[nycbug-talk] HP DL165G5 Issues

[Mark Saad]

John
  thanks for the info, I figured out what the issue was. Each time we tried to fix the issue I was using the same type of 1T drives from hp. Turns out they are the issue. Something on their disk firmware is causing them to flush their buffer every couple of seconds. I have no idea why and it makes the server useless when they are installed, regardless of what os you are running. For now I have swapped back to a OEM drive from seagate and its working fine. 

 


----- Original Message ----
> From: John Baldwin <jhb at freebsd.org>
> To: talk at lists.nycbug.org; mark.saad at ymail.com
> Sent: Thu, April 1, 2010 1:18:54 PM
> Subject: Re: [nycbug-talk] HP DL165G5 Issues
> 
> On Thursday 01 April 2010 10:31:51 am Mark Saad wrote:
> Hello 
> Talk
>   Any one out there using Hp DL165G5 servers. I have a about 
> 10 now and a 
number of the exhibit an odd issue. With freebsd , under i/o 
> load the system 
hangs for 6 to 10 seconds every 30 seconds or so , and  
> processes are stuck in 
state getblk or biord . I was thinking it could be a 
> disk or sata controller 
issue so I swapped them both out sata for a p400 sas 
> card and swapped drives a 
few times sas,sata mdl dual port single port  
> and no change .   So. Installed 
8.0 7.2 7.3 all have the same issue. I 
> then tried CentOS 5.1 and 5.4 and  
similar issues popped up. Formatting 
> a 1T drive took over 45min .  Any one 
have any ideas on what's toast 
> ?  What would a process stuck in getblk or 
biord refer to ? 
> 

"biord" is a thread waiting for an I/O read request to be completed (the 
> 
requesting thread sleeps in "biord" while GEOM passes the request down via 
> 
g_down and g_up).  "getblk" is probably also blocked on disk I/O, but 
> 
indirectly.  For example it may need to claim a buffer that is 
> currently in 
use for an in-process transaction for some other thread waiting 
> in "biord" or 
"biowr".

-- 
John Baldwin

--
Mark Saad
mark.saad at ymail.com