LVM trava ao tentar criar meu nó de dispositivo raiz

1

Acompanhamento para Como posso verificar se há blocos ruins em um volume físico LVM?

O título resume principalmente. Basicamente, eu tenho uma caixa particionada com uma partição /boot regular e, em seguida, um volume físico LVM preenchendo o restante da unidade. No LVM, tenho um grupo de volumes com uma partição raiz, uma partição /home e uma partição swap.

Quando o LVM cria os nós de dispositivos em /dev/mapper , ele cria as partições swap e home particionai. No entanto, ele geralmente trava ao tentar criar o nó do dispositivo raiz. Isso acontece a partir de um CD ao vivo ( pvscan; vgscan; vgchange -ay é o que eu usei, IIRC) e também do ramdisk inicial, impedindo que a caixa seja inicializada. Eu também tentei a partir do shell de recuperação initrd ( lvm pvscan; lvm vgscan; lvm vgchange -ay é o que eu usei, IIRC), que também falha da mesma maneira.

Às vezes, vgchange -ay realmente cria o nó do dispositivo raiz (após um longo atraso), mas nunca sai, deixando que eu o mate manualmente. Quando isso acontece eu tento montar o dispositivo, mas ele sempre trava indefinidamente. Observe que, enquanto os dois comandos estão em execução, o console exibe um monte de mensagens sobre o comando com falha "READ DMA" ou algo assim.

Eu corri smartctl -a /dev/sda algumas vezes. Cada vez que dá uma quantidade razoável de erros sobre blocos ruins (IIRC), mas finalmente diz que a unidade está em boas condições.

Eu coloquei um pastebin de dmesg na máquina afetada. Os logs vêm da inicialização de um live CD do Arch Linux e, em seguida, são executados pvscan; vgscan; vgchange -ay . vgchange -ay ficou pendurado para sempre dessa vez e eu acabei matando. Aqui está o fim de dmesg , para a posteridade (e, portanto, eu não uso um pastebin 2 ):

[   46.332920] end_request: I/O error, dev fd0, sector 0
[   58.503496] end_request: I/O error, dev fd0, sector 0
[167992.304649] EXT4-fs (sda1): recovery complete
[167992.304660] EXT4-fs (sda1): mounted filesystem with ordered data mode. Opts: (null)
[168092.874016] EXT4-fs (sda1): mounted filesystem with ordered data mode. Opts: (null)
[168163.318923] EXT4-fs (sda1): mounted filesystem with ordered data mode. Opts: (null)
[168459.839738] end_request: I/O error, dev fd0, sector 0
[168472.010337] end_request: I/O error, dev fd0, sector 0
[168614.642035] bio: create slab <bio-2> at 2
[168630.045526] ata1.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x0
[168630.045649] ata1.00: BMDMA stat 0x65
[168630.045710] ata1.00: failed command: READ DMA
[168630.045787] ata1.00: cmd c8/00:08:00:10:10/00:00:00:00:00/e6 tag 0 dma 4096 in
         res 51/40:08:00:10:10/00:00:00:00:00/e6 Emask 0x9 (media error)
[168630.046006] ata1.00: status: { DRDY ERR }
[168630.046071] ata1.00: error: { UNC }
[168630.066286] ata1.00: configured for UDMA/100
[168630.079493] ata1.01: configured for UDMA/66
[168630.079514] sd 0:0:0:0: [sda] Unhandled sense code
[168630.079517] sd 0:0:0:0: [sda]  
[168630.079520] Result: hostbyte=0x00 driverbyte=0x08
[168630.079523] sd 0:0:0:0: [sda]  
[168630.079525] Sense Key : 0x3 [current] [descriptor]
[168630.079530] Descriptor sense data with sense descriptors (in hex):
[168630.079532]         72 03 11 04 00 00 00 0c 00 0a 80 00 00 00 00 00 
[168630.079544]         06 10 10 00 
[168630.079549] sd 0:0:0:0: [sda]  
[168630.079551] ASC=0x11 ASCQ=0x4
[168630.079554] sd 0:0:0:0: [sda] CDB: 
[168630.079556] cdb[0]=0x28: 28 00 06 10 10 00 00 00 08 00
[168630.079567] end_request: I/O error, dev sda, sector 101715968
[168630.079665] Buffer I/O error on device dm-3, logical block 0
[168630.079775] ata1: EH complete
[168634.564062] ata1.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x0
[168634.564165] ata1.00: BMDMA stat 0x64
[168634.564225] ata1.00: failed command: READ DMA
[168634.564301] ata1.00: cmd c8/00:08:80:0f:10/00:00:00:00:00/e6 tag 0 dma 4096 in
         res 51/10:00:83:0f:10/00:00:00:00:00/e6 Emask 0x81 (invalid argument)
[168634.564527] ata1.00: status: { DRDY ERR }
[168634.564592] ata1.00: error: { IDNF }
[168634.584336] ata1.00: configured for UDMA/100
[168634.597559] ata1.01: configured for UDMA/66
[168634.597578] ata1: EH complete
[168639.087353] ata1.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x0
[168639.087462] ata1.00: BMDMA stat 0x64
[168639.087521] ata1.00: failed command: READ DMA
[168639.087596] ata1.00: cmd c8/00:08:80:0f:10/00:00:00:00:00/e6 tag 0 dma 4096 in
         res 51/10:00:83:0f:10/00:00:00:00:00/e6 Emask 0x81 (invalid argument)
[168639.087822] ata1.00: status: { DRDY ERR }
[168639.087886] ata1.00: error: { IDNF }
[168639.105791] ata1.00: configured for UDMA/100
[168639.118999] ata1.01: configured for UDMA/66
[168639.119017] ata1: EH complete
[168645.896986] ata1.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x0
[168645.897095] ata1.00: BMDMA stat 0x64
[168645.897155] ata1.00: failed command: READ DMA
[168645.900373] ata1.00: cmd c8/00:08:80:0f:10/00:00:00:00:00/e6 tag 0 dma 4096 in
         res 51/40:00:83:0f:10/00:00:00:00:00/e6 Emask 0x9 (media error)
[168645.906936] ata1.00: status: { DRDY ERR }
[168645.910263] ata1.00: error: { UNC }
[168645.931315] ata1.00: configured for UDMA/100
[168645.944504] ata1.01: configured for UDMA/66
[168645.944525] sd 0:0:0:0: [sda] Unhandled sense code
[168645.944529] sd 0:0:0:0: [sda]  
[168645.944531] Result: hostbyte=0x00 driverbyte=0x08
[168645.944534] sd 0:0:0:0: [sda]  
[168645.944537] Sense Key : 0x3 [current] [descriptor]
[168645.944541] Descriptor sense data with sense descriptors (in hex):
[168645.944543]         72 03 11 04 00 00 00 0c 00 0a 80 00 00 00 00 00 
[168645.944554]         06 10 0f 83 
[168645.944559] sd 0:0:0:0: [sda]  
[168645.944561] ASC=0x11 ASCQ=0x4
[168645.944564] sd 0:0:0:0: [sda] CDB: 
[168645.944566] cdb[0]=0x28: 28 00 06 10 0f 80 00 00 08 00
[168645.944578] end_request: I/O error, dev sda, sector 101715843
[168645.947946] Buffer I/O error on device dm-2, logical block 10485744
[168645.951439] ata1: EH complete
[168650.445911] ata1.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x0
[168650.449275] ata1.00: BMDMA stat 0x65
[168650.452579] ata1.00: failed command: READ DMA
[168650.455873] ata1.00: cmd c8/00:08:00:10:10/00:00:00:00:00/e6 tag 0 dma 4096 in
         res 51/40:08:00:10:10/00:00:00:00:00/e6 Emask 0x9 (media error)
[168650.462537] ata1.00: status: { DRDY ERR }
[168650.465714] ata1.00: error: { UNC }
[168650.486063] ata1.00: configured for UDMA/100
[168650.499326] ata1.01: configured for UDMA/66
[168650.499344] sd 0:0:0:0: [sda] Unhandled sense code
[168650.499348] sd 0:0:0:0: [sda]  
[168650.499350] Result: hostbyte=0x00 driverbyte=0x08
[168650.499353] sd 0:0:0:0: [sda]  
[168650.499355] Sense Key : 0x3 [current] [descriptor]
[168650.499360] Descriptor sense data with sense descriptors (in hex):
[168650.499362]         72 03 11 04 00 00 00 0c 00 0a 80 00 00 00 00 00 
[168650.499373]         06 10 10 00 
[168650.499378] sd 0:0:0:0: [sda]  
[168650.499380] ASC=0x11 ASCQ=0x4
[168650.499383] sd 0:0:0:0: [sda] CDB: 
[168650.499385] cdb[0]=0x28: 28 00 06 10 10 00 00 00 08 00
[168650.499396] end_request: I/O error, dev sda, sector 101715968
[168650.502757] Buffer I/O error on device dm-3, logical block 0
[168650.506189] ata1: EH complete
[168798.816025] usb 9-2: new high-speed USB device number 2 using ehci-pci

Este é apenas o final do log, onde os erros começaram, porque eu atingi o limite de postagem. Para ver a coisa toda, olhe para o pastebin.

Desculpas por não fornecer informações específicas, mas não estou na frente da caixa afetada no momento.

    
por strugee 06.10.2013 / 05:41

1 resposta

2

A partir das informações extras que você forneceu, parece que você tem um mau drive (blocos defeituosos). Você pode tentar resolver esses problemas se quiser, mas eu consideraria seriamente substituir a unidade.

Se você quer contornar o problema, basicamente você terá que encontrar as extensões físicas do LVM que ficam sobre os blocos defeituosos, e adicionar essas extensões físicas a um volume lógico que não deve ser usado. / p>

Na verdade, existe uma corrente de e-mails bastante recente na lista de discussão linux-lvm sobre esse assunto (eu li toda a cadeia, ela contém muitas informações):

link

Nesta mensagem específica, parece que alguém criou um script python para ajudar na tarefa:

link

Having helped people in such situations (where internet at least was working), I've used the attached script to help find affected LVs and files.

#!/usr/bin/python
# Identify partition, LV, file containing a sector 

# Copyright (C) 2010,2012 Stuart D. Gathman
# Shared under GNU Public License v2 or later
#   This program is free software; you can redistribute it and/or modify
#   it under the terms of the GNU General Public License as published by
#   the Free Software Foundation; either version 2 of the License, or
#   (at your option) any later version.

#   This program is distributed in the hope that it will be useful,
#   but WITHOUT ANY WARRANTY; without even the implied warranty of
#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
#   GNU General Public License for more details.

#   You should have received a copy of the GNU General Public License along
#   with this program; if not, write to the Free Software Foundation, Inc.,
#   51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.

import sys
from subprocess import Popen,PIPE

ID_LVM = 0x8e
ID_LINUX = 0x83
ID_EXT = 0x05
ID_RAID = 0xfd

def idtoname(id):
  if id == ID_LVM: return "Linux LVM"
  if id == ID_LINUX: return "Linux Filesystem"
  if id == ID_EXT: return "Extended Partition"
  if id == ID_RAID: return "Software RAID"
  return hex(id)

class Segment(object):
  __slots__ = ('pe1st','pelst','lvpath','le1st','lelst')
  def __init__(self,pe1st,pelst):
    self.pe1st = pe1st;
    self.pelst = pelst;
  def __str__(self):
    return "Seg:%d-%d:%s:%d-%d" % (
      self.pe1st,self.pelst,self.lvpath,self.le1st,self.lelst)

def cmdoutput(cmd):
  p = Popen(cmd, shell=True, stdout=PIPE)
  try:
    for ln in p.stdout:
      yield ln
  finally:
    p.stdout.close()
    p.wait()

def icheck(fs,blk):
  "Return inum from block number, or 0 if free space."
  for ln in cmdoutput("debugfs -R 'icheck %d' '%s' 2>/dev/null"%(blk,fs)):
    b,i = ln.strip().split(None,1)
    if not b[0].isdigit(): continue
    if int(b) == blk:
      if i.startswith('<'):
    return 0
      return int(i)
  raise ValueError('%s: invalid block: %d'%(fs,blk))

def ncheck(fs,inum):
  "Return filename from inode number, or None if not linked."
  for ln in cmdoutput("debugfs -R 'ncheck %d' '%s' 2>/dev/null"%(inum,fs)):
    i,n = ln.strip().split(None,1)
    if not i[0].isdigit(): continue
    if int(i) == inum:
      return n
  return None

def blkid(fs):
  "Return dictionary of block device attributes"
  d = {}
  for ln in cmdoutput("blkid -o export '%s'"%fs):
    k,v = ln.strip().split('=',1)
    d[k] = v
  return d

def getpvmap(pv):
  pe_start = 192 * 2
  pe_size = None
  seg = None
  segs = []
  for ln in cmdoutput("pvdisplay --units k -m %s"%pv):
    a = ln.strip().split()
    if not a: continue
    if a[0] == 'Physical' and a[4].endswith(':'):
      pe1st = int(a[2])
      pelst = int(a[4][:-1])
      seg = Segment(pe1st,pelst)
    elif seg and a[0] == 'Logical':
      if a[1] == 'volume':
    seg.lvpath = a[2]
      elif a[1] == 'extents':
    seg.le1st = int(a[2])
    seg.lelst = int(a[4])
    segs.append(seg)
    elif a[0] == 'PE' and a[1] == 'Size':
      if a[2] == "(KByte)":
    pe_size = int(a[3]) * 2
      elif a[3] == 'KiB':
    pe_size = int(float(a[2])) * 2
  if segs:
    for ln in cmdoutput("pvs --units k -o+pe_start %s"%pv):
      a = ln.split()
      if a[0] == pv:
        lst = a[-1]
    if lst.lower().endswith('k'):
      pe_start = int(float(lst[:-1]))*2
      return pe_start,pe_size,segs
  return None

def findlv(pv,sect):
  res = getpvmap(pv)
  if not res: return None
  pe_start,pe_size,m = res
  if sect < pe_start:
    raise Exception("Bad sector in PV metadata area")
  pe = int((sect - pe_start)/pe_size)
  pebeg = pe * pe_size + pe_start
  peoff = sect - pebeg
  for s in m:
    if s.pe1st <= pe <= s.pelst:
      le = s.le1st + pe - s.pe1st
      return s.lvpath,le * pe_size + peoff

def getmdmap():
  with open('/proc/mdstat','rt') as fp:
    m = []
    for ln in fp:
      if ln.startswith('md'):
    a = ln.split(':')
    raid = a[0].strip()
    devs = []
    a = a[1].split()
    for d in a[2:]:
      devs.append(d.split('[')[0])
    m.append((raid,devs))
    return m

def parse_sfdisk(s):
  for ln in s:
    try:
      part,desc = ln.split(':')
      if part.startswith('/dev/'):
        d = {}
        for p in desc.split(','):
      name,val = p.split('=')
      name = name.strip()
      if name.lower() == 'id':
        d[name] = int(val,16)
      else:
        d[name] = int(val)
    yield part.strip(),d
    except ValueError:
      continue

def findpart(wd,lba):
  s = cmdoutput("sfdisk -d %s"%wd)
  parts = [ (part,d['start'],d['size'],d['Id']) for part,d in parse_sfdisk(s) ]
  for part,start,sz,Id in parts:
    if Id == ID_EXT: continue
    if start <= lba < start + sz:
      return part,lba - start,Id
  return None

if __name__ == '__main__':
  wd = sys.argv[1]
  lba = int(sys.argv[2])
  print wd,lba,"Whole Disk"
  res = findpart(wd,lba)
  if not res:
    print "LBA is outside any partition"
    sys.exit(1)
  part,sect,Id = res
  print part,sect,idtoname(Id)
  if Id == ID_LVM:
    bd,sect = findlv(part,sect)
    # FIXME: problems if LV is snapshot
  elif Id == ID_LINUX:
    bd = part
  else:
    if Id == ID_RAID:
      for md,devs in getmdmap():
    for dev in devs:
      if part == "/dev/"+dev:
        part = "/dev/"+md
        break
        else: continue
    break
    res = findlv(part,sect)
    if res:
      print "PV =",part
      bd,sect = res
    else:
      bd = part
  blksiz = 4096
  blk = int(sect * 512 / blksiz)
  p = blkid(bd)
  try:
    t = p['TYPE']
  except:
    print bd,p
    raise
  print "fs=%s block=%d %s"%(bd,blk,t)
  if t.startswith('ext'):
    inum = icheck(bd,blk)
    if inum:
      fn = ncheck(bd,inum)
      print "file=%s inum=%d"%(fn,inum)
    else:
      print "<free space>"
    
por 06.10.2013 / 08:12

Tags