Re: Does solaris require disks to be zeroed before using

From: David A.Lethe (davidATsantools.com)
Date: 09/25/04 

Date: Fri, 24 Sep 2004 17:20:42 -0500

On Fri, 24 Sep 2004 21:22:49 +0000, Beardy <beardy@beardy.net> wrote:

>David A.Lethe wrote:
>> On 23 Sep 2004 22:31:46 GMT, Scott Howard <scott@hunterlink.net.au>
>> wrote:
>>
>>
>>>David A.Lethe <davidATsantools.com> wrote:
>>>
>>>>This is causing some issues at our company, and we can't get any
>>>>definitive answers. Is there any requirement, or even best-practices
>>>>policies that SCSI and/or fibre channel disk drives need to have all
>>>>zeros written to them before partitioning and building a file system
>>>>on them?
>>>
>>>In a word, no.
>>>
>>>Solaris will never attempt to read a block of data which it hasn't
>>>previously written something to.
>>>
>>> Scott
>>
>>
>> Not true, and this explains the whole point of the question.
>>
>> Example .. .when you run the format command. Obviously it doesn't
>> write before reading, or every time you ran it, it would blow the
>> partitions away ;)
>
>OK format will check for the existence of a valid label on the disk
>first, and if not present, allow you to label the disk at that point.
>This is an exception, and I think the OP was referring to whether data
>blocks on disk should be zeroed, which they need not be; except in the
>unusual case that Casper referred to.

so data need not be zeroed except in unusual cases where it needs to
be zeroed?

So Scott seems to be saying solaris does a read-before-write on a
block? That couldn't possibly be correct either.
>
>> This brings up the reason for the question to begin with. The test
>> suites don't wipe the disks clean, and we had what might have been a
>> zillion-to-one shot where fdisk (on LINUX) exited with an invalid
>> partition error when the sysadmin tried to partition things.
>
>Your case is not relevant to the OP's question. The fact that the test
>suites may leave patterns on the disk is still not a problem. A random
>pattern and a regular pattern are identical if Solaris doesn't know they
>are there.

As I am the OP, I think I have the right to decide what is relevent ;)
The test suites leaving info on the disks IS relevent. Does solaris
require, or is it even best practice to have a disk zeroed before
using it? I couldn't see anything online that said one way or
another, only that disks must be formatted before using.

>
>> After I figured out problem wasn't hardware related, and using dd to
>> blow the first few MB of the disk away, we have to deal with problem
>> prevention ... do we adjust testing utilties to insure that all or
>> part of the disk is zeroed before we let people use the disks.
>
>Do what you feel is correct on Linux. This is not relevant to the OP's
>question, and to the correct answers that have been provided to him.

Relevant Pages

  • Re: Exchange 2003 Transaction Log Performance
    ... Instead of one thread writing, ... The IO pattern will ... The seek time is the major component of disk access time. ... order to achieve a purely sequential workload, you'd need to be able to ...
    (microsoft.public.exchange.design)
  • Re: Exchange 2003 Transaction Log Performance
    ... the storage guru has spoken. ... continue the pattern, we not that when compared to a purely sequential ... The seek time is the major component of disk access time. ... In order to achieve a purely sequential workload, you'd need to be able to ...
    (microsoft.public.exchange.design)
  • Re: "secure" file flag?
    ... > this could be done on in the kernel, ... The DOD has a specific pattern it requirs, ... > likely to be effective against a dedicated inspection of the disk; ... you really want to erase. ...
    (freebsd-hackers)
  • Re: Does solaris require disks to be zeroed before using
    ... first, and if not present, allow you to label the disk at that point. ... > partition error when the sysadmin tried to partition things. ... pattern and a regular pattern are identical if Solaris doesn't know they ...
    (comp.sys.sun.admin)
  • Re: Design help...an explosion of interfaces......
    ... > That pattern is a beautiful OO pattern that is useful in many other places. ... or serial port then typically on through I/O filters doing ... logical layers above the disk, ...
    (comp.object)