Operating System - Linux
1828227 Members
3336 Online
109975 Solutions
New Discussion

rm: cannot remove `TGSmakeit': No such file or directory

 
Niraj Kumar Verma
Trusted Contributor

rm: cannot remove `TGSmakeit': No such file or directory

Dear Experts,

We have an strange problem on Redhat EL 3.0 U4 WS ( 2.4.xxxx kernel).

While running build I get the following error:

gmake[4]: Leaving directory `/project/dvdrw2/users/dvdr_blr/build/C1_5/1303/software-build_c1_5/software/dev
/appl/comp/ola/test'
rm: cannot remove `TGSmakeit': No such file or directory
rm: cannot remove `TGSgenerate': No such file or directory
rm: cannot remove `TGSgenerate.c': No such file or directory
gmake[3]: *** [test] Error 1

The strange part is most of the time the build runs without any problem .. but some time it fails. I did few chek to see if the file exists before removing it .. and it does ..

The project area is on NFS mounted FS from and EMC NAS box ...

Any clue to where will the problem be ? Does it some thing to do with inode ? ( Just a wise guess)

Regards
Niraj.Verma@philips.com
2 REPLIES 2
Stuart Browne
Honored Contributor

Re: rm: cannot remove `TGSmakeit': No such file or directory

What NFS options have you set for the filesystem?

In particular, are you using NFS over TCP, and with what block/buffer sizes for read & write?

It looks as if it is making a file on the filesystem, but when it goes to read it back, the filesystem hasn't got it yet. As if the NAS is buffering the request to create the file, and hasn't actually done it yet.

What sort of buffering/caching options does the NAS have? Can you turn them off for testing?
One long-haired git at your service...
thibault ketterer
New Member

Re: rm: cannot remove `TGSmakeit': No such file or directory

Hi, I have the exact same problem
EMC NAS clarion + with RH EL 3 up 4 2.4.21-32.ELsmp

/foo mounted by nfs

ls /foo
/foo
/foo/bar

In most cases it works, but
sometime rm says :
$ rm -r /foo
rm: cannot remove `/foo/bar': No such file or directory

/foo/bar is erased, but the message is wrong and rm exits with not 0.

I have a full encapsulated procedure which stops when rm fails. Very annoying problem

There is many levels of cache:
- nfs client rsize=32768 and wsize=4096
- nfs server attribute cache
- EMC disk cache

While my system Is in production it is pretty hard to turn off EMC or nfs server cache, cause it would dracstically impact performances.