I'm running low on patience with how this is working out. I have a RHEL 5.4 64bit server running 389 directory services and NFS4. I set up the schema for automount, and I am having some issues.
I'm relatively new to both LDAP and Automount, and NFS4 has thrown me some curves vs NFS2/3.
Here's what I'd like to be able to do:
User on a remote box logs in, via Ldap (that works)
Their home directory is automounted from the NFS server and they can write to it (It can mount if the directory exists on the NFS server [can I have it automagically created?], but it is read only)
How can I configure this so that the user logs in, gets their home directory, and it just works (Without Kerberos).
you must a create an specific OU in LDAP and configure this for automount..( if this is ok , then their home directorys are already ok because of your maps (in automount) on the clients )..
The LDAP piece worked fine (NFSv3 works, NFSv4 fails). I have just about given up on NFSv4, and though I got it all working with NFSv3, RHEL seems to have a bug where if I do an ls -l when a file copy or write operation is being done to the remote volume, it lags for a LONG time (>20 seconds) before returning the listing.
Yes, whenever a large file transfer is going on. This is, I believe, due to the need for write locking to quiesce. Switching to UDP only mode in NFSv3 made it a little better, but I am investigating moving to iSCSI.
as for i know nfsv3 default already uses udp.maybe it is tcp in redhat.
at the now,between clients and server of network line is at least giga/10g and local and clear? is it?
and (>20 seconds) is big time so you must have really big files:)
for this maybe you can modify rsize/wsize settings on clients and remount and lets see what happened about times.
your power helper is below.. 5.*Optimizing NFS Performance
I'm sorry, but I am having a hard time understanding you. I suppose English is not your first language, so I will try my best.
The connection is NOT 10GbE, it is not even 1GbE. The wiring is old, so it is connecting at 100Mb. The file I copied was ~10GB in size, and 20s was the best observed performance. It was often much worse.
I am trying to convince my boss' boss who thinks 8 workstations can somehow be served over 7Mbit if they were to go across a WAN link at another site.
The r/w settings are not the issue, by the way. The issue has been narrowed down to the need for write locks to quiesce when doing a stat on the file system to provide for the ls -l. On its own, ls is fast.
