gvinum behaviour on disk failure

[Alastair Watts]

Hi list,

The other night we had a drive die under gvinum on FreeBSD 5.4. The 
setup was ad0 and ad2 in a mirror. Log messages were:

Aug 26 23:11:49 mail kernel: ad2: FAILURE - READ_DMA 
status=51<READY,DSC,ERROR> error=4<ABORTED> LBA=80176
Aug 26 23:11:49 mail kernel: GEOM_VINUM: subdisk swap.p1.s0 state 
change: up -> down
Aug 26 23:11:49 mail kernel: GEOM_VINUM: plex swap.p1 state change: up 
-> down
Aug 26 23:11:49 mail kernel: GEOM_VINUM: subdisk root.p1.s0 state 
change: up -> down
Aug 26 23:11:49 mail kernel: GEOM_VINUM: plex root.p1 state change: up 
-> down
Aug 26 23:11:49 mail kernel: GEOM_VINUM: subdisk tmp.p1.s0 state change: 
up -> down
Aug 26 23:11:49 mail kernel: GEOM_VINUM: plex tmp.p1 state change: up -> 
down
Aug 26 23:11:49 mail kernel: GEOM_VINUM: subdisk usr.p1.s0 state change: 
up -> down
Aug 26 23:11:49 mail kernel: GEOM_VINUM: plex usr.p1 state change: up -> 
down
Aug 26 23:11:49 mail kernel: swap_pager: I/O error - pagein failed; 
blkno 9977,size 4096, error 0
Aug 26 23:11:49 mail kernel: vm_fault: pager read error, pid 30400 (httpd)
Aug 26 23:11:49 mail kernel: pid 30400 (httpd), uid 1002: exited on 
signal 11

Am I missing something (perhaps data didn't get written to ad0, but then 
shouldn't it have been in memory), or shouldn't gvinum have absorbed 
this error and supplied the data from ad0?

Or is this a known issue/bug? I searched but couldn't find a similar 
reference on this list..

Cheers,

Al