Changeset 4abace9 for fedkit


Ignore:
Timestamp:
Feb 29, 2008 9:45:35 AM (16 years ago)
Author:
Ted Faber <faber@…>
Branches:
axis_example, compt_changes, info-ops, master, version-1.30, version-2.00, version-3.01, version-3.02
Children:
f64fa81
Parents:
906c763
Message:

Changes to imporve reliability of routes coming up. Restructuring of
federate leaves the stsyem in a state where we can safely restart the router
(specifically, talking to the local boss again) and we restart it in the
final topology. This introduces a lag in routing actually being established.
It may be useful to add a lag to the startcmd to take this into account.

Location:
fedkit
Files:
1 added
4 edited

Legend:

Unmodified
Added
Removed
  • fedkit/fed-tun.pl

    r906c763 r4abace9  
    265265           ($tunnel_mask ? " netmask $tunnel_mask" : ""));
    266266    warn "configuration of tunnel interface failed" if ($?);
    267        
    268     system("route add $opts{'peer'} $tunnel_router");
    269     warn "configuration routes via tunnel interface failed" if ($?);
     267
     268    # Sometimes the insertion of DNS names lags a bit.  Retry this
     269    # configuration a few times to let DNS catch up.  Might want to really
     270    # check the DNS name before we try this...
     271    my $config_succeeded = 0;
     272    my $tries = 0;
     273    my $max_retries = 6;
     274
     275    do {
     276        system("route add $opts{'peer'} $tunnel_router");
     277        if ( $? ) {
     278            warn "configuration routes via tunnel interface failed";
     279            $tries++;
     280            sleep(10);
     281        }
     282        else { $config_succeeded = 1; }
     283    } until ( $config_succeeded || $tries > $max_retries );
    270284
    271285    print "setup_tunnel_cfg done\n" if ($debug);
  • fedkit/federate.sh

    r906c763 r4abace9  
    11#!/bin/sh
    22
    3 # Take the user and the user's startcmd as arguments
     3# Set up a federated environment on a client.  Unmount all local file systems,
     4# erase all local accounts and then bring in accounts and file systems from the
     5# master testbed.  Once all that's done, restore client communication to the
     6# local bossnode and execute any startup command passed in as arguments to this
     7# script.
    48
     9
     10# The file containing the DNS name of the current boss
    511BOSSNODE="/usr/local/etc/emulab/bossnode"
    6 # GATEWAY="detertunnel"
    712
     13# Ports that the master gateway will forward
    814PORTS="139 7777"
     15# network cat command (to listen for the gateway)
    916NC="/usr/bin/nc"
     17# Emulab rc scripts directory
    1018RCDIR="/usr/local/etc/emulab/rc"
    1119
     20# Emulab rc script to reset accounts and federation srcipt to start the smb
     21# automount process.
    1222RCACCT="rc.accounts"
    1323SMBMOUNT="smbmount.pl"
     
    2636SCRIPTDIR=`$EMUDIR/tmcc -b status | perl -ne '/ALLOCATED=([^\/]+)\/([^\s]+)/ && print "/proj/$1/exp/$2/tmp\n";'`
    2737
    28 # Die if things are weird
     38# Die if Scripts are not where they should be.  NB, this is checked by the
     39# bootstrapper, too, so a failure here is unlikely.
    2940if [ -z "$SCRIPTDIR" ] ; then
    3041    echo "Can't find federation scripts.  Tmcc status says:"
     
    3344fi
    3445
    35 # Move the SMB mounting script out of NFS so we can call it after the umount
    36 cp "$SCRIPTDIR/$SMBMOUNT" /tmp
    37 
    3846# Copy the global hostnames file into /etc/hosts so this node can address
    39 # others in the experiment on other testbeds.
     47# others in the experiment on other testbeds.  The extra steps preserves teh
     48# localhost entry that's aliased to this host's name.
    4049grep 127\\.0\\.0\\.1 /etc/hosts > /tmp/hosts
    4150cat "$SCRIPTDIR/hosts" >> /tmp/hosts
     
    6776
    6877echo "Configuring TMCC."
    69 
    70 $EMUDIR/sethostname
    71 
    7278echo $GATEWAY > $BOSSNODE
    7379
     
    8591
    8692#
    87 # Copy and exec the smbmount perl script
     93# Call the smbmount perl script, this invokes the automounter to mount teh
     94# needed smb filesystems.
    8895#
    89 
    9096echo "Mounting via SMB."
    91 
    9297"/tmp/$SMBMOUNT" $SHARE $GATEWAY $SMBUSER
    9398
     99echo "Restoring old bossnode"
     100# remove bossnode override
     101/bin/rm -f $BOSSNODE
     102# Also clear any cached data that might interfere with rourte construction.
     103/bin/rm -f /var/emulab/boot/tmcc/ifconfig
     104
     105# I remain unclear why ospfd seems screwed up after the resetting of accounts
     106# and mounts.  This seems to fix it, but it's not very satisfying.  At this
     107# point the federated topology is complete, so any problems with routing in a
     108# partial topology -e.g. failed or partial bridging - that might have confused
     109# the router should be gone.
     110/bin/pkill ospf
     111/bin/pkill gated
     112/usr/local/etc/emulab/rc/rc.route boot
     113
     114# Execute any command passed in as a startcmd.  This basically daisy chains the
     115# startcmd.
    94116if [ ! -z "$1" ]; then
    95117        su -l "$1" -c "$2"
  • fedkit/splitter.pl

    r906c763 r4abace9  
    99use File::Copy;
    1010
    11 my @scripts = ("federate.sh", "smbmount.pl", "make_hosts", "fed-tun.pl");
     11my @scripts = ("fed_bootstrap", "federate.sh", "smbmount.pl",
     12    "make_hosts", "fed-tun.pl");
    1213my $local_script_dir = ".";
    1314my($pid, $gid);                 # Process and group IDs for calling parse.tcl
  • fedkit/testbeds.example

    r906c763 r4abace9  
    4949# These slave commands are appropriate for generic testbeds, which emulab and
    5050# WAIL are.  The -H on sudo is crucial when calling federate.sh
    51 SlaveNodeStartCmd: sudo -H /bin/sh FEDDIR/federate.sh >& /tmp/federate
     51SlaveNodeStartCmd: sudo -H /bin/sh FEDDIR/fed_bootstrap >& /tmp/federate
    5252SlaveConnectorStartCmd: sudo -H FEDDIR/fed-tun.pl -f GWCONF>& /tmp/bridge.log
    5353
    5454# These master commands are placeholders
    55 MasterNodeStartCmd: sudo -H /bin/sh FEDDIR/federate.sh &> /tmp/mfederate
     55MasterNodeStartCmd: sudo -H /bin/sh FEDDIR/fed_bootstrap &> /tmp/mfederate
    5656MasterConnectorStartCmd: sudo -H /users/faber/fed-tun.pl >& /tmp/bridge.log
    5757FileServer: fs
     
    6666ConnectorType: pc
    6767ConnectorImage: FBSD61-TUNNEL2
    68 SlaveNodeStartCmd: sudo -H /bin/sh FEDDIR/federate.sh >& /tmp/federate
     68SlaveNodeStartCmd: sudo -H /bin/sh FEDDIR/fed_bootstrap >& /tmp/federate
    6969SlaveConnectorStartCmd: sudo -H FEDDIR/fed-tun.pl -f GWCONF>& /tmp/bridge.log
    70 MasterNodeStartCmd: sudo -H /bin/sh FEDDIR/federate.sh &> /tmp/mfederate
     70MasterNodeStartCmd: sudo -H /bin/sh FEDDIR/fed_bootstrap &> /tmp/mfederate
    7171MasterConnectorStartCmd: sudo -H /users/faber/fed-tun.pl >& /tmp/bridge.log
    7272FileServer: fs
Note: See TracChangeset for help on using the changeset viewer.