netOrk keeps one list of NTP servers and a time zone per site and aligns
a device that differs (NetOrk/netork#2360). A Windows host takes part
through w32time and its time zone, in one PowerShell script per read.
The servers and the mode are read from w32time's registry values, group
policy's first, not from w32tm /query /configuration, whose text is in
the host's language; synchronised is the leap indicator on the first
line of w32tm /query /status, which reads the same in any language. The
servers are written the way w32tm /config /manualpeerlist
/syncfromflags:manual writes them (NtpServer with 0x8, Type NTP),
followed by w32tm /config /update: no value passes a native command
line, whose quoting Windows PowerShell 5.1 mangles. A domain member that
takes its time from the domain hierarchy (NT5DS, AllSync) and a host
whose w32time comes from group policy refuse the write: their time is
set elsewhere, and a local write would be overridden or break the
domain's time.
Windows keeps a Windows zone, which stands for several IANA zones, so
Europe/Vienna written would read back as Europe/Berlin and the poll
would align the host forever. The IANA name written is kept beside it
in HKLM\SOFTWARE\napalm-windows and read back while the host is still
in that Windows zone. The zone is read from the registry, which
[TimeZoneInfo]::Local caches for the life of the remote session.
The zone table is now CLDR's windowsZones.xml (release 49) both ways,
generated by tools/windows_zones.py with tzdata's links and committed
with its source and checksum; nothing is fetched at runtime. CLDR's old
names (Asia/Calcutta) read as their current ones (Asia/Kolkata).
Writes the host already has are skipped. Needs napalm-device-types
4.35.0.
Refs NetOrk/netork#2367
netOrk shows and controls every scheduled job of a host; on Windows that
is the Task Scheduler. The driver now implements napalm-device-types'
ScheduledJobsMixin over PowerShell Remoting, one script per call as every
getter does.
get_scheduled_jobs reads every task with its run info in one script
(Get-ScheduledTask, Get-ScheduledTaskInfo) as task:<path><name>, its
triggers summarised. Tasks under \Microsoft\ are the platform's, others
the admin's: Windows does not record which software installed a task.
The time zone is mapped to IANA for the common Windows zones, else None.
A reading cut short, or holding fewer tasks than it counted, raises.
manage_scheduled_job switches the task's own flag or starts it;
Start-ScheduledTask returns once the task has started, so the result says
that and no more. A disabled task, or one that refuses a start on demand,
is not offered run.
put_scheduled_job registers an application's task \<name> in the root
folder from XML, so a create and a change write the whole definition the
same way: the command runs in cmd.exe /d /s /c, the principal is SYSTEM
or a local user without a stored password (S4U), and the cron expression
becomes triggers only where Task Scheduler can say it exactly; any other
shape is refused, naming why. The description keeps the job's name and
cron expression, so the reading gives back what was written, and the
digest also covers the triggers, so a change by hand is seen.
delete_scheduled_job unregisters such a task and no other. Every write
reads the task again first, and every value goes into PowerShell
single-quoted with its quotes doubled.
Requires napalm-device-types 4.26.0.
Closes#8
netOrk runs every driver in its own containers on Python 3.12, so 3.10
and 3.11 are no longer supported or tested. requires-python, mypy and
ruff move to 3.12, along with what ruff's pyupgrade rules then rewrite.
- CI moves to .gitea/workflows/ci.yml and calls the shared driver workflow
of NAPALM/driver-standard; .github/ is gone.
- requires-python >=3.10 (napalm-device-types needs it), LICENSE file,
py.typed, setuptools>=77 with the SPDX license string, Issues and
Changelog links on git.netork.io, napalm-device-types>=4.21.1 from PyPI,
floors on the dev extra.
- mypy strict for the whole package, no module exempt: optional_args is
dict[str, Any] | None.
- tests/quality_baseline.json records get_facts and get_route_to, over
complexity 10. Coverage floor 95 % (measured 95.67 %).
- CHANGELOG.md in Keep a Changelog form; the README says no host output is
recorded yet, links napalm-device-types in the NAPALM org and pypsrp on
PyPI.
- get_route_to: "0.0.0.0" and "::" are the unspecified next hop of an
on-link route as the host reports it, not an address anything binds to
(nosec B104).
Mechanical: the rule set every NAPALM driver uses (E, F, I, UP, B at line
length 100, Python 3.10), with E501 off under tests/. The code was already
formatted and clean under it; what ruff cannot fix comes in the next commit.
The same job as napalm-fritzbox and napalm-opnsense: Python 3.10, 3.11 and
3.12, pytest, then wheel and sdist. napalm-device-types comes from
git.netork.io first, because PyPI has an unrelated package of that name.
One script: Get-NetTCPConnection -State Listen, Get-NetUDPEndpoint, the
processes holding them and Win32_Service's PIDs. A process running one
service names it; a shared svchost runs several, and which of them holds
the socket Windows does not say, so it names none. http.sys sockets
(WinRM, IIS, SMB) belong to the System process.
The fixture is synthetic: no Windows host was at hand. tools/harvest.py
records the real output. For netOrk#673.
Talks PSRP over WinRM via pypsrp, one runspace pool per connection. Each
getter sends one PowerShell script that projects cmdlet results onto flat
fields and ends in ConvertTo-Json, so the Python side parses JSON, not text.
Covers facts, interfaces, IP addresses, ARP, routes and services, plus
service start/stop/restart/enable/disable. Service names are validated and
quoted as PowerShell verbatim strings, typographic quotes included.
Fixtures are synthetic: they pin down the JSON the scripts are designed to
emit. tools/harvest.py records the real output from a host.
Refs christianmanivong/netork#300