13.07.2015 Views

Linux System Administration Recipes A Problem-Solution Approach

Linux System Administration Recipes A Problem-Solution Approach

Linux System Administration Recipes A Problem-Solution Approach

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

CHAPTER 2 ■ CENTRALIZING YOUR NETWORK: KERBEROS, LDAP, AND NFS■ Note If you’re using a system on which the file /etc/default/nfs-kernel-server exists (for example,Debian/Ubuntu), then you’ll also need to set the new RPCNFSDCOUNT value in this file.If you’re experiencing load or performance issues, you can also increase the memory queue limits.These are frequently low by default, especially on 2.4 kernels, and because the memory queue is sharedbetween all the server instances, it’s probably a good idea to increase the memory queue if you increasethe number of servers. The limits are set using proc. Use /proc/sys/net/core/rmem_default and/proc/sys/net/core/rmem_max for the read limits, and use /proc/sys/net/core/wmem_default and/proc/sys/net/core/wmem_max for the read limits. Reset them to 256k as follows:# echo 262144 > /proc/sys/net/core/rmem_default# echo 262144 > /proc/sys/net/core/rmem_max# echo 262144 > /proc/sys/net/core/wmem_default# echo 262144 > /proc/sys/net/core/wmem_maxThen restart the NFS server with /etc/init.d/nfs-kernel-server.■ Note There are some reports of problems generated in other systems if these values are left unchanged forlong periods of time. You can reset them after you’ve restarted the NFS server (using the same lines as shownpreviously, but substitute 65536 for 262144). However, bear in mind that in this case, the new values will bediscarded the next time you have to restart NFS. Perhaps better is to watch for problems and reset the values ifthey arise.nfsstat -s (on the server) and nfsstat -c (on the client) may also be helpful. These both show NFSstatistics. In particular, if badcalls is greater than zero, you may have network problems, and if writes isgreater than 10 percent, it may be worth looking into write caching to speed up operation.nfsstat -m from the client provides server information from the client’s point of view. The srttgives a smoothed round-trip time value. If this is greater than 50ms, the mount point is slow. cur givesthe timeout value. If cur > 80ms for lookups, cur > 150ms for reads, and/or cur > 250ms for writes, thenthe requests are taking too long.61Download at WoweBook.Com

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!