VA7400 logprn messages

[Sylvia Welch]

I was looking thru armlog the other day, and noticed quite a few of these:

Device Unique Name = 50060B000008AD80
Device Name = HP StorageWorks Virtual Array 7400
Date = September 16, 2005
Time = 6:11:12 AM EDT
Controller Time Stamp = 22402005419101
Event Number = 142611
Sequence Number = 1075
Event Type = Controller Event
Event Code = 348
Severity = 1
Event Count = 1
FRU Location = M/C1.H1
FRU Data = enclosureId/slot/component/subcomponent : 0x00/0x00/0x70/0xff
Reporting Controller = M/C1
Event Source Info = serialNum/moduleId/processId : 00PR05B51074/0x47/0xffffffff
Event Description = FRONTEND_FC_ABTS_EVENT_EH This error code indicates that the Host sent a Fibre Channel ABTS (Ab
ort Sequence) BLS frame to the abort an IO. The array will log this event for informational and debug purposes only.
It does not necessarily indicate a problem with the array.
Event Unique Info =
0x17028a: 0x0006
frontend osPortCB: ABTSAbort-JCB
NPortID=0x000002,OXID=0x0006,LUN=0x4050
CDB=0x2a00005b624800000800000000000000
et=10.2 qd=17 hwm=70 jl=0x00001000 dl=0x00001000 ro=0x00000000 js=1

The Queue Threshold is set to 74. We recenty added a tray of disks resulting in
32 disks in RG1 and 28 in RG2. Also, the space for LUNs is well below 50%. I was seeing these errors before the additional drives, but more frequently now.

All disk paths are optimal and pvtimeout is set to 90. Is the Queue Threshold too
low?

Thanks

Sylvia

[Sameer_Nirmal]

Which firmware and CV-SDM version you have?
Which OSs and HBAs you are using?

It seems that I/O load of the hosts connected is quite high. The Queue threshold should be increased to take care of the load.
But I would advise the calculated approch to the issue.

You can monitor the CV-SDM metrics like Queue Full Response / Queue Full Status Sent,
Command Frames Received , Command Aborts Received/Abort Sequences Received.
Looking at the event log, host is aborting I/O on PortID=0x000002,OXID=0x0006,LUN=0x4050
while accessing this LUN the theshold value was already reached. Array would send an Queue Full response to the host which will make the outstanding I/O aborted by the host. This LUN may be highly used. You may see such entries for other LUNs as well.

I won't surprise to know as the frequency of the events is increased. It is quite usual as more disks are added , more LUNs are created , the host I/O is going to increase.
Remember common queue threshold value is applied to all host ports in VA7400.

You need to monitor those metrics and increase the threshold value suitable for the setup.

Be aware, setting this value vary high would result in longer response time and time-outs
So my advise, be generous in setting up this value.

[Sylvia Welch]

Firmware is HP19

OS 11.00, patched up

HBA's HP Tachyon TL/TS Fibre Channel Mass Storage Adapter HP5158A.

What is the formula for the Queue Threshold?

Syl

[Sameer_Nirmal]

The simple formula for Queue Threshold would be
Q = H * P * q * L
Where Q = Queue Threshold
H = Number of Hosts connected to the
array port
P = Number of hosts paths connected to
the array port.
q = queue depth per LUN on hosts
for HP-UX it is 8.
L = Number of LUNs configured on the
array port seen from hosts

Be aware the maximum Queue Threshold value would be 750 for VA7400 .

[Jim Viera]

Sameer,

I was told by numerous HP engineers at the response center that the queue full threshold parameter should be set to 2 to 3 times the number of disk drives per redundancy group (RG). My experience has been that when the array gets too busy or there is a problem (disk failed/rebuild), that I receive these events.

I've had 2 va7400 arrays for over 3 years and the setting of this parameter per HP or other settings had much effect.

Has HP changed their recommendations?????

[Sylvia Welch]

So if I go with the 2 to 3 times disks in each RG, does that mean LUNS in each RG or physical disks assigned to each RG? Example, I have 60 disks total, that would be 30 in RG1 and 30 in RG2 which would make the Threshold 60-90? On the other hand I have 24 LUNS assigned to RG1 and 22 assigned to RG2.

[Jim Viera]

Sylvia,

It's physical disks not LUNs. Since your already at 74 you could bump it up to 96.

I want to clarify my previous post. I meant that the setting of this parameter in my environment (HPUX, Windows, Netware) has NOT made any significant effect in regards to Event 348.

[Sylvia Welch]

Thanks. Think I'll bump it up to 90 for now and keep an eye on it. Think I'll also do the math and figure what the threshold will be. I've also heard of the 2 to 3 times the disk/RG rule.

Syl

[Sylvia Welch]

I'm still seeing the aborts in armlog. Trying to track down the LUN responsible. How do I translate the error to a LUN?

Example:


NPortID=0x000002,OXID=0x0006,LUN=0x4050


Thanks.

[Jim Viera · Accepted solution]

Sylvia,

The lun would be the decimal value of the last 2 digits in LUN 0x4050 which is lun 80.