Jay wrote:
> hi *,
>
> i'm currently playing around with the setup of an opensolaris server as
> home nas and am experiencing occasional read/write problems with the
> zfs pool.
>
> the short version (details below/attached):
> * 6-disk raidz pool attached to the sata controller on an nvidia MCP78S
> chipset
> * first scrub of the pool with some data on it marks device sd5 as faulted
> due to "WARNING: ahci0: watchdog port 5 satapkt 0xffffff01c7d8d660 timed
> out"
> and a plethora of "Error for Command: read(10)" (see attached messages)
>
Jay,
if you search the bugs database for this error message,
http://bugs.opensolaris.org
you will find a number of hits. Many possibly related bugs
have been fixed by b101, but there may be more.
You should also ask this question on the drivers-discuss forum
as that is where the device driver writers hang out.
-- richard
> * these messages appeared also for sd1, sd2 and sd3, but only sd5 failed in
> the end
> * replaced the disk, resilvering started
> * the same timeouts appear for sd0 and sd1 while resilvering, to prevent the
> pool from
> failing completely, i (rather brute force) rebooted the machine
> * resilvering ends eventually, data seems intact
> * everything seems normal for a few days, reading/writing is ok, no errors
> show up, the
> data is accessible
>
> today, i saw the same errors reported for sd4 in the logfile and when trying
> a 'zpool status'
> it became unresponsive, with timeouts showing up for sd0. after another
> reboot, everything still looks ok, zpool status is ok, read and write access
> are ok.
>
> the disks themselves should be ok, i had them running a burn-in before
> installing opensolaris and the WD diagnostics passed them - even the faulted
> one i replaced passed another test as being perfectly ok.
>
> can anybody shed some light on this? i'm guessing it's related to the sata
> controller, but i'd appreciate any help or insight.
>
> (at the moment, i'm not really worried about data loss as you might guess
> from the brute
> force rebooting, all the data on the pool is also stored on an old linux
> machine. i'm reacquainting myself with solaris, so it's more or less a
> playground for now. but i'd like to replace the old linux server sometime -
> mainly because of zfs)
>
> thanks,
> jay
>
>
_______________________________________________
zfs-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
