MSA1500cs Large AvQDep Value

[Eric K. Miller]

Anyone know why every so often a LUN on an MSA1500cs will show a very large AvQDep (queue depth) value?

A "show perf" is included below. Notice LUNs 5, 6, and 7 (especially 7)?

Thanks!

Eric

CLI-1> show perf

Sample interval: 778.2s
Structure version: 101
% CPU utilization: 40%
Command count: 113447
Avg Command Latency: 0.64ms
% Time Active Reqs: 5.6%
Max Outst Cmd Lists: 28
Logical Req Count: 113447
Avg Free Phys Reqs: 1587
Avg Free Log. Reqs: 948
Avg Free Xbuf Sects: 0
Avg Lrgst Avail Xb S: 0
Avg Free CPU DRAM KB: 0
Max DMA Xfer Q Depth: 0
% Time DMA Xfr Stall: 0.0%
Avg Locked Stripes: 0
Avg Stripes Waiting: 0
Avg Completions/Sec: 145.7
Avg Log Req Comp/Sec: 145.7

Logical volume data: (All values in hex except Vol and AvQDep)
Vol LWrites LReads AvQDep CHits CMiss RACount LSectRd LSectWr
0 4615 156D 153 13B8 1B5 0 2D1A5 64DBE
1 61A 32 10 B 27 0 1173 3DCD
2 106A 6B 0 43 28 0 ED3 92F4
3 96A 27 0 A 1D 0 1363 895A
4 40C A 0 0 A 0 1141 40C
5 5961 B71 8664 305 879 0 EC6C 594EB
6 3177 DDC 3305 B0 D2C 0 EF34 422A8
7 3FBF 7A6 134902 639 16D 0 F9C1 24911
8 2725 275 0 25F 16 0 51B9 1D782
9 0 0 0 0 0 0 0 0
10 1484 3B 2 23 18 0 1A5B 111E9
11 10F 38 0 B 2D 0 17A3 10F
12 C 2B 0 C 2F 0 14A3 C
13 427 36 1 16 20 0 B4F 2310
14 575 75 0 C 69 0 BD5 35E4
15 435 24 0 A 1A 0 B23 23AD

[Srikanth.D]

Hi Eric, could you please provide us with details of the Setup. The performance can be hindered due to various reasons. Probably the most required data is mapped to Lun7. etc..

[Eric K. Miller]

This MSA1500cs is connected to 4 MSA30's fully-populated with 146GB disks and some 300GB disks.

I attached a "show tech_support" to provide some more details.

There are 4 ESX 3.5 U1 hosts connected to this MSA1500cs.

Eric

[Eric K. Miller]

I forgot to mention that I also did a "clear perf" and "show perf" at the end of that attached "show tech_support" to show that an almost clean show perf has a huge value for these 3 LUNs.

Thanks!

Eric

[John Kufrovich]

Erik,
It's a miscalculation bug in cli portion of the code.

jk

[Eric K. Miller]

Thanks John! We'll assume there is nothing to worry about.

Eric