Changes between Initial Version and Version 1 of Internal/ControlNetSeparation

Timestamp:

Mar 24, 2011 5:04:29 PM (13 years ago)

Author:

sklower

Comment:

initial version of page

Internal/ControlNetSeparation

@@ +1 @@
+= 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
+
+comments::
+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.
+
+To enable control net separation, both commands are required.
+The separation process requires about 12 seconds for a normal swapped in experiment
+(independent of the number of the nodes in the experiment!) and will take about 30 seconds for a 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.