home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!cs.utexas.edu!news
- From: fletcher@cs.utexas.edu (Fletcher Mattox)
- Newsgroups: comp.sys.sun.admin
- Subject: automounter swamps mountd
- Date: 14 Sep 1992 19:05:44 -0500
- Organization: CS Dept, University of Texas at Austin
- Lines: 22
- Message-ID: <lbaa6oINNlei@deepthought.cs.utexas.edu>
- NNTP-Posting-Host: deepthought.cs.utexas.edu
-
- Why is it so easy to swamp rpc.mountd with mount requests?
-
- The use of the automounter on the clients seems to exacerbate the problem,
- though I have seen it happen without the automounter, too. Once mountd
- falls behind, it never catches up. Even if no new requests arrive.
- The unread data on the mountd socket never gets drained.
-
- This example is on a Sequent NFS server, but I've seen similar behaviour
- under SunOS 4.1.1.
-
- cs.utexas.edu% rpcinfo -p localhost | grep mountd
- 100005 1 udp 1553 mountd
- cs.utexas.edu% netstat -n -a | egrep "Proto|1553"
- Proto Recv-Q Send-Q Local Address Foreign Address (state)
- udp 32572 0 *.1553 *.*
- cs.utexas.edu%
-
- 32k of unread data represents an enormous number of outstanding mount
- requests. Why can't mountd keep up? Or perhaps this is really a
- problem with inetd? (I run mountd under inetd).
-
- Fletcher
-