Description

Operation

disable

/usr/testbed/sbin/wap /usr/testbed/sbin/ctlsep_setup uninstall

enable

/usr/testbed/sbin/wap /usr/testbed/sbin/ctlsep_setup preinstall

/usr/testbed/sbin/wap /usr/testbed/sbin/ctlsep_setup install

post-tagger-reboot

/usr/testbed/sbin/wap /usr/testbed/sbin/ctlsep_setup tagger

Discussion

Disabling Control net separtion::

When running uninstall the first thing it does is to move all control interfaces in all swapped in experiments to the common CONTROL VLan which takes under 20 seconds. Experimenters should be able to ssh to their nodes, which should have normal NFS connectivity to the server users.isi.deterlab.net

It then performs some mop up operations which disable the traffic interfaces for both the ISI and Berkeley taggers, and removes all the other per-experiment control VLans. The act of untagging the output traffic interface for each tagger is a snmpit corner case which will produce some diagnostics, but it actually does the correct thing and by that point should be irrelevant to the historic operation of the testbed.

The entire cleanup operation should take less than 3 minutes.

Installing Control Net Separation::

To enable control net separation, both commands are required. The separation process requires about 12 seconds for each normal swapped in experiment (independent of the number of the nodes in the experiment!) and will take about 30 seconds for each firewalled experiment. The site variable swap/lockout is set, and must be reset manually after the operator verifies that it seems safe for users to be able to swap in/out experiments.

Wedged Tagger::

If the tagger node gets wedged and must be rebooted, it needs to be informed about which mac address go to which vlans.

The command ctlsep_setup tagger assumes that the tagger is up and running, formats the information derived from the database state and synchronizes the tagger via an ssh command.