Bug 1961 - NFS mageia-mandriva slow depending on who establishes connection
Summary: NFS mageia-mandriva slow depending on who establishes connection
Status: RESOLVED INVALID
Alias: None
Product: Mageia
Classification: Unclassified
Component: RPM Packages (show other bugs)
Version: 1
Hardware: x86_64 Linux
Priority: Normal normal
Target Milestone: ---
Assignee: Guillaume Rousse
QA Contact:
URL:
Whiteboard:
Keywords: NEEDINFO
Depends on:
Blocks:
 
Reported: 2011-06-29 09:58 CEST by Arthur Dent
Modified: 2011-12-22 10:19 CET (History)
2 users (show)

See Also:
Source RPM:
CVE:
Status comment:


Attachments

Description Arthur Dent 2011-06-29 09:58:09 CEST
I have set up nfs in mageia1 and mandriva 2010.2 (with kernel 2.6.38.8), both are 64bit machines
If i establish the nfs connection from mageia to mandriva, the speed is as it should be - quick:

e.g. user@mageia # mount mandriva:/data /mnt
I can copy files in both directions with nearly 12 MB/sec, which is fine.

If i mount the mageia machine on mandriva, then things slow down considerably when copying files from mandriva to mageia (around 230 kB/sec):
user@mandriva # mount mageia:/data /mnt


What is different is the mount-string appearing on mageia/mandriva:
- mageia mounting nfs of mandriva:
(excerpt of mount-command)
mandriva:/data on /mnt type nfs (rw,vers=4,addr=192.168.3.10,clientaddr=192.168.3.11)

mandriva mounting nfs of mageia:
(excerpt of mount-command)
mageia:/data on /mnt type nfs (rw,addr=192.168.3.11)

- There are no nfs-related entries in etc/fstab, the connection is done on demand in the shell

- i also tried to connect using expicit rsize and wsize parameters, so far without any change in speed: the copy of files from mandriva to mageia is still slow (~230kB/sec)

- commands tried so far:
user@mandriva # mount mageia:/data /mnt -o rsize=8192,wsize=8192
user@mandriva # mount mageia:/data /mnt
user@mandriva # mount -t nfs mageia:/data /mnt

The following files are identical on both mandriva and mageia machines (setup for testing purposes)

/etc/hosts.allow
portmap: 192.168.3.0/255.255.255.0
nfsd: 192.168.3.0/255.255.255.0
mountd: 192.168.3.0/255.255.255.0
EOF

- /etc/hosts.deny
portmap: ALL
nfsd: ALL
mountd: ALL
EOF

- /etc/exports
/data 192.168.3.0/255.255.255.0(root_squash,sync,secure,subtree_check,rw)
Comment 1 Samuel Verschelde 2011-10-01 16:12:50 CEST
Guillomovitch, is there something you could do to help debugging this? (if not, and if you have another name to give to me, I take it :))

CC: (none) => guillomovitch, stormi

Comment 2 Guillaume Rousse 2011-10-04 11:23:35 CEST
From the mount string, you're probably using NFSv4 in the one case (mageia -> mandriva), and NFSv3 in the other case (mandriva -> mageia). That's probably due to protocol negociation failure in the second case, due to a missing NFSv4-specific component (rpc.idmapd, probably). Also, you need to use rpcbind, not portmap, for NFSv4. And if running on a private local network segment, using tcpwrapper is probably overkill (but that's up to you).

On both sides, check your /etc/sysconfig/nfs-common file for NEED_IDMAPD parameter, and use 'service nfs-common status' to check if rpc.idmapd is running. If not enough, you can force protocol version with nfsvers parameter, and also use mount -v flag for details. Also, don't bother with rsize and wsize parameters, they're quite useless excepted in a few scenarios.

Status: NEW => ASSIGNED
Assignee: bugsquad => guillomovitch

Comment 3 Guillaume Rousse 2011-12-13 22:49:41 CET
Ping ?

Keywords: (none) => NEEDINFO

Comment 4 Guillaume Rousse 2011-12-22 10:19:10 CET
This problem is now 6 month old, and is likely to be a local configuration issue. I'm closing it, given the lack of interest from original reporter.

Status: ASSIGNED => RESOLVED
Resolution: (none) => INVALID


Note You need to log in before you can comment on or make changes to this bug.