Commit 782caff4 authored by Leigh Stoller's avatar Leigh Stoller

Updates.

parent 92eb1d5e
[This file is not kept entirely up to date.]
* Fix capserver to require reserved port number. It was always
supposed to be that (its why I picked port 855) but I forgot to
actually finish it!
* Look at using CISCO NAM modules.
* Beef up ntpstart to deal with timesys version.
* Add cron job to poll db for unverified users.
* Fix length of osname and osid to equal imageid and imagename. The
EZ image page assumes they are the same, and they aren't. They should
be.
* Fix /opt and software install inside a jail. Perhaps point /opt to
/local (outside the jail).
* Allow bw=0 in delay_config for unlimited bw.
* Fix staticroutes for 2000x2000 topos! Too many giant hashes.
* Eric wants to be able to specify units of `rate' in events to trafgen.
* Get experiment restart released, also with batch experiments.
* Tiptunnel build environment for windows.
* Analyze mysqld usage wuth swapping in big vnode experiment (using 50% CPU
as reported by Jay).
* Upgrade SFS to version 7. Try with rexd only in jails. Work with DM?
Use DNS SRV records for port?
* Windows version of ssh-mime.pl and instructions for installing.
* Fix tbacl files with IE (Jay says they do not work).
* Fix sshdport in nodes table to default to 22 instead of 11000, even
though the port is not used unless its a jailed node without its
own IP address.
* From: Mike Hibler <mike@flux.utah.edu>
To: testbed-ops@flux.utah.edu
Subject: "hardware off" experiment
Date: Fri, 17 Oct 2003 09:35:34 -0600 (MDT)
Chad Lake sent some mail today about high-temp alarms in the machine room
and how they need to adjust the threshold that triggers an alert to the
campus police.
Anyway, that got me thinking about ways that we can help out if temps
start to climb and we need to start shutting things down. While I know
that our heat generation pales compared to the SCI cluster, and that
shutting it down is a prerequisite to lowering the temps in that room,
it still wouldn't hurt to at least put this on the todo list.
One easy, not too invasive way would be to shutdown all free nodes.
At the same time, we would want to remove them from the free list so that
experiment setup didn't power cycle them and have them wind up on again.
With a couple of options it seems we could do this. At some level of
allocation (nalloc?) we have an option to allocate all free nodes to an
experiment. Then we add an option to power to work on all nodes in an
experiment (currently this is only an option in node_reboot).
Then we could do something like:
nalloc -A emulab-ops hwoff
power off emulab-ops/hwoff
Another useful, but more invasive, alternative is to shutdown all
experimental nodes. We could do this by adding an option to power
to operate on all nodes, for some appropriate definition of "all":
power -A off
which of course would prompt out the yazoo to make sure you really
mean it. Or maybe more explicitly, a "type" option which tells power
to interpret its arguments as types of machines:
power -T off pc600 pc850 pc1500 pc2000
This is mostly so that, if I get a call at 3am sometime about how
temps-are-rising-can-you-shut-stuff-down-quick, I can login and do something
quick without having to try to figure out a query to list all free nodes
or how to construct a list of all nodes and then do it. Ideally, we would
have some command that someone (not us) in the machine room could run to
do it. Maybe they login to snake's console as "killswitch" and it takes
everything down.
* Fix problem Eric reported with instance limited images and swapmod:
Eric> *** Update aborted; old state restored.
This appears to have happened because I wanted to move the instance-limited
image UAV-TSL31-NET from one node of this experiment to another. This
experiment has only one instance of UAV-TSL31-NET; the other is currently
allocated by another experiment, so there are no spares lying around.
Eric> *** /usr/testbed/libexec/assign_wrapper:
Eric> Cannot load pces-UAV-TSL31-NET on one or more nodes.
Eric> Too many nodes are already running this OSID!
Eric> *** Failed (67) to map to reality.
* From: Robert P Ricci <ricci@cs.utah.edu>
Subject: Re: Disabled Web
Date: Fri, 8 Aug 2003 11:12:35 -0600
......@@ -93,12 +200,6 @@
* Adjust web page width to actual display width the user is using.
* From: Mike Hibler <mike@flux.utah.edu>
Date: Tue, 8 Jul 2003 11:01:00 -0600 (MDT)
BTW: shouldn't we have a "node types" page where people can find out info
about the various machine types (CPU, memory, etc.)?
* From: Mac Newbold <newbold@flux.utah.edu>
Date: Tue, 8 Jul 2003 10:57:22 -0600 (MDT)
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment