Some will not take an image without being told which kind it is, and the
only way to find out was to attempt an update and read the error, which
writes to the bmc before it gets that far.
A stray trailing comma made the update detail a one element tuple, so a
firmware error printed as a python tuple. A missing status printed the whole
response dict. A failure that named no node was dropped entirely, which is how
a service data request that the server refused came out as silence and a
success exit code, and nodestorage, nodelicense and nodesupport exited zero
even when they had reported an error.
nodeconsole crashed decoding an absent screenshot, and again on the terminal
calls behind a pipe, where a log replay crashed too; refuse the terminal only
modes cleanly and dump the log when there is no terminal to replay into.
nodedefine raised a ValueError on an argument without an equals sign, and
firmware for a category the target does not describe printed usage as though
the question had been malformed.
On the server side the readability check was applied to the path a download is
saved to, so asking for service data or saved licences at a path that does not
exist yet failed claiming the destination was not readable.
Three things stopped a redfish firmware update on MegaRAC. The AMI handler
opened by asking the bmc to preserve fourteen named settings, and a build that
knows a different set rejects the whole request, which aborted the update
before anything was uploaded; send only the keys the bmc advertises. The
multipart push carried the image alone, and the specification has it carry an
UpdateParameters part too, which this firmware enforces. AMI also wants an
OemParameters part naming the kind of image, and nothing was supplying one.
The kind of image is asked for rather than worked out from the file. The
extension is vendor habit rather than format, and the leading bytes answer just
as confidently about an image they have never seen, while being wrong means a
bmc flashed with a bios image. So nodefirmware takes --type, it travels as far
as the handler that wants it, and where the bmc publishes the types it accepts,
an unknown one is refused with the list, as is asking with none. A platform
that reads the kind of firmware out of the image itself refuses the option
rather than dropping it, so nobody aims an update somewhere they did not mean
to. A parameter file still wins, since it can carry more than the image type.
Updating the bmc takes the bmc, and the task being watched, away for minutes.
That is the update working rather than the monitoring failing, so wait a
bounded while for it to answer again instead of reporting a successful flash
as an error.
For nodemedia, nodelicense, and nodefirmware, support
for expressions in filenames was
fouled when pass by
filehandle was added.
Restore this by adding all the files matching an expression.
Some platforms can have a very slow category,
like disks. Give CLI a way to ask for the desired
categories and a chance to optimize away the uninteresting.
RHEL8 will no longer tolerate implicit use of python.
For now relent to being python2, though ideally one day
it could be either.
Unfortunately, this means once code is ready for python3, we have to
probably implement build time changes for python3 enabled distros
to have different shebangs than python2 distros.
The suppression was unable to be accomplished for bash without
somehow otherwise breaking the shell. zsh and csh could be better at
one-off glob disabling though.
Markus Hilger
Jarrod Johnson
Jarrod Johnson