Possible RAID Controller performance issue

[JFMorgan]

This is my first post to Technibble. I support small business clients in northern Alabama. Recently we installed a brand new HP Proliant ML150 Gen 9 entry level server for a client. The server is running Windows Server 2012 R2 Essentials. The server has an Intel Xeon CPU E5-2609 v3 @ 1.90GHz 6-core CPU with 8GB RAM. The customer has 7 users and a lot of customer data and many large graphic files. They have about 700 GB of customer data. The customer uses a Client/Server Line of Business (LoB) application which has a SQL Express backend. The LoB application has a server applet and a client application. The Operating system and the LoB app run off a RAID 5 array. We are using the built-in HP Dynamic Smart Array B140i Controller. We have five 1TB Seagate Barracuda SATA drives configured as a RAID 5 array with a hot spare. We have two Intel 160GB SSD drives configured as a RAID 1 (mirrored) array on which we have the SQL database for the LoB application. We installed this new server in late November. We are experiencing continuing problems with the LoB server applet crashing. In addition, the customer experiences occasional slow responsiveness from the LoB application and when opening and saving customer files to/from the RAID 5 array. The delay in response is about 30 to 40 seconds. We use CrashPlan PROe to backup the server and we thought the backup processing may be the cause of the slow responsiveness. We changed the backup schedule so that backups do not occur during business operations, but the problems continue to occur. I've found the following entries in the Windows Application Event Logs which I believe are related to this issue:

Date: 12/30/2015 12:26:36 AM
Event Type: Warning
Event Source: ESENT
Event Category: Performance
Event ID: 508
User: N/A
Event Description: svchost (3380) A request to write to the file "C:\Windows\system32\LogFiles\Sum\Svc.log" at offset 319488 (0x000000000004e000) for 4096 (0x00001000) bytes succeeded, but took an abnormally long time (27 seconds) to be serviced by the OS. This problem is likely due to faulty hardware. Please contact your hardware vendor for further assistance diagnosing the problem.

We are finding numerous such entries in the Event Logs. The time stamp roughly coincides with the LoB Server applet shutting down.

Any thoughts as to the root cause?

 

[Metanis]

Just 2 thoughts at first glance. 1.) Only 8GB of RAM? 2.) This might be a job for Process Explorer to help zero in on the offending hardware sub-system

 

[JFMorgan]

That's a good suggestion. Thank you!

 

[freedomit]

Also make sure you apply to latest HP Proliant Service Pack to update all drivers and firmware.

And check the shadow copy schedule

 

[DataMedics]

Also could be that the hard drive's are starting to degrade, it's all too common with cheap consumer grade Seagate's. I'm recovering an array right now where 3 out of 5 Seagate drives in a RAID 5 all had SMART errors relating to re-allocated sectors. The problem is the controller doesn't notice the issue until you reboot the machine, and by then you've got multiple drive failures. Try using HGST drives when you upgrade and see the difference.

 

[freedomit]

Also could be that the hard drive's are starting to degrade, it's all too common with cheap consumer grade Seagate's. I'm recovering an array right now where 3 out of 5 Seagate drives in a RAID 5 all had SMART errors relating to re-allocated sectors. The problem is the controller doesn't notice the issue until you reboot the machine, and by then you've got multiple drive failures. Try using HGST drives when you upgrade and see the difference.

They are brand new so unlikely but SATA and Raid5 and not going to be quick anyway

 

[YeOldeStonecat]

Couple of quick thoughts...

*Server Essentials...much like SBS...is extra heavy/bloated. Gotta up the horsepower of the rig it's on. 8 gigs was barely tolerable for Server 2008 Standard...but for SBS or Essentials or Server 2012...ouch! And a SQL app to boot?

*I want speed on the drive/volume that the data is on. SSD for boot drive is great...for the server doing reboots, or booting up. But how often is that? The clients use the data on the data shares all day long. SATA...ouch. RAID 5....long write times, and not good for frequently changing data such as databases. Good for storage of accessed files..as it has decent read times. But not good for changing data and heavy writes. SATA on RAID 5...double ouch.

*I love HP servers and their RAID controllers...but that's a fake RAID controller..not a typical Smart Array controller. no BBU, uses the CPU (software based) and uses system RAM.

 

[DataMedics]

They are brand new so unlikely but SATA and Raid5 and not going to be quick anyway

With the 38% annual failure rate that their ST3000DM001 models are facing, don't be so sure. I've already seen a number of brand new Seagate arrays show up with multiple failures. Never hurts to hook up a drive and at least pull the SMART stats.