Re: errors on shadow sets and their members

From: Michael Moroney (moroney_at_world.std.spaamtrap.com)
Date: 04/25/04 

Date: Sun, 25 Apr 2004 21:46:51 +0000 (UTC)

helbig@astro.multiCLOTHESvax.de (Phillip Helbig---remove CLOTHES to reply) writes:

>I've rearranged some hardware in my hobbyist cluster. In one BA353
>there is an RZ26 which I can't remove, so I'm "forced" to use it. It is
>a member of a shadow set (the other member is on another node). During
>SHADOW MERGE or COPY operations, it collects 10--15 errors, but only on
>the node it has a direct connection to, not to all the other nodes, to
>which it is MSCP served. The shadow set is mounted by all nodes and
>shows no errors.

>What does this mean?

You have to look at the error log to see what the error really is.

One possibility is a parity error on a disk block. DKDRIVER will log an
error whenever it attempts to read a block with a parity error.

How did the error get there? It could mean bits flipped (mega badness)
but due to how shadowing works, it may have been deliberately put there
by shadowing (perhaps the corresponding block on the other drive was
bad at one time, and this was "copied" during a copy operation in the
past).

You may want to track down which files contain the bad blocks, if any.
$ ANALYZE/DISK/READ will read every block of every file but if you do it
to the shadowset it may go to the wrong member and miss the error blocks.
You can do it to the individual member if you are willing to break up
the shadowset and endure a copy when putting it back in (do you have
a version of VMS with minicopy?) 

-- 
-Mike

Relevant Pages

  • Re: SHADDETINCON, SHADOWING detects inconsistent state
    ... Each system has a 2- or 3-member shadow set as its system disk. ... sets distributed among the nodes (each member has a direct connection to ... long as the spontaneous reboots are happening. ...
    (comp.os.vms)
  • Re: Shadow set problem finally solved
    ... pure-vanilla problem of a shadow set having a forced ... error that VMS replicates on every shadow copy, ... Break out a member. ...
    (comp.os.vms)
  • Re: VMS analogue of FBSD and linux hier(7) man pages
    ... x ROSRVC x VMS V7.3-2 x MEMBER xmqqqqqqqqqqqqqqqqqqqj ... a full shadow copy. ...
    (comp.os.vms)
  • Re: Shadow set problem finally solved
    ... pure-vanilla problem of a shadow set having a forced ... you have to allocate the block somehow. ... init/erase a target drive, then back up to it from your master copy]. ... Break out a member. ...
    (comp.os.vms)
  • Re: Shadow set problem finally solved
    ... pure-vanilla problem of a shadow set having a forced ...  Do an back/image on the member ... and if not a boot disk the following [after being certain nothing will be ... My temp file technique solves the problem outright. ...
    (comp.os.vms)