[gpfsug-discuss] Biggest file that will fit inside an inode?
Yuri L Volobuev
volobuev at us.ibm.com
Fri Oct 7 17:20:01 BST 2016
Exactly how much data can fit into an inode depends on whether any extended
attributes (EAs) are present, which makes this complicated, as EAs may be
set explicitly by the user or implicitly by the system. In any inode,
there's a fixed footprint of the inode header, 128 bytes on recent GPFS
streams. So if there's no possibility to fit more than (inodeSize - 128)
bytes into an inode. If some EAs are present, this becomes more
complicated. GPFS can store EAs either in the inode, or in a special "EA
overflow block". Certain EAs used by GPFS internally (for encryption, AFM,
DMAPI, clones) must reside in the inode, due to the need to have those set
at certain junctures when it's tricky to accommodate EA overflow allocation
with proper atomicity guarantees. Other subsystems, e.g. SELinux, may
implicitly use EAs as well. So a given inode may have less space for data
than described above. This is another big part of the reason why 4K is the
current default inode size. And of course this is not an external spec set
in stone, but rather the way the current implementation works. A future
version of GPFS may have a bigger inode header, for example. So it would
be imprudent to bank on file/directory data fitting into an inode with a
great deal of precision. A better way to view this is as a kind of a
performance optimization.
DMAPI won't "punch out" data that resides in inode. Such a file has zero
blocks allocated, so there's nothing for an HSM app to migrate. TCT can
copy in-inode data to cloud for co-resident mode. And yes, TCT doesn't
rely on the same DMAPI API as HSM, per se, but naturally the underlying
code shares some of the infrastructure.
yuri
From: Luke Raimbach <luke.raimbach at googlemail.com>
To: gpfsug main discussion list <gpfsug-discuss at spectrumscale.org>,
Date: 10/03/2016 09:16 AM
Subject: Re: [gpfsug-discuss] Biggest file that will fit inside an
inode?
Sent by: gpfsug-discuss-bounces at spectrumscale.org
It doesn't, but the end result is the same... data shipped off 'somewhere
else' with a stub file?
I have in my mind that DMAPI support was around before data-in-inode (or at
least 4K inodes) was introduced, so it had to be made a bit cleverer to
cope, but I may be misremembering that.
On Mon, 3 Oct 2016 at 17:10 Simon Thompson (Research Computing - IT
Services) <S.J.Thompson at bham.ac.uk> wrote:
TCT doesn't use dmapi though I thought?
________________________________________
From: gpfsug-discuss-bounces at spectrumscale.org [
gpfsug-discuss-bounces at spectrumscale.org] on behalf of Luke Raimbach [
luke.raimbach at googlemail.com]
Sent: 03 October 2016 17:07
To: gpfsug main discussion list
Subject: Re: [gpfsug-discuss] Biggest file that will fit inside an inode?
Surely it wouldn't go? Maybe the data would get copied out rather than
stubbed... DMAPI can't be stupid enough to stub data out of an inode? Can
it? Interesting question.
Maybe I'll test that one.
On Mon, 3 Oct 2016 at 17:00 Simon Thompson (Research Computing - IT
Services) <S.J.Thompson at bham.ac.uk<mailto:S.J.Thompson at bham.ac.uk>>
wrote:
Would you tier an in-inode file to the cloud?
I mean, I wouldn't tier an in-inode file out to tape?
Simon
________________________________________
From: gpfsug-discuss-bounces at spectrumscale.org<mailto:
gpfsug-discuss-bounces at spectrumscale.org> [
gpfsug-discuss-bounces at spectrumscale.org<mailto:
gpfsug-discuss-bounces at spectrumscale.org>] on behalf of Oesterlin, Robert
[Robert.Oesterlin at nuance.com<mailto:Robert.Oesterlin at nuance.com>]
Sent: 03 October 2016 16:56
To: gpfsug main discussion list
Subject: Re: [gpfsug-discuss] Biggest file that will fit inside an inode?
What's going be taken away if you use Encryption or Transparent Cloud
Tiering?
Bob Oesterlin
Sr Storage Engineer, Nuance HPC Grid
From: <gpfsug-discuss-bounces at spectrumscale.org<mailto:
gpfsug-discuss-bounces at spectrumscale.org>> on behalf of Marc A Kaplan <
makaplan at us.ibm.com<mailto:makaplan at us.ibm.com>>
Reply-To: gpfsug main discussion list <gpfsug-discuss at spectrumscale.org
<mailto:gpfsug-discuss at spectrumscale.org>>
Date: Monday, October 3, 2016 at 10:46 AM
To: gpfsug main discussion list <gpfsug-discuss at spectrumscale.org<mailto:
gpfsug-discuss at spectrumscale.org>>
Subject: [EXTERNAL] Re: [gpfsug-discuss] Biggest file that will fit
inside an inode? 3968!!
On a non-SELINUX system the answer is 3968 of data in a 4K inode, just
128 bytes of metadata.
Caution: it's possible in some future release, this could change ... I
don't know of any plans, I'm just saying ...
Inode 16346892 [16346892] snap 0 (index 12 in block 255420):
Inode address: 6:123049056 size 4096 nAddrs 330
indirectionLevel=INODE status=USERFILE
objectVersion=1 generation=0xC0156CB nlink=1
owner uid=0 gid=0 mode=0200100644: -rw-r--r--
blocksize code=5 (32 subblocks)
lastBlockSubblocks=0
checksum=0xAD8E0B4B is Valid
fileSize=3968 nFullBlocks=0
currentMetadataReplicas=1 maxMetadataReplicas=2
currentDataReplicas=1 maxDataReplicas=2
...
Data [3968]:
0000000000000000: BCA91252 2B64BEDC A7D7BA9D D5BE8C30 *...R+d.........0*
...
0000000000000F70: DA925E2F 16A68C01 03CA5E37 08D72B7F *..^/......^7..+.*
trailer: is NULL
_______________________________________________
gpfsug-discuss mailing list
gpfsug-discuss at spectrumscale.org<http://spectrumscale.org>
http://gpfsug.org/mailman/listinfo/gpfsug-discuss
_______________________________________________
gpfsug-discuss mailing list
gpfsug-discuss at spectrumscale.org
http://gpfsug.org/mailman/listinfo/gpfsug-discuss
_______________________________________________
gpfsug-discuss mailing list
gpfsug-discuss at spectrumscale.org
http://gpfsug.org/mailman/listinfo/gpfsug-discuss
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://gpfsug.org/pipermail/gpfsug-discuss_gpfsug.org/attachments/20161007/a797290f/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: graycol.gif
Type: image/gif
Size: 105 bytes
Desc: not available
URL: <http://gpfsug.org/pipermail/gpfsug-discuss_gpfsug.org/attachments/20161007/a797290f/attachment.gif>
More information about the gpfsug-discuss
mailing list