nfs corrupts files
Subsystems (2)
hi everyone,
ok, here is what i did:
i had some files on an nfs-server (v3) using kernel 2.6.8-gentoo-r8 (updated because of the trouble). the 9 files had an average size of about 180mB and i built the md5sums of these files via ssh _on_ this server.
after that i transferred these files to my workstation, once via nfs and once via smb and let them check with md5sum again. the files transferred with smb were ok, but 6 of the 9 files tranferred via nfs were corrupt. then i made cmp between the local files (transferred with nfs of course) and the files on the nfs-server. the result:
1) multiple use of cmp on files which md5sum mentioned as corrupt resulted in reporting differences at different positions
2) sometimes the usage of cmp on files which md5sum mentioned as correct resulted in reporting differences...
so, as you see, there's definitly something wrong
now some further infos:
there are no log messages either on the client nor on the server.
export-config:
---------------------------------------------------
mephisto root # exportfs -v
/raid/data/public/media
baal(rw,wdelay,nohide,no_root_squash)
/raid/data baal(rw,wdelay,no_root_squash)
---------------------------------------------------
mephisto is my servername, baal is my clientname and the subdirectory has to be reexported as it is a new mount-point
the line from the /etc/fstab on client:
---------------------------------------------------
thomas@baal thomas $ grep nfs /etc/fstab
mephisto:/raid/data/ /data nfs rw,hard,intr 0 0
---------------------------------------------------
i'm thankful for any help as i don't know, where the mistake might be....
thomas
P=E5 to , 14/10/2004 klokka 18:14, skreiv Thomas Lenherr:
Quoted 7 lines
ok, here is what i did:
i had some files on an nfs-server (v3) using kernel 2.6.8-gentoo-r8 (updated because of the trouble). the 9 files had an average size of about 180mB and i built the md5sums of these files via ssh _on_ this server.
=20
after that i transferred these files to my workstation, once via nfs and
Quoted 8 lines
once via smb and let them check with md5sum again. the files transferred with smb were ok, but 6 of the 9 files tranferred via nfs were corrupt. then i made cmp between the local files (transferred with nfs of course) and the files on the nfs-server. the result:
1) multiple use of cmp on files which md5sum mentioned as corrupt resulted in reporting differences at different positions
2) sometimes the usage of cmp on files which md5sum mentioned as correct resulted in reporting differences...
You don't mention which kernel the client was running. Was it the same as the server?
Have you tried just grabbing a kernel 2.6.8.1 from ftp.kernel.org, and compiling it up (using the same .config as the gentoo kernel if you like) and then doing the same thing? I ask 'cos I just don't know what is in the gentoo kernels.
Cheers,
Trond
-------------------------------------------------------
This SF.net email is sponsored by: IT Product Guide on ITManagersJournal Use IT products in your business? Tell us what you think of them. Give us Your Opinions, Get Free ThinkGeek Gift Certificates! Click to find out more
http://productguide.itmanagersjournal.com/guidepromo.tmpl
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs
sorry, for forgetting the kernel of the workstation, it's yet another gentoo-kernel: 2.6.7-gentoo-r6. i looked up the difference between the vanilla kernel 2.6.7 and this patched gentoo-version and it seems it's not really much, just stuff like bootsplash, console speakup and something about making it ready for supermount...
anyway, as you said i installed the original 2.6.8.1 on the server and tried the whole thing again, what means i copied the 9 files over nfs and let them check with md5sum. same result :(
i did not yet update the kernel of the workstation, because i accessed the nfs server with an windows-implementation of nfs and experienced the same troubles. so i think the server is the faulty one...
cheers,
thomas
Quoted 45 lines · Trond Myklebust
På to , 14/10/2004 klokka 18:14, skreiv Thomas Lenherr:
Quoted 6 lines
ok, here is what i did:
i had some files on an nfs-server (v3) using kernel 2.6.8-gentoo-r8 (updated because of the trouble). the 9 files had an average size of about 180mB and i built the md5sums of these files via ssh _on_ this server.Quoted 11 lines
after that i transferred these files to my workstation, once via nfs and once via smb and let them check with md5sum again. the files transferred with smb were ok, but 6 of the 9 files tranferred via nfs were corrupt. then i made cmp between the local files (transferred with nfs of course) and the files on the nfs-server. the result:
1) multiple use of cmp on files which md5sum mentioned as corrupt resulted in reporting differences at different positions
2) sometimes the usage of cmp on files which md5sum mentioned as correct resulted in reporting differences...You don't mention which kernel the client was running. Was it the same as the server?
Have you tried just grabbing a kernel 2.6.8.1 from ftp.kernel.org, and compiling it up (using the same .config as the gentoo kernel if you like) and then doing the same thing? I ask 'cos I just don't know what is in the gentoo kernels.
Cheers,
Trond-------------------------------------------------------
This SF.net email is sponsored by: IT Product Guide on ITManagersJournal Use IT products in your business? Tell us what you think of them. Give us Your Opinions, Get Free ThinkGeek Gift Certificates! Click to find out more
http://productguide.itmanagersjournal.com/guidepromo.tmpl
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs
it's me again...
is there really nobody who might help me? i just don't know what else i should try....
here for short again the problem and what i tried to solve it:
ok: i had 9 big files (each ~180mB) on an nfsv3-server, and they got corrupt when i copied
them down (of course not only these files, but of the files on the nfs-server).
(copied the files via smb for testing, they got not corrupted)
i tried all of the following to solve it:
- updated the kernel of the server and workstation to vanilla 2.6.8.1
- tried nfsv2
- tried nfsv4
- tried nfs over tcp
- had ethereal checking the checksums of all transferred packets via udp while copying one file, all sums were ok
- copied the files _on_the_server_ on an other drive (which is connected to an other (p)ata-controller )
it might be i forgot some of the things i tried, but this list should contain most of it...
btw: is there some kind of a verbose-flag to toggle so i might get some more log-entries about nfs? right now i don't get any entry, neither on the server, nor on the client, so it's somehow like nfs reads wrong data from the drives and builds all the checksums after this so error is detected....
please somebody help me, i don't want switch to samba giving lousy 4mB/s transfer-speed instead of the 11mB/s of nfs... :(
thanks anyway....
thomas
Quoted 65 lines · Thomas Lenherr
sorry, for forgetting the kernel of the workstation, it's yet another gentoo-kernel: 2.6.7-gentoo-r6. i looked up the difference between the vanilla kernel 2.6.7 and this patched gentoo-version and it seems it's not really much, just stuff like bootsplash, console speakup and something about making it ready for supermount...
anyway, as you said i installed the original 2.6.8.1 on the server and tried the whole thing again, what means i copied the 9 files over nfs and let them check with md5sum. same result :(
i did not yet update the kernel of the workstation, because i accessed the nfs server with an windows-implementation of nfs and experienced the same troubles. so i think the server is the faulty one...
cheers,
thomasQuoted 45 lines · Trond Myklebust
På to , 14/10/2004 klokka 18:14, skreiv Thomas Lenherr:
Quoted 6 lines
ok, here is what i did:
i had some files on an nfs-server (v3) using kernel 2.6.8-gentoo-r8 (updated because of the trouble). the 9 files had an average size of about 180mB and i built the md5sums of these files via ssh _on_ this server.Quoted 10 lines
after that i transferred these files to my workstation, once via nfs and once via smb and let them check with md5sum again. the files transferred with smb were ok, but 6 of the 9 files tranferred via nfs were corrupt. then i made cmp between the local files (transferred with nfs of course) and the files on the nfs-server. the result:
1) multiple use of cmp on files which md5sum mentioned as corrupt resulted in reporting differences at different positions
2) sometimes the usage of cmp on files which md5sum mentioned as correct resulted in reporting differences...You don't mention which kernel the client was running. Was it the same as the server?
Have you tried just grabbing a kernel 2.6.8.1 from ftp.kernel.org, and compiling it up (using the same .config as the gentoo kernel if you like) and then doing the same thing? I ask 'cos I just don't know what is in the gentoo kernels.
Cheers,
Trond-------------------------------------------------------
This SF.net email is sponsored by: IT Product Guide on ITManagersJournal Use IT products in your business? Tell us what you think of them. Give us Your Opinions, Get Free ThinkGeek Gift Certificates! Click to find out more
http://productguide.itmanagersjournal.com/guidepromo.tmpl
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs