geom_raid5 livelock?

[R. B. Riddick]

--- CyberLeo Kitsana <cyberleo at cyberleo.net> wrote:
> Good morning!
>
Thx. U2! :-)

> I'm not sure what error 5 is, but it looks ominous.
>
That is EIO... Dont know where it comes from.

We would know better, if u change the line around 1292 or so from
  G_RAID5_LOGREQ(bp, "[done]");
to
  G_RAID5_LOGREQ(bp, "[done err%d]",bp->bio_error);
and after line 1160 (the line with = EIO;)
  G_RAID5_LOGREQ(bp, "[not-equal]");
Then wait for this suspected live lock again and look for messages and then set
debug level to 2...

> The first log shows two seconds of the first test, where only ad2s2 was 
> showing up. The second log is after a restart of everything, and ad0s2, 
> ad2s2, and ad6s2 show up, indicating that this most likely isn't just a 
> drive, bus, or controller failure.
>
Possibly... Maybe U need to run a verify manually (graid5 configure -R
<device>)... Are u sure, that a complete verify run took place after label'ing
the graid5 device?

> The machine is on a UPS, so power loss isn't too much of an issue. What 
> other impacts would reducing kern.geom.raid5.wdt have?
> 
OK - UPS is nice...
<wdt> is the number of seconds, graid5 waits before it executes a write
request... So lower wdt means, lower chances for write optimization in favor to
lower chances for data loss...

Btw.: A kernel crash could result in such data loss, too (just like a power
failure)...

-Arne


 
____________________________________________________________________________________
8:00? 8:25? 8:40? Find a flick in no time 
with the Yahoo! Search movie showtime shortcut.
http://tools.search.yahoo.com/shortcuts/#news