Re: bin/38471: Raidframe crashes on reconstruction of RAID5 (5 disks @ 298GB)

[Greg Oster <oster%cs.usask.ca@localhost>]

The following reply was made to PR bin/38471; it has been noted by GNATS.

From: Greg Oster <oster%cs.usask.ca@localhost>
To: thomas%t-online.de@localhost
Cc: gnats-bugs%NetBSD.org@localhost
Subject: Re: bin/38471: Raidframe crashes on reconstruction of RAID5 (5 disks @ 
298GB) 
Date: Sun, 20 Apr 2008 15:47:32 -0600

 thomas%t-online.de@localhost writes:
 > >Number:         38471
 > >Category:       bin
 > >Synopsis:       Raidframe crashes on reconstruction of RAID5 (5 disks @ 298G
 > B)
 > >Confidential:   no
 > >Severity:       critical
 > >Priority:       high
 > >Responsible:    bin-bug-people
 > >State:          open
 > >Class:          sw-bug
 > >Submitter-Id:   net
 > >Arrival-Date:   Sun Apr 20 19:20:00 +0000 2008
 > >Originator:     Thomas Feddersen
 > >Release:        4.0_BETA2
 > >Organization:
 > Dipl.-Ing Thomas Feddersen, Beratender Ingenieur
 > >Environment:
 > NetBSD bremen 4.0_BETA2 NetBSD 4.0_BETA2 (XEN3_DOM0) #0: Thu Mar  1 04:57:05 
 > PST 2007  builds@wb33:/home/builds/ab/netbsd-4/i386/200703010002Z-obj/home/bu
 > ilds/ab/netbsd-4/src/sys/arch/i386/compile/XEN3_DOM0 i386
 > >Description:
 > One disk of the raid-set has sectors pending. 
 > Apr 20 18:09:26 bremen smartd[201]: Device: /dev/wd5d, 1 Currently unreadable
 >  (pending) sectors
 > Apr 20 18:09:26 bremen smartd[201]: Device: /dev/wd5d, 1 Offline uncorrectabl
 > e sectors
 > 
 > I attempted an in-place reconstruction of the drive in question, but reconstr
 > uction lets the system drop into debugger.
 > 
 > Second attempt by adding a spare component and reconstructing to the spare ha
 > s the same effect.
 > 
 > The problem occurs with XEN3 (128MB total memory) or i386 (1GB total Memory) 
 > kernel likewise.
 > 
 > Need assistance to restore the redundancy of the raid set.
 
 You'll likely have to build a kernel with a bumped up NKMEMPAGES 
 value... e.g.:
 
  options NKMEMPAGES=32768
 
 You can see what your current values are with:
 
  sysctl vm.nkmempages
 
 and use a larger value.
 
 > >How-To-Repeat:
 [snip]
 > Give the command for in-place reconstruction:
 > 
 > bremen# raidctl -R /dev/wd5a raid0
 > 
 > the system will drop into debugger (green letters on console):
 > 
 > raid0: initiating in-place reconstruction on column 4
 > panic: malloc: out of space in kmem_map
 > stopped in pid 885.1 (raid_reconip) at  netbsd:cpu_Debugger+0x4:  popl
 > %
 > ebp
 > db>
 > 
 > the only way out is to reboot the system and operate the raid set in degraded
 >  mode
 
 I have a patch nearly done testing that will probably fix this...  
 
 Later...
 
 Greg Oster