error 3
Lost communication 4
Degraded 5
Quiesced 6

Note:

By default, this measure reports the States listed in the table above to indicate the status of a LUN. The graph of this measure however, represents the status of a LUN using the numeric equivalents - 0 to 6.
Total_paths Indicates the number of storage paths through which the host communicates with this LUN. Number  
Queue_depth Indicates the number of outstanding I/O requests to this LUN for which a response has not been received from the LUN. Number A low value is desired for this measure. A high value is indicative of a large number of pending requests for the LUN and hints at a potential processing bottleneck on the LUN.
Isssd Indicates whether/not this LUN is SSD or non-SSD.   Solid State Disks (SSD) offer a much higher throughput and much lower latency than traditional magnetic hard disks, since they are based on flash memory. vSphere hosts can use locally attached SSDs as a host swap cache, as virtual flash, as a vSAN, or as a regular datastore.

If the LUN being monitored is SSD-based, then the value of this measure wwill be Yes. If not, then the value of this measure will be No.

The numeric values that correspond to each of the aforesaid tests have been discussed in the table below:

Measure Value Numeric Value
Yes 1
No 0

Note:

By default, this measure reports the Measure Values listed in the table above. In the graph of this measure however, the SSD status of the LUN is represented using the corresponding numeric equivalents only.
Health Indicates the current health status of this LUN.   The values that this measure can report and their corresponding numeric values are discussed in the table above:

State Value
Healthy 0
Failed 4
Offline 5
Decommissioned 6
Permanent Diak Failure 26

Note:

By default, this measure reports the Measure Values listed in the table above. In the graph of this measure however, the health status of the LUN is represented using the corresponding numeric equivalents only.
Capacity Indicates the total capacity of this LUN. MB  
Capacity_used Indicates the amount of space used in this LUN. MB Compare the value of this measure across LUNs to know which LUN is consuming space excessively.
Capacity_reserved Indicates the amount of space reserved for this LUN. MB  
Capacity_free Indicates the amount of free space currently available for this LUN. MB Compare the value of this measure across LUNs to know which LUN is rapidly running out of space.
Capacity_usage_perc Indicates the percentage of space in this LUN that is in use. Percent A low value is desired for this measure. A value close to 100% is a cause for concern, as it indicates a potential space crunch on the LUN. You may want to compare the value of this measure across LUNs to know which LUN is utilizing disk space excessively.