Re: qla2300 driver stability, (was Re: 2.4.20, .text.lock.swap cpu usage?)

Re: qla2300 driver stability, (was Re: 2.4.20, .text.lock.swap cpu usage?)

[Brian Tinsley]

>>problem. I've had this driver running in my lab and at numerous client 
>>sites for quite some time and have never seen it even burp.
>>    
>>
>
>how hard did you push it in testing?
>
I've hit it with simulations of client loads as well as outright abusive 
load testing for several days nonstop without problem.

>in some configurations i've had to subject it to many hours of
>intensive IO (ie multiple concurrent and continious bonnie++ runs of
>varying file sizes) in order to get it to spin in
>qla2x00_intr_handler. but it will eventually hang given enough IO ime.  
>(in other configurations, heavy sustained IO will lock it up in
>minutes, even seconds).
>
Maybe I need to load bonnie++ in the lab and see what happens.


-- 

Brian Tinsley
Chief Systems Engineer
Emageon

Re: 2.4.20, .text.lock.swap cpu usage? (ibm x440) [rescued]

[Brian Tinsley]

>
>
>Which driver are you using for the QLA23xx?
>
The 6.01.00-fo driver from QLogic.

>I've missed the beginning of this thread, but problem here may not 
>have anything to with the number of threads you're runnng, rather it 
>may well be the qla2300 driver, if you are using the qlogic v6 driver.
>
I was getting nailed by an issue involving reaping inodes (or the lack 
thereof). I don't believe the QLogic driver was contributing to my 
problem. I've had this driver running in my lab and at numerous client 
sites for quite some time and have never seen it even burp.

>I have a test system in work which is similar to yours (2x qla2310F,
>SMP (dual athlon) attached to FC RAID storage) and it is easy to live
>lock with any kind of intensive IO, eg bonnie++.
>
>The Redhat 5.31-RH driver is about the most stable one, but i havnt
>extensively stress tested it. All of the qlogic v6 drivers are trivial
>to lock. (they spin forever in the qla2300 ISR - qlogic have beta
>drivers, but they still have the same problem).
>
Interesting. Again, I've never seen this behavior, but I appreciate your 
mentioning it. It's definitely something to keep an eye out for.

-- 

Brian Tinsley
Chief Systems Engineer
Emageon

qla2300 driver stability, (was Re: 2.4.20, .text.lock.swap cpu usage?)

[Paul Jakma]

On Sun, 12 Jan 2003, Brian Tinsley wrote:

> The 6.01.00-fo driver from QLogic.

hmm..

> problem. I've had this driver running in my lab and at numerous client 
> sites for quite some time and have never seen it even burp.

how hard did you push it in testing?

in some configurations i've had to subject it to many hours of
intensive IO (ie multiple concurrent and continious bonnie++ runs of
varying file sizes) in order to get it to spin in
qla2x00_intr_handler. but it will eventually hang given enough IO ime.  
(in other configurations, heavy sustained IO will lock it up in
minutes, even seconds).

> Interesting. Again, I've never seen this behavior, but I appreciate
> your mentioning it. It's definitely something to keep an eye out
> for.

regards,
-- 
Paul Jakma	Sys Admin	Alphyra
	paulj@alphyra.ie
Warning: /never/ send email to spam@dishone.st or trap@dishone.st