On Oct 13, 2005 20:31 -0600, Kumaran Rajaram wrote:> i) In Lustre OSTs, 400MB is used to storage journal data and ldiskfs > reserves additional 5% for root. Is it recommended to bring down this > reserved space to say 1% or 0% using tune2fs ? This would make more disk > capacity available to the users. Also, on large storage systems 5% would > mean a significant storage lost. If it is recommended, can it be specified > using config.xml or we need to manually tune2fs after OST is formatted.These are FAQs: https://bugzilla.lustre.org/show_bug.cgi?id=2378 How to reclaim the 5% of disk space reserved for root https://bugzilla.lustre.org/show_bug.cgi?id=2381 Why is so much disk space used, when I just reformatted? There are many FAQs answered in the Knowledge Base: https://bugzilla.lustre.org/buglist.cgi?product=KB> ii) As I recall in Lustre 1.0, df reports space available based a simple > quick formula: OST with least space available * Number of OSTs > > Please correct if I am wrong. We seem to see this df behavior in > 1.4.x as well although stripe count defaults to 1. The above formula can > be misleading sometimes if a particular OST is full or if you add new > OSTs with some old OSTs almost full. Are there plans to refine this ? > Currently, du -sh on Lustre FS would reflect the actual Lustre-capacity > used.Actually, I don''t think this has ever been the case. The free space (in blocks) has always been the aggregate of the total amount of space in the filesystem. In some cases this is misleading, as a full OST may cause ENOSPC writing to a file even when there is free space on other OSTs. New files would avoid that full OST so it isn''t necessarily fatal. Cheers, Andreas -- Andreas Dilger Principal Software Engineer Cluster File Systems, Inc.
This is a MIME message. If you are reading this text, you may want to
consider changing to a mail reader or gateway that understands how to
properly handle MIME multipart messages.
--=__Part0022017C.0__Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: quoted-printable
Hi,=20
=20
Couple of questions:=20
=20
i) In Lustre OSTs, 400MB is used to storage journal data and ldiskfs reserves
additional 5% for root. Is it recommended to bring down this reserved space to
say 1% or 0% using tune2fs ? This would make more disk capacity available to the
users. Also, on large storage systems 5% would mean a significant storage lost.
If it is recommended, can it be specified using config.xml or we need to
manually tune2fs after OST is formatted.=20
=20
ii) As I recall in Lustre 1.0, df reports space available based a simple quick
formula:=20
OST with least space available * Number of OSTs=20
=20
Please correct if Iam wrong. We seem to see this df behavior in 1.4.x as well
although stripe count defaults to 1. The above formula can be misleading
sometimes if a particular OST is full or if you add new OSTs with some old OSTs
almost full. Are there plans to refine this ?=20
Currently, du -sh on Lustre FS would reflect the actual Lustre-capacity used.=20
=20
Thanks,=20
-Kums=20
--=__Part0022017C.0__Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable
Content-Description: HTML
<html>
<head>
<style type=3D"text/css">
<!--
body { margin-bottom: 1px; font-variant: normal; margin-left: 4px;
line-height: normal; margin-right: 4px; margin-top: 4px }
-->
</style>
=20
</head>
<body>
<DIV>
Hi,
</DIV>
<DIV> </DIV>
<DIV>
Couple of questions:
</DIV>
<DIV> </DIV>
<DIV>
i) In Lustre OSTs, 400MB is used to storage journal data
and ldiskfs reserves additional 5% for root. Is it recommended to bring
down this reserved space to say 1% or 0% using tune2fs ?
This would make more disk capacity available to the users. Also, on
large storage systems 5% would mean a significant storage lost. If it is
recommended, can it be specified using config.xml or we need to manually
tune2fs after OST is formatted.
</DIV>
<DIV> </DIV>
<DIV>
ii) As I recall in Lustre 1.0, "df"
reports space available based a simple quick formula:
</DIV>
<DIV>
OST with least space available * Number of OSTs
</DIV>
<DIV> </DIV>
<DIV>
Please correct if Iam wrong. We seem to see this df behavior in 1.4.x as
well although stripe count defaults to 1. The above formula can be misleading
sometimes if a particular OST is full or if you add new OSTs with some
old OSTs almost full. Are there plans to refine this ?
</DIV>
<DIV>
Currently, "du -sh" on Lustre FS would reflect
the actual Lustre-capacity used.
</DIV>
<DIV> </DIV>
<DIV>
Thanks,
</DIV>
<DIV>
-Kums
</DIV>
</body>
</html>
--=__Part0022017C.0__=--