GitLab

"server 127.127.xxx.xxx ..."

ntp each time.

first talking to tmcd to get the ntpdrift and ntp server/peer lists.
This is a strict wrapper so it should be invoked from /etc/rc.conf
like this:

	xntpd_program="/usr/local/etc/emulab/ntpstart"
	xntpd_flags="/usr/sbin/ntpd -p /var/run/ntpd.pid"

which is to say that this program passes it entire argument list off
to the shell once it sets up the config file. We fire off ntp no
matter what happens though, and allow for no actual changes to the
config file if tmcd does not provide anything.

TMCD: Add ntpinfo and ntpdrift commands, as per Mike's specification.
ntpinfo returns lines like:
	SERVER=nnn.nnn.nnn.nnn
	...
	PEER=nnn.nnn.nnn.nnn
	...
	DRIFT=nnn.nn

and "ntpdrift nnn.nn" allows the current drift setting to be placed
into the DB, although thats not currently happening. In the DB,
ntpdrift is a new float slot in the nodes table, that defaults to
NULL. The servers/peers are stored in a new table that looks like
this:

	+---------+---------+--------+
	| node_id | IP      | type   |
	+---------+---------+--------+
	| wa32    | 1.1.1.1 | server |
	| wa32    | 1.1.1.2 | server |
	| wa32    | 1.1.1.3 | peer   |
	| wa32    | 1.1.1.4 | peer   |
	+---------+---------+--------+

Add ntpinfo table for storing server/peer IPs for each node.
These values are to be returned by tmcd.

devices in jail

Add some missing os_features options and op_mode, along with guidelines for setting them properly. Add them to form handler and display as well.

Fix small problem that was causing a failure in the test suite: If there are no nodes in the expt, don't die(), just exit().

Fix last fix to nfree... turns out that if you alias a table ('from nodes as n') that the old name ('nodes') is no longer available, and you _must_ use the new alias ('n') to refer to it.

Add an 'osid' field to the nodes table. This will store the currently running OS. The only time it gets updated is by (the new) stated when the node goes into the BOOTING state (sent by the new proxydhcp when a node asks for its info.) Without this, we have no way of knowing what OS we told the node to boot last time it came up.

not already exist).

well (requires new libsetup.pm to be installed too). Still a work in
progress.

sillyness.

headers in order for the new assign to function properly.

Add a new 'patches' directory to store these, and other patches.

we didn't start.

- Increased the point penalty for using a link
  which is marked as -1 ("not-connected") in the physical matrix.
  (Was 10,000; is now 100,000.)

- Now we actually report an error, and return errorcode 6,
  if the final reported solution uses any not-connected links.

jailed environments.
Fix up sfs stuff so that warnings and errors from sfskey do not
get printed out on the console.
Add a jailedsetup stub for testing with jails; work in progress.

Remove the bootinfo_ack functionality. Nothing uses it and it couldn't properly do the db changes it needed to.

getting into the system.

Remove -e flag from calls to power. node_reboot sends an event only when ssh reboot or ipod are successful in rebooting the node, and only calls power when they are not successful. So an event should be sent by power every time node_reboot calls it. This explains some of the problems we were having with tons of email from stated about invalid transitions: since the state changes weren't always happening, it appeared to skip over states.

1) Set 'BOSSNODE' in the defs file to `hostname` - required for the
   event system
2) Include virtual nodes, and randomize order, when picking nodes
   to run on
3) Run exports_setup after allocating nodes, so they can mount
   homedirs, etc.
4) Enable events in full mode
5) When creating experiment in experiments table, set gid and testdb
6) Run our own copy of stated, so that transitions get recorded

mode.

tbswapin from starting an event scheduler for the experiment.

For use with the testsuite in 'full' mode.