Re: [sqwebmail] Re: [Courier-imap] Problem with Netapp and readdir
Brian Candler <[EMAIL PROTECTED]> writes: > I've just been able to test another platform: Solaris-2.8 client with a > NetApp server. In this case I get 200/200 transferred. In fact I get > 2000/2000 if I make the test that big as well. I've also confirmed Solaris client works OK; /mnt is the NetApp, the client is Solaris: > ./testnfs /mnt/home/cshenton/SOLNFS/ Transferred 200 out of 200 files > uname -a SunOS thing 5.9 Generic_112233-08 sun4u sparc SUNW,Ultra-60 > So this is starting to look like a FreeBSD issue Grumble...
Re: [sqwebmail] Re: [Courier-imap] Problem with Netapp and readdir
Brian Candler <[EMAIL PROTECTED]> writes: > I've just been able to test another platform: Solaris-2.8 client with a > NetApp server. In this case I get 200/200 transferred. In fact I get > 2000/2000 if I make the test that big as well. > > So this is starting to look like a FreeBSD issue; would be interesting to > see if someone could try it with Linux client and/or NFS server. First a local-disk test for sanity: [EMAIL PROTECTED](292> ./testnfs Transferred 200 out of 200 files Then one on the NetApp: [EMAIL PROTECTED](294> mkdir /netapp/home/cshenton/NFSTEST [EMAIL PROTECTED](295> ./testnfs /netapp/home/cshenton/NFSTEST Transferred 169 out of 200 files I'm running FreeBSD: [EMAIL PROTECTED](296> uname -a FreeBSD Palimpsest.saic.hq.nasa.gov 4.9-PRERELEASE FreeBSD 4.9-PRERELEASE #17: Thu Sep 4 10:55:56 EDT 2003 [EMAIL PROTECTED]:/usr/obj/usr/src/sys/Palimpsest i386 I'd love to try this in a FreeBSD-5.1 system but don't have one here right now. Maybe tonight when I get home...
Re: [sqwebmail] Re: [Courier-imap] Problem with Netapp and readdir
On Tue, Oct 07, 2003 at 12:36:35PM +0200, Stefan Kaltenbrunner wrote: > > So this is starting to look like a FreeBSD issue; would be interesting to > > see if someone could try it with Linux client and/or NFS server. > > works here too (200/200) client is Linux 2.4.22, server is > 2.4.9+patches(redhat enterprise server kernel) > > using a FreeBSD 4.9RC-client against the same server fails(169/200) Many thanks. I will submit a FreeBSD PR, although now we have the workaround it's not such a problem. Cheers, Brian.
Re: [sqwebmail] Re: [Courier-imap] Problem with Netapp and readdir
On Mon, Oct 06, 2003 at 09:49:17AM -0700, Doug Clements wrote: > > Well, I thought it was a Netapp problem, but now it looks like NFS and/or > > FreeBSD. The attached program replicates it: > > > > bash-2.05a# ./testnfs /na0/testdir > > Transferred 169 out of 200 files > > > > That was with a FreeBSD-4.8p3 frontend with the Netapp backend. However > > re-running it with FreeBSD-4.7 talking to a Solaris-2.8 NFS backend gives > > exactly the same result :-( > > I also get 169/200 on FreeBSD-4.8 with a FreeBSD-4.8 NFS server. I've just been able to test another platform: Solaris-2.8 client with a NetApp server. In this case I get 200/200 transferred. In fact I get 2000/2000 if I make the test that big as well. So this is starting to look like a FreeBSD issue; would be interesting to see if someone could try it with Linux client and/or NFS server. Cheers, Brian.
Re: [sqwebmail] Re: [Courier-imap] Problem with Netapp and readdir
On Mon, Oct 06, 2003 at 09:49:17AM -0700, Doug Clements wrote: > > bash-2.05a# ./testnfs /na0/testdir > > Transferred 169 out of 200 files > > > > That was with a FreeBSD-4.8p3 frontend with the Netapp backend. However > > re-running it with FreeBSD-4.7 talking to a Solaris-2.8 NFS backend gives > > exactly the same result :-( > > I also get 169/200 on FreeBSD-4.8 with a FreeBSD-4.8 NFS server. Thanks. I just installed the current dev build of courier-imap though and it successfully works around the problem - i.e. on the first POP3 connection absolutely all mail is available. Cheers, Brian.
Re: [sqwebmail] Re: [Courier-imap] Problem with Netapp and readdir
On Mon, Oct 06, 2003 at 05:29:14PM +0100, Brian Candler wrote: > On Mon, Oct 06, 2003 at 11:59:08AM -0400, Chris Shenton wrote: > > > After renaming 40 or so files, the readdir() terminates, probably because > > > the state of the directory is being changed while it is being read. Arguably > > > this is a Netapp bug, > > > > This doesn't occur if you mount some UNIX box via NFS? What version > > of DataOntap are you running? > > 6.4R1 > > Well, I thought it was a Netapp problem, but now it looks like NFS and/or > FreeBSD. The attached program replicates it: > > bash-2.05a# ./testnfs /na0/testdir > Transferred 169 out of 200 files > > That was with a FreeBSD-4.8p3 frontend with the Netapp backend. However > re-running it with FreeBSD-4.7 talking to a Solaris-2.8 NFS backend gives > exactly the same result :-( I also get 169/200 on FreeBSD-4.8 with a FreeBSD-4.8 NFS server. --Doug