Re: MFC of "Large set of CAM improvements" breaks I/O to Adaptec

On Apr 29, 2010, at 10:56 PM, Alexander Motin wrote:
> Scott Long wrote:
>> On Apr 29, 2010, at 7:47 AM, Robert Noland wrote:
>>>=20
>>> Scott Long wrote:
>>>> On Apr 29, 2010, at 2:50 AM, Pete French wrote:
>>>>>> Thanks. First step successful - I can steadily reproduce problem =
on
>>>>>> CURRENT. raidtest with 200 I/O streams over gmirror of two disks =
on same
>>>>>> channel triggers issue in seconds. Any I/O on channel dying after =
both
>>>>>> disks report "Queue full" error same time. The rest of system =
works
>>>>>> fine. If I preliminarily manually adjust queue depth of one disk =
-
>>>>>> everything works fine. I'll investigate it tomorrow.
>>>>> Glad you have managed to dupliate it - the queue depth thing is
>>>>> inetersting, what changes did you make ? I can try them here and =
see
>>>>> if they improve the situation on either of my two machines.
>>>>>=20
>>>> For the record, queue-full is a common, expected condition in CAM.  =
It's not something that should be avoided =3D-)
>>> Should we maybe have a counter in sysctl rather than flooding the =
console with these messages then?
>>=20
>> That's a pretty good idea.  I'll make it happen.
>=20
> It is already hidden behind bootverbose. Hiding it deeper will make
> unclear why CAM requeues the rest of commands (also reported under
> bootverbose). I've tuned log messages a bit recently and they seem to =
be
> more consistent and readable now IMHO.
>=20

We used to run FreeBSD at Yahoo with bootverbose turned on in order to =
help with debugging.  After years of doing this, I finally turned =
bootverbose off last year, partially because of the excessive console =
spam produced by these queue-full messages.  Even when we were writing =
the ahc/ahd drivers at Adaptec years ago, I never really liked these =
messages, and we rarely ran with bootverbose turned on unless we were =
actively developing code or debugging a problem.  I like Robert's =
suggestion because not only does it make running with bootverbose less =
painful, it can also provide counters and also calculate and report rate =
measurements that might be more useful than just the printf.

If you feel strongly against it, I won't push it, but I do like the =
suggestion.

Scott

_______________________________________________
[e-mail] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "[e-mail]"