[prev in list] [next in list] [prev in thread] [next in thread]
List: evms-devel
Subject: Re: [Evms-devel] evms_clean...where's the metadata please?
From: Steve Dobbelstein <steved () us ! ibm ! com>
Date: 2007-03-27 14:59:06
Message-ID: OFD93A2E57.B3E4E7E8-ON052572AB.0050FCF7-052572AB.00525111 () us ! ibm ! com
[Download RAW message or body]
[Attachment #2 (multipart/alternative)]
thomas62186218@aol.com wrote on 03/26/2007 10:48:35 PM:
> Hi there,
>
> Having some issues here that appear to be related to left-over
> metadata that is not getting cleaned out when deleting RAID groups.
> It occurs most frequently when using "nested" RAID groups using
> md...for example, creating a RAID 10 using RAID 0 across multiple
> RAID 1's. After deleting the RAID groups and regions, I run
> evms_clean to wipe them further,
When you delete the RAID groups and regions the metadata would be wiped=
out. When you run evms_clean it should not find any RAID metadata to w=
ipe
out.
> but if I create new RAID groups
> with these same hard drives, EVMS 2.5.5 almost always reports a
> degraded state from the get-go and some other incorrect information.
>
> I know evms_clean is undocumented, but is there source code
> available?
steved@steved:~/evms/engine2> find . -name evms_clean\*
./tests/evms_clean.c
> Better yet, can you tell me where (what sector range
> perhaps) on a hard drive the metadata is being stored? I'd be happy
> to write a script that nukes those sectors but I need some guidance
> here. Thanks!
The RAID metadata are stored near the end of the object. I believe its=
location is something like "take the size of the object, round down to =
the
next 64 KB boundary, then back up another 64 KB and put the metadata
there." See plugins/md/md_super.h for the real calculations.=
[Attachment #5 (text/html)]
<html><body>
<p>thomas62186218@aol.com<tt> wrote on 03/26/2007 10:48:35 PM:<br>=
<br>
> Hi there,<br>
> <br>
> Having some issues here that appear to be related to left-over <br=
>
> metadata that is not getting cleaned out when deleting RAID groups=
. <br>
> It occurs most frequently when using "nested" RAID group=
s using <br>
> md...for example, creating a RAID 10 using RAID 0 across multiple =
<br>
> RAID 1's. After deleting the RAID groups and regions, I run <br>
> evms_clean to wipe them further,</tt><br>
<br>
<tt>When you delete the RAID groups and regions the metadata would be w=
iped out. When you run evms_clean it should not find any RAID met=
adata to wipe out.</tt><br>
<br>
<tt>> but if I create new RAID groups <br>
> with these same hard drives, EVMS 2.5.5 almost always reports a <b=
r>
> degraded state from the get-go and some other incorrect informatio=
n. <br>
> <br>
> I know evms_clean is undocumented, but is there source code <br>
> available?</tt><br>
<br>
<tt>steved@steved:~/evms/engine2> find . -name evms_clean\*</tt><br>=
<tt>./tests/evms_clean.c</tt><br>
<br>
<tt>> Better yet, can you tell me where (what sector range <br>
> perhaps) on a hard drive the metadata is being stored? I'd be happ=
y <br>
> to write a script that nukes those sectors but I need some guidanc=
e <br>
> here. Thanks!<br>
</tt><br>
<tt>The RAID metadata are stored near the end of the object. I be=
lieve its location is something like "take the size of the object,=
round down to the next 64 KB boundary, then back up another 64 KB and =
put the metadata there." See plugins/md/md_super.h for the r=
eal calculations.</tt></body></html>=
-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
Evms-devel mailing list
Evms-devel@lists.sourceforge.net
To subscribe/unsubscribe, please visit:
https://lists.sourceforge.net/lists/listinfo/evms-devel
[prev in list] [next in list] [prev in thread] [next in thread]
Configure |
About |
News |
Add a list |
Sponsored by KoreLogic