Re: why does raidframe retry I/O 5 times

[christos%astron.com@localhost (Christos Zoulas)]

In article <20120625165303.GA11930%asim.lip6.fr@localhost>,
Manuel Bouyer  <bouyer%antioche.eu.org@localhost> wrote:
>On Mon, Jun 25, 2012 at 09:39:45AM -0700, Brian Buhrow wrote:
>>      Hello.  While I agree with Manuel's assertion that the system becomes
>> stalled while all these retries are happening, I actually like all the
>> retries, because I've actually seen drives finally wake up and either remap
>> and write sectors they were complaining about, or figure out how to correct
>> the read error they were getting and deliver the data.  FreeBSD times out
>> much more quickly and loses drives seemingly more often.  I'd rather see
>
>In my experience, if a wd(4) drive couldn't get the data after 5
>retries, it's very unlikely it will get it after 25.
>
>> the wd(4) and raidframe(4) drivers made mp save which I think will address
>> the stalling issue, mostly, while retaining the persistent nature of the
>> current scheme.
>
>No, it won't. While a drive is retrying, I/O to this drive are
>stalled (obviously). 25 retries stalls I/O to this drive for 25 seconds,
>which stalls anything else which also needs this drive for 25 seconds.

We could add a sysctl...

christos