Unix/Linux Go Back    


HP-UX HP-UX (Hewlett Packard UniX) is Hewlett-Packard's proprietary implementation of the Unix operating system, based on System V.

Bypass stale PE ?

HP-UX


Closed    
 
Thread Tools Search this Thread Display Modes
    #8  
Old Unix and Linux 06-10-2014   -   Original Discussion by port43
vbe's Unix or Linux Image
vbe vbe is offline Forum Staff  
Moderator
 
Join Date: Sep 2005
Last Activity: 17 June 2018, 9:38 AM EDT
Location: Switzerland - GE
Posts: 6,544
Thanks: 308
Thanked 563 Times in 525 Posts
From what I see you are in mirror using 2 controllers : 0/10/0... and 1/10/0...
If you look, all your LV are like :

Code:
   Current LE                  256
   Allocated PE                512

PE is twice the value of LE ... except the one giving you trouble... only Im a bit stressed now at work, will try to finish rapidly and if no one else had helped you, join back from my laptop at home...

All the best
Sponsored Links
    #9  
Old Unix and Linux 06-10-2014   -   Original Discussion by port43
port43's Unix or Linux Image
port43 port43 is offline
Registered User
 
Join Date: Aug 2010
Last Activity: 12 June 2018, 10:50 AM EDT
Location: Detroit Suburb
Posts: 70
Thanks: 17
Thanked 12 Times in 12 Posts
Ah, yes, there are three disks in use on each controller. Whoever initially set this up wasn't thinking and didn't keep the mirrors on distinct controllers since some volumes have data on both sides but the mirror on the other. For instance:


Code:
 # lvdisplay -v /dev/vg00/lvol5
--- Logical volumes ---
LV Name                     /dev/vg00/lvol5
VG Name                     /dev/vg00
LV Permission               read/write
LV Status                   available/stale
Mirror copies               1
Consistency Recovery        MWC
Schedule                    parallel
LV Size (Mbytes)            3584
Current LE                  896
Allocated PE                1792
Stripes                     0
Stripe Size (Kbytes)        0
Bad block                   on
Allocation                  strict
IO Timeout (Seconds)        default

   --- Distribution of logical volume ---
   PV Name            LE on PV  PE on PV
   /dev/dsk/c5t5d0    583       583
   /dev/dsk/c9t3d0    313       313
   /dev/dsk/c9t5d0    896       896

   --- Logical extents ---
   LE    PV1                PE1   Status 1 PV2                PE2   Status 2
   00000 /dev/dsk/c5t5d0    00818 current  /dev/dsk/c9t5d0    00818 current
   00001 /dev/dsk/c5t5d0    00819 current  /dev/dsk/c9t5d0    00819 current
   00002 /dev/dsk/c5t5d0    00820 current  /dev/dsk/c9t5d0    00820 current
   00003 /dev/dsk/c5t5d0    00821 current  /dev/dsk/c9t5d0    00821 current
   ... (more of the same until)
   00579 /dev/dsk/c5t5d0    02165 current  /dev/dsk/c9t5d0    01397 current
   00580 /dev/dsk/c5t5d0    02166 current  /dev/dsk/c9t5d0    01398 current
   00581 /dev/dsk/c5t5d0    02167 current  /dev/dsk/c9t5d0    01399 current
   00582 /dev/dsk/c5t5d0    02168 current  /dev/dsk/c9t5d0    01400 current
   00583 /dev/dsk/c9t3d0    00000 current  /dev/dsk/c9t5d0    01401 current
   00584 /dev/dsk/c9t3d0    00001 current  /dev/dsk/c9t5d0    01402 current
   00585 /dev/dsk/c9t3d0    00002 current  /dev/dsk/c9t5d0    01403 current
   00586 /dev/dsk/c9t3d0    00003 current  /dev/dsk/c9t5d0    01404 current
   ... (and still more up to the stale PE)
   00650 /dev/dsk/c9t3d0    00067 current  /dev/dsk/c9t5d0    01468 current
   00651 /dev/dsk/c9t3d0    00068 current  /dev/dsk/c9t5d0    01469 current
   00652 /dev/dsk/c9t3d0    00069 current  /dev/dsk/c9t5d0    01470 stale
   00653 /dev/dsk/c9t3d0    00070 current  /dev/dsk/c9t5d0    01471 current
   00654 /dev/dsk/c9t3d0    00071 current  /dev/dsk/c9t5d0    01472 current
   ... (and on until the bottom)
   00893 /dev/dsk/c9t3d0    00710 current  /dev/dsk/c9t5d0    01711 current
   00894 /dev/dsk/c9t3d0    00711 current  /dev/dsk/c9t5d0    01712 current
   00895 /dev/dsk/c9t3d0    00712 current  /dev/dsk/c9t5d0    01713 current

So if I lose controller 1/10/0 I lose lvol5. But that's a whole other issue.
Sponsored Links
    #10  
Old Unix and Linux 06-10-2014   -   Original Discussion by port43
vbe's Unix or Linux Image
vbe vbe is offline Forum Staff  
Moderator
 
Join Date: Sep 2005
Last Activity: 17 June 2018, 9:38 AM EDT
Location: Switzerland - GE
Posts: 6,544
Thanks: 308
Thanked 563 Times in 525 Posts
I wonder if its not because the strict allocation, that is not truly suitable when using 3 disks… Doesn't pvdisplay show anything?
    #11  
Old Unix and Linux 06-11-2014   -   Original Discussion by port43
port43's Unix or Linux Image
port43 port43 is offline
Registered User
 
Join Date: Aug 2010
Last Activity: 12 June 2018, 10:50 AM EDT
Location: Detroit Suburb
Posts: 70
Thanks: 17
Thanked 12 Times in 12 Posts
The pvdisplay is on the first page of this thread and looks fine to me, except for the one stale part.

I looked into the fact that it has a "strict" allocation policy and from what I glean from the lvchange command documentation, that just means that the mirror copies of extents cannot share the same physical volume, which makes sense because you wouldn't want your data mirrored on the same disk because if it failed ...



Code:
         -s strict                Set the strict allocation policy.
                                    Mirror copies of a logical extent can be
                                    allocated to share or not share the same
                                    physical volume or physical volume
                                    group.  This option only makes sense
                                    when the physical volumes of the volume
                                    group that owns the specified logical
                                    volume reside on different physical
                                    disks.  strict can have one of the
                                    following values:

                                    y    Set a strict allocation policy.
                                         Mirrors of a logical extent cannot
                                         share the same physical volume.

                                    g    Set a PVG-strict allocation policy.
                                         Mirrors of a logical extent cannot
                                         share the same physical volume
                                         group.

                                    n    Do not set a strict or a PVG-strict
                                         allocation policy.  Mirrors of a
                                         logical extent can share the same
                                         physical volume.

                                    When a logical volume is mirrored, the
                                    following changes are not allowed:

                                    +    From nonstrict to strict
                                    +    From nonstrict to PVG-strict
                                    +    From strict to PVG-strict

I thought I was onto something when I saw the -C contiguous option so I could skip the bad extent but that's not available for the root volume group.
Sponsored Links
    #12  
Old Unix and Linux 06-11-2014   -   Original Discussion by port43
vbe's Unix or Linux Image
vbe vbe is offline Forum Staff  
Moderator
 
Join Date: Sep 2005
Last Activity: 17 June 2018, 9:38 AM EDT
Location: Switzerland - GE
Posts: 6,544
Thanks: 308
Thanked 563 Times in 525 Posts
I never had stale on HP-UX... (cant say that of the aix's I was in charge...) So I am trying to interpolate from my bad AIX experience... About Contiguous - You do have some... HP-UX11.11 have in vg00 3 Lvols that use contiguous: the swap, root and stand, but its not the case here, how did you change disk? (what commands were passed...) because I think I did have something of the sort with a D class when I forgot to do a ??? forgot.. trying hard to remember you can help since I have no more documentation, you should do a ioscan ( full in doubt, not funC...) to update devs but there is something else I wonder if its not insf -e can you check what that does? because my idea would be to remove the disk if you are sure it is good with no bad blocks etc.. the rm all associated files /dev/rdsk/c.... then add the disk do an insf if its the right command and see if that changes after mirroring again
Sponsored Links
    #13  
Old Unix and Linux 06-11-2014   -   Original Discussion by port43
port43's Unix or Linux Image
port43 port43 is offline
Registered User
 
Join Date: Aug 2010
Last Activity: 12 June 2018, 10:50 AM EDT
Location: Detroit Suburb
Posts: 70
Thanks: 17
Thanked 12 Times in 12 Posts
I always use the HP doc ""When Good Disks Go Bad"
and my steps boil down to:
  1. lvreduce
  2. vgreduce
  3. mv /etc/lvmtab
  4. pull disk
  5. rmsf -H 1/10/0/0.5.0
  6. put in new disk
  7. ioscan to verify disk is there
  8. insf
  9. vgscan
  10. pvcreate
  11. vgextend
  12. lvextend

I'm going to call HP and see if I can't beg my way to some help.
Sponsored Links
    #14  
Old Unix and Linux 06-11-2014   -   Original Discussion by port43
vbe's Unix or Linux Image
vbe vbe is offline Forum Staff  
Moderator
 
Join Date: Sep 2005
Last Activity: 17 June 2018, 9:38 AM EDT
Location: Switzerland - GE
Posts: 6,544
Thanks: 308
Thanked 563 Times in 525 Posts
You do have a disk that is not like the others: c5t5d0 which has 2169 instead of 2170 so there is a bad block there and I wonder if that is not what is giving trouble
Sponsored Links
Closed

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Linux More UNIX and Linux Forum Topics You Might Find Helpful
Thread Thread Starter Forum Replies Last Post
Stale lock installed for pkgadd poyato Solaris 2 04-05-2012 07:57 AM
Stale users and no process LisaS HP-UX 8 10-06-2010 01:29 PM
how to remove the stale entry without a reboot?? navadeep AIX 6 03-26-2009 10:53 PM
removing stale partitions karthikosu UNIX for Dummies Questions & Answers 2 11-21-2006 02:54 AM
Unix Stale Mounts derf912 UNIX for Dummies Questions & Answers 2 08-01-2006 02:15 PM



All times are GMT -4. The time now is 07:58 PM.