Re: SCSi issues after 10.0 update

Hi,

I stressed the system for days...
I have it serving over NFS the same stuff as before.
In the meanwhile I had it compile for 33 hours something quite heavy, 
lots of filesystem as well as swapping.
Up to now, no issue.

I would exclude a hard disk issue (do these older parallel SCSI drives 
have something like SMART?).
So more a kernel issue... of course, pretty wild to guess what and where?

Riccardo

Riccardo Mottola wrote:
> Hi,
>
> Today, my freshly updated Netra T1 which serves NFS to all my 
> sparcs... (including pkgsrc and package binaries) started to be not 
> responsive.
>
> I attach serial console and find myself in the debugger, stacktrace 
> looks like in an drive,  I type continue and see a stream of this:
>
>
> [ 152678.9197479] sd1a: error writing fsbn 15551936 of 
> 15551936-15551967 (sd1 bn 15551936; cn 4330 tn 21 sn 113)
> [ 152679.0383219] sd1a: error writing fsbn 15552032 of 
> 15552032-15552063 (sd1 bn 15552032; cn 4330 tn 22 sn 76)
> [ 152679.1558558] sd1a: error writing fsbn 15552192 of 
> 15552192-15552223 (sd1 bn 15552192; cn 4330 tn 23 sn 103)
> [ 152679.2744306] sd1a: error writing fsbn 15551840 of 
> 15551840-15551871 (sd1 bn 15551840; cn 4330 tn 21 sn 17)
> [ 152679.3919657] sd1a: error writing fsbn 15551968 of 
> 15551968-15551999 (sd1 bn 15551968; cn 4330 tn 22 sn 12)
> [ 152679.5094991] sd1a: error writing fsbn 15552096 of 
> 15552096-15552127 (sd1 bn 15552096; cn 4330 tn 23 sn 7)
> [ 152679.6259930] sd1a: error writing fsbn 15552224 of 
> 15552224-15552255 (sd1 bn 15552224; cn 4330 tn 24 sn 2)
> [ 152679.7445680] sd1a: error writing fsbn 15551872 of 
> 15551872-15551903 (sd1 bn 15551872; cn 4330 tn 21 sn 49)
> [ 152679.8600212] sd1a: error writing fsbn 15552000 of 
> 15552000-15552031 (sd1 bn 15552000; cn 4330 tn 22 sn 44)
> [ 152679.9775560] sd1a: error writing fsbn 15552128 of 
> 15552128-15552159 (sd1 bn 15552128; cn 4330 tn 23 sn 39)
> [ 152680.0950897] sd1a: error writing fsbn 15552256 of 
> 15552256-15552287 (sd1 bn 15552256; cn 4330 tn 24 sn 34)
> [ 152680.2147047] sd1a: error writing fsbn 15551904 of 
> 15551904-15551935 (sd1 bn 15551904; cn 4330 tn 21 sn 81)
> [ 152680.3301591] sd1a: error writing fsbn 15552064 of 
> 15552064-15552095 (sd1 bn 15552064; cn 4330 tn 22 sn 108)
> [ 152680.4487334] sd1a: error writing fsbn 15552160 of 
> 15552160-15552191 (sd1 bn 15552160; cn 4330 tn 23 sn 71)
> [ 152680.5662672] sd1a: error writing fsbn 15552320 of 
> 15552320-15552351 (sd1 bn 15552320; cn 4330 tn 24 sn 98)
> [ 152680.6858827] sd1: sync (50.00ns offset 31), 16-bit (40.000MB/s) 
> transfers, tagged queueing
> [ 152680.9942874] esiop0: selection timeout without command, target -1 
> (sdid 0x4), slot 10
> [ 152681.0884124] esiop0: unhandled scsi interrupt, sist=0x400 
> sstat1=0xf DSA=0xc034f130 DSP=0x3d0
> [ 152681.1924234] esiop0: scsi bus reset
> [ 152681.2361087] sd1(esiop0:0:1:0): command with tag id 0 reset
> [ 152681.3047562] sd1(esiop0:0:1:0): command with tag id 1 reset
> [ 152681.3734043] sd1(esiop0:0:1:0): command with tag id 2 reset
> [ 152681.4420526] sd1(esiop0:0:1:0): command with tag id 3 reset
> [ 152681.5107008] sd1(esiop0:0:1:0): command with tag id 4 reset
> [ 152681.5793498] sd1(esiop0:0:1:0): command with tag id 5 reset
> [ 152681.6479976] sd1(esiop0:0:1:0): command with tag id 6 reset
> [ 152681.7166465] sd1: async, 8-bit transfers, tagged queueing
> [ 152681.7842901] sd1: sync (50.00ns offset 31), 16-bit (40.000MB/s) 
> transfers, tagged queueing
> [ 152681.8869196] esiop0: DMA IRQ: Illegal instruction DMA fifo empty, 
> DSP=0x4f0 DSA=0xc034e790:  current T/L/Qd
> [ 152682.0127705] esiop0: scsi bus reset
> [ 152682.0564558] sd1(esiop0:0:1:0): command
> CTRL-A Z for help | 9600 8N1 | NOR | Minicom 2.8 | VT102 | Offline | 
> tty00
>
> of course, haard disks can go havoc... the Netra has two hard disks, 
> sd1 is the one which serves files.
> Or could it be that NetBSD 10 isn't that stable? Network and Disk I/O 
> caused trouble?
>
> I quickly power-cycled. System rebooted, fsck'd volumes mounted and 
> NFS worked, no more SCSI errors. NFS resumed and worked until I shout 
> down everything for Friday evening.
>
> Will retry next week... but looks like esiop0 did hang up on sd1!
>
> Riccardo
>