Any minor thing makes server freeze

[Fern H2O]

Hello,

I just installed Dec 2007 HPUX 11.23 version on 2 of my test servers. They are both 7410's. I was pretty much done and turned it over to the DBA's but when they tried to ftp a 6gig file the server just freezes. So i noticed that once i try to ftp it basically uses up ALL of the disk I/O so thats why the server just freezes. But i also noticed that it does this for pretty much every command. Like when i finally got the file on the server, I tried to tar it and again the same problem. The Disk I/O sky rockets and the server goes ridiculously slow.

Does anyone know why this is?? Is there a patch that i'm missing? If anyone can help me out it would be greatly appreciated!

Thanks in advance!

[Duncan Edmonstone]

To what filesystem are you putting this file? Is it going onto the system disk (i.e. an LV in the root VG?)

Large sequential IOs to system disks should generally be avoided. See this thread for details...

http://forums11.itrc.hp.com/service/forums/questionanswer.do?threadId=1116138

HTH

Duncan

[Tim Nelson]

What is your dbc_max_pct set to ?

If oracle is installed it is recommended to be at or around 2-3% depending on total memory. ( 500MB - 800MB )

[Fern H2O]

Hello,

Yes, it's going to the root disk but like I said it's just a simple ftp and untar.

dbc_max_pct = 50.

Below is info from glance on mem.

Total VM : 438mb Sys Mem : 2.0gb User Mem: 330mb Phys Mem : 16.0gb
Active VM: 206mb Buf Cache: 8.0gb Free Mem: 5.7gb FileCache: na

Should I increase that parameter then??

[Fern H2O]

Also,

NOTHING is installed on it yet. Not SAP or Oracle or any other app. It's just a basic cold install right now which is why I don't understand why this is happeneing.

But soon, we'll be installing all those on it which is why i want to figure this out first.

[Duncan Edmonstone · Accepted solution]

Did you read the thread I pointed you at?

A simple ftp and untar can still generate significant amounts of sequential IO. If you have additional disks/LUNs available, try the same test on a non-system disk and see if the problem persists.

HTH

Duncan

[Tim Nelson]

The 8GB of buffer cache can and will bog down a server. The buffer needs to flush and it takes priority over many other processes.

The default of 50% has been a bad default for the longest time.

kctune dbc_mix_pct=3
kctune dbc_max_pct=5

These changes take effect immediately.

As I will assume you have not tuned your kernel yet you will want to change these and other application recommended changes before you turn things over.

And certainly, massive sequential IOs to the system internal disk will cause issue as Duncan mentioned. Address this also, but first things first.

[Fern H2O]

Thanks For your help Duncan & Tim.

I tried both your advice. Duncan, your right, when i tried the same thing on other disks besides the root disk it works pretty much normally.