ffs issue was Re: ataraid issue was Re: [netbsd-7] Critical issue with ffs+log

[BERTRAND Joël <joel.bertrand%systella.fr@localhost>]

Michael van Elst a écrit :
> On Fri, Jan 22, 2016 at 11:03:51AM +0100, BERTRAND Joël wrote:
> > Michael van Elst a écrit :
> > > joel.bertrand%systella.fr@localhost (=?UTF-8?Q?BERTRAND_Jo=c3=abl?=) writes:
> > >
> > > > 	By dkctl wd0/1 setcache none save ? I will try next saturday. But if
> > > > both disks are unsynchronized, how can I force a resynchronization ?
> > >
> > > There is no sane way. For a resynchronization you'd need at least a single
> > > bit that tells you that both sides are out of sync (and better some
> > > kind of journal or a parity map).
> > >
> > > Copying the raid to itself will of course synchronize both disks.
> >
> > 	I don't understand. If I use a real raid controler (or raidframe), I can
> > remove and readd a disk and I have to resynchronize raid volume. Maybe is my
> > question trivial, but with ataraid, how can I swap device when a physical
> > disk is dying ?
>
>
> The ataraid driver doesn't have many features. It can perform reads/writes
> on RAID0 and RAID1 and handle disk failures for RAID1. You can query
> the current state with the bioctl utility.
>
> Creation, verification or recovery of raid volumes is not supported by
> the driver. There is usually a BIOS extension that lets you do this.

	Motherboard manual indicates there is only one way to resynchronize a 
raid array : destroy and rebuild (!). I don't like and I have migrated 
my system on a pure raidframe configuration I use for a long time on 
Sparc64.

	I haven't reinstalled my system, I only have copied it from ataraid to 
raidframe with pax. Now my /etc/fstab contains :

/dev/raid0a             /       ffs     rw       1 1
/dev/raid0b             none    swap    sw,dp    0 0
/dev/raid0e             /usr    ffs     rw       1 2
/dev/raid0f             /var    ffs     rw       1 2
/dev/raid0g             /home   ffs     rw       1 2
kernfs                  /kern   kernfs  rw
ptyfs                   /dev/pts        ptyfs   rw
procfs                  /proc   procfs  rw
/dev/cd0a               /cdrom  cd9660  ro,noauto
tmpfs                   /var/shm        tmpfs   rw,-m1777,-sram%25

	I only have configured one raid-1 volume :

legendre# raidctl -v -s raid0
Components:
           /dev/wd0a: optimal
           /dev/wd1a: optimal
No spares.
Component label for /dev/wd0a:
   Row: 0, Column: 0, Num Rows: 1, Num Columns: 2
   Version: 2, Serial Number: 2016012301, Mod Counter: 65
   Clean: No, Status: 0
   sectPerSU: 128, SUsPerPU: 1, SUsPerRU: 1
   Queue size: 100, blocksize: 512, numBlocks: 1953517440
   RAID Level: 1
   Autoconfig: Yes
   Root partition: Force
   Last configured as: raid0
Component label for /dev/wd1a:
   Row: 0, Column: 1, Num Rows: 1, Num Columns: 2
   Version: 2, Serial Number: 2016012301, Mod Counter: 65
   Clean: No, Status: 0
   sectPerSU: 128, SUsPerPU: 1, SUsPerRU: 1
   Queue size: 100, blocksize: 512, numBlocks: 1953517440
   RAID Level: 1
   Autoconfig: Yes
   Root partition: Force
   Last configured as: raid0
Parity status: clean
Reconstruction is 100% complete.
Parity Re-write is 100% complete.
Copyback is 100% complete.

	Please note I have disabled log option in fstab to mount all slices in 
pure FFSv2 mode.

	I see again lot of data corruption. For example :
legendre# make clean
===> Cleaning for kbruch-4.14.3nb2
rm: fts_read: No such file or directory
*** Error code 1

Stop.
make: stopped in /usr/pkgsrc/misc/kbruch
legendre# rm -rvi work/
remove 'work'? y
remove 'work/.buildlink'? y
remove 'work/.buildlink/include'? y
remove 'work/.buildlink/include/boost'? y
remove 'work/.buildlink/include/boost/interprocess'? y
remove 'work/.buildlink/include/boost/interprocess/mem_algo'? y
remove 'work/.buildlink/include/boost/interprocess/mem_algo/'? y
rm: work/.buildlink/include/boost/interprocess/mem_algo/: No such file 
or directory
rm: fts_read: No such file or directory
legendre# cd work/.buildlink/include/boost/interprocess/
legendre# ls -l
total 6
drwxr-xr-x  2 root  wheel   512 Jan 25 00:40 mem_algo
lrwxr-xr-x  1 root  wheel    50 Jan 25 00:39 offset_ptr.hpp -> 
/usr/pkg/include/boost/interprocess/offset_ptr.hpp
lrwxr-xr-x  1 root  wheel    51 Jan 25 00:39 permissions.hpp -> 
/usr/pkg/include/boost/interprocess/permissions.hpp
lrwxr-xr-x  1 root  wheel    55 Jan 25 00:39 segment_manager.hpp -> 
/usr/pkg/include/boost/interprocess/segment_manager.hpp
lrwxr-xr-x  1 root  wheel    60 Jan 25 00:39 shared_memory_object.hpp -> 
/usr/pkg/include/boost/interprocess/shared_memory_object.hpp
drwxr-xr-x  3 root  wheel   512 Jan 25 00:39 smart_ptr
drwxr-xr-x  7 root  wheel  1024 Jan 25 00:39 sync
lrwxr-xr-x  1 root  wheel    61 Jan 25 00:39 windows_shared_memory.hpp 
-> /usr/pkg/include/boost/interprocess/windows_shared_memory.hpp
lrwxr-xr-x  1 root  wheel    47 Jan 25 00:39 xsi_key.hpp -> 
/usr/pkg/include/boost/interprocess/xsi_key.hpp
lrwxr-xr-x  1 root  wheel    57 Jan 25 00:39 xsi_shared_memory.hpp -> 
/usr/pkg/include/boost/interprocess/xsi_shared_memory.hpp
legendre# cd mem_algo/
legendre# ls -la
ls: .: No such file or directory
legendre#

	I have tried to remode faulty inode with clri without any result and I 
only found one solution : adding -fy in /etc/rc.d/fsck and rebooting 
this server. I have tried to remount /usr in readonly mode to run fsck 
without result :

legendre# mount -ur /usr
mount_ffs: /dev/raid0e on /usr: Operation not supported

	Kernel is GENERIC.201512202240Z

	Regards,

	JKB