solaris8 - Day 7

Goals

• use automount
• tune your network

1. what is a "backfile system" on CacheFS?
2. what is a "frontfile system"?
3. what is a "cache miss"?
4. what is a "cache hit"?
5. set up a cached filesystem (cachefs) on your box. Set the maximum diskspace usage to about 15%.
6. look at the filesystems being exported by the instructor's box: mount the one that mentions day7post on your cachefs you just created
7. what is in that directory?
8. how can you find out some stats on the cache?
9. how many filesystems can be mounted on the same cache?

autofs/automounter

• client-side; the remote fs is always exported
• allows user-enabled mounting
• speeds up boot
• minimizes memory overhead
• minimizes network traffic
• minimizes lockups do to NFS server failures
• falls off after a few minutes
• easier to shutdown/reboot an exporting box
• do not mix manually and auto mounting for the same mount!
• autofs behaviour controlled by maps.
• allows NIS management of certain types of filesystems (usually fstab is box-specific)

setting up an automount

• eyeball /etc/init.d/autofs
• eyeball /etc/auto_master: format: mountpoint map -options
• build direct and/or indirect maps

indirect maps

• most common type
• the "trigger" is a directory; the indirect map controls the entire directory
• directory is the key
• mountpoints do not exist; created by autofs
• can be several indirect maps on the box
• used to forge a consistent structure from disparate sources
• strange mounting artifacts like
• less likely to cause "mount storms"

direct maps

• used less frequently
• more like a normal mount, with a normal mountpoint
• good for odd mounts, or
• uses special /- key; directories are spelled out in the map
• alteration of the direct map should entail a reloading of the aufofs system.
• use $VARS in NIS-managed networks

automounter weirdness

• automounted dirs in your PATH can cause delays
• automounting local FSs can cause strange paths
• calendar and other site-walking tools (find!) can thrash a NFS-serving box
• pwd can produce odd or unexpected results:
• to kill automount, use kill -TERM pid to avoid NFS hangs.

network tuning

• automount
  • use failover (two -line \ syntax)
  • increase the automagic unmount time if users keep files open longer than the timeout.
  • recurrimg jobs occur (cron?) that have a period longer than the timeout
  • note that longer connection times are more efficient but increase the risk of a hang.
  • length of task execution will affect user perceptions
  • be aware a collection of direct-mapped FSs in a directory; ls can cause a storm
  • subdirectory:fields help with overmapped indirects (warning, can cause strange-looking paths)
• tcp
  • ensure a consistent and vendor approved broadcast address syntax (ie, 1s or 0s): ifconfig
  • arp -a to see the IP-MAC mapping on the network
  • ping -s (solaris) to see if a host is accessible, to learn about host/IP inconsistencies. Various NFS services ping the various hosts. Random timeouts or peaks can be a termination or cable problem.
  • spray to test the network, NICs, etc. use various -l lengths. Writes tend to be large. Reads and attributes are tiny. Try passing a -d delay of 1 microsecond to see if it can keep up then. Note the bandwidth usage.
  • traceroute to see the route to a host
  • rpcinfo -t/u host nfs to see if portmapper has registered RPC services.
  • ypcat map (note -k option, and apparent randomness)
  • ypmatch key map
  • ypwhich shows current binding
  • ypwhich -x shows nicknames
  • ypwhich -m map shows master for a map
  • ypset server force a binding if possible

  • exportfs shows exports and options
  • /etc/rmtab can get staled by dying clients, or by changing hostname without umount -a
  • df and friends can be inaccurate in heterogenous environments.
  • nfsstat -s/-c
    • calls/strong> total NFS calls
    • badcalls rejected RPC becore NFS handoff
    • nullrecv nfsd with nothing to do... decrease them.
    • badcalls damaged or poorly-formed packet
    • retrans no response; asked again
    • timeout hang or fail (hard/soft)
    • badxids mean the server is receiving retrans but is not answering...
    • wait waiting on a file id handle

  time service

  Time can affect NIS polls, makefiles, etc. May want to define a timehost
  • rdate host
  • public timeservers

  things that go wrong

  • renegade nis servers
  • interfering IPs
  • interfereing hostnames

    ---

    http://www.mousetrap.net/syllabus/solaris8/day7.html
    $Id: day7.orb,v 1.4 2002/11/21 00:08:37 mouse Exp $