confluent_client and confluent_server rpms took the PEP 440 version
setup.py needs, 4.0.1.dev5+g..., which rpm sorts after 4.0.1: a
released package did not upgrade a development build of it. Stamp the
rpm with the ~dev form the other packages already use and keep .dev only
for setup.py and the sdist name. The debs already sort right, since
stdeb rewrites .dev to ~dev.
Python 3.13 dropped crypt from the stdlib, and openSUSE Leap 16 packages
neither legacycrypt nor crypt_r, so the server did not start there
without a pip install. Both of those only call crypt(3) from libcrypt,
which every distribution ships, so do that through ctypes as the last
resort, serialized because crypt(3) returns a static buffer. Its hashes
match the stdlib module byte for byte.
Move the import chain both callers carried into confluent.pwcrypt. The
legacycrypt Recommends is no longer needed.
Neither is packaged by EPEL or Ubuntu, so the hard dependency made the
server uninstallable there without --nodeps or a stub package. Both are
already optional in the code: httpapi drops passkey support when
webauthn does not import, and the Lenovo IMM2 BIOS settings path reports
the missing module as unsupported functionality. Recommend them instead,
as the SUSE branch already does for webauthn.
See xcat2/confluent#192.
For one, at the time an XCC emits a NOTIFY, it's DeviceDescription.json is also very sluggish. Extend timeout
knowing this may be delayed.
Further, if snoop comes back with only one address family, do not
perturb existing address families.
- A manager that is not a confluent node is pinned on the guest's node
(pubkeys.tls_hardwaremanager), as the console already does.
- user@realm!tokenid with the token secret is accepted as the login;
VNC needs a password login and says so.
- Route every API call through one checked path: non-2xx answers raise
with PVE's message, a 401 logs in again, a migrated guest is re-mapped.
- Report login failures other than 401 instead of a TypeError.
- Power changes are bounded and reported per node; 'on' when on is a
no-op; every live QEMU state counts as on.
- Read and set the boot order in every form PVE accepts, keeping disks.
- Stop toggling Content-Type on the shared client headers.
- Inventory from the current config, all SMBIOS forms.
- Refuse ambiguous guest names; skip templates.
- Keep the serial console alive and report how it ends.
- Let the self-test run without a config manager.
asyncssh relationship with cryptography can be messy, but it's not up to us, so discard those.
Change osdeploy to not use fork to avoid that deprecationwarning.
The vncwebsocket URL used the name of the Proxmox node running the
guest as host. That name often does not resolve on the management
node, and the fingerprint checked is the one of the manager anyway.
Connect to the manager instead, which forwards the websocket to the
right node.
Some devices have multiple BMCs, but we don't know
the 'right' one.
In such a scenario, search for BMC NICs for any that match the address we are using.
In the process of reworking cache logic, a temporary disabling was left in.
Before the logic change, 'cache' of '1' meant 'True', but the intent was a max age to decline to reply. This allows very close together requests to be cached.
Proivde support for API keys. Users can register by doing a POST like:
/confluent-api/sessions/current/apikey/create with body of '{"expiration":null}
API keys can be reset by doing a POST to /confluent-api/sessions/current/apikey/revokeall
A secret is provisioned per user to allow them to clear all keys, while remaining mostly stateless to avoid having to manage every key server side.
The password change and the account update set If-Match: * on the
connection itself, and after a forced password change that connection
is handed back, so every later request carried it. Pass the header
with the PATCH alone. Also drop an unused asyncio import.
get_health looked only at each node's State, so an Enabled node whose
Health was Warning or Critical left the chassis reported as ok. It also
skipped a node whose resource could not be read, counting it as
healthy while an unreadable chassis was flagged. Report the node's
Health when it is not OK, and a node that cannot be read as
Unreachable.
The sensor collection has 668 members, and reading each one on its own
takes about 92 seconds and rebuilds on every nodesensors call. Opt in
to $expand=. for that collection, checking once whether the firmware
really inlines the members, so firmware that ignores $expand keeps the
per-member reads.
ForceRestart only restarts the host, so the node BMC kept running and
a reseat did not recover a hung one. The EUREKA firmware provides a
Reseat reset type that removes all power from the slot.
A node whose BMC is not reporting returns Reading 0 from its
temperature sensors, dragging averages down with meaningless
values. The ComputerSystem Oem data flags this via HasBMCMetrics;
skip such nodes, and keep collecting when the flag is absent.
get_health collected an issues list but returned it nowhere, and
flattened every problem to Warning. Return the findings as
badreadings using SensorReading, honor verbose, and map chassis
health through _healthmap so Critical is no longer downgraded.
The sensor URL was built as BMC{N}CpuCPU{X}Temp instead of
BMC{N}CPU{X}Temp, so every request returned 404. The entries also
referenced const.SensorUnits, which does not exist, and used a
dict shape the only consumer, get_average_processor_temperature,
cannot read - it expects thermal-style dicts with ReadingCelsius.
util.json_loads does not exist, so the PasswordChangeRequired flow
raised AttributeError on every 401, silently swallowed by the blanket
except. Use json.loads, which accepts the bytes body directly.
ManagedBy is optional, and when a system does not link a manager
get_bmcurl() answers None. Most callers passed that straight into a
web request, which failed with "Constructor parameter should be str"
from the url library. Route those callers through bmcinfo(), which now
raises UnsupportedFunctionality so confluent reports it plainly. The
event log falls through to its system and chassis fallback, and
list_media lists nothing, since neither needs a manager.
SensorReading computed health and states with .get() and then
overwrote both with a strict lookup, so a sensor whose Status lacks
Health, or carries a value outside the health map, raised KeyError
and aborted the whole sensor listing. Keep the tolerant lookup, and
drop the states of a reading that is OK, as the copy in command.py
already does.
Markus Hilger
Jarrod Johnson
Laurence
Jarrod Johnson