GitLab

devices with network interfaces. Emulab's spp and bbg nodes are
examples, but I did all that by hand inserting sql. An spp node is a
shared node with some interfaces. Users can allocate one or more of
those interfaces and establish vlans to the interfaces. The node is a
"fakenode" in "shared" mode, and everything else falls out. The mapper
assigns virtual nodes until all of the interfaces are allocated,
snmpit does its work on the interfaces, and the user then does the
rest.

Anyway, to added a special device:

  boss> wap addspecialdevice -s -t goober goober1

The -t argument is the name of the node type, created if it does not
exists. The last argument is the name of the fakenode to create in the
DB. The -s option says the special device is shared. Without -s, the
device is allocated exclusively.

Then to add interfaces to the device:

  boss> wap addspecialiface -b 1Gb -s cisco4,100,100 goober1 eth0

The -b option is the speed (either 100Mb or 1Gb). The -s option is the
switch side of the interface (switchname,card,port). The last two
arguments are the nodename and iface name for the interfaces table.

After the interface and wires table entry are added to the DB, snmpit
is called to put the switch port into tagged mode (if the node is
shared). To skip the snmpit step, add the -t option.

ilopswd.

Validate those SQL args!

NOTE: we also ensure that the reporting node is listed as a legit tip server
in the tipservers table. This means that capture may stop working on nodes
whose servers are not in the table! SQL update 291 will add any servers
listed in tiplines entries that are not in tipservers to prevent this breakage.

lengthy comment in the code (text from Ryan).

the parent osid, and can be used to select a different image to load,
sau FEDOR15-OPENVZ-STD. Does not work on shared nodes of course.

them to KEY.

two different domain tags. Confused Firefox 12.

Also, when calling with -A (showall) and -n (number of records),
lets not read the entire 5 million rows and then throw away everything
but the first 20. Takes about an hour.

In general this script needs some work.  Would also be nice to page
through records by specifying an index to start with.

The problem is with the chunking that apache does to avoid having
to read the entire response, so it can set the Content-Length header
properly. I have no idea who apache does chunking, but on the 9.0
client it was getting confused and trash characters were leaking into
the data. The trash is probably the result of the client getting out
of sync. 

Anyway, I do not know who is at fault (client code or our ancient
apache server), but after many hours of screwing around I found out
how to prevent the client from getting chunked data from the server.

Basically, you have to force the request to be sent as HTTP/1.0
instead of 1.1, which tells apache to not chunk the data. In general
this is not a good thing to do, but in this case the RPCs are pretty
small and not on a critical performance path. We can revisit when we
upgrade our boss to apache 2.X ...

Add interface/host for foamvm

be true, which requires a file back to $BOOTDIR after running prepare.

In testmode, pretends the ref was just created.

Renamed old '-C' option '-c'

update/join was not doing what I expected, and it looks like I cannot
actually do what I was intended cause mixing multiple table update and
left join does not work like I expect (read: I cannot make any sense
of the mysql manual). So, recast as a multiple table update without
the join, over individual rows in the vinterfaces table.