Use the same uptime representation as FRR does.
Example output:
DESTINATION PREF NEXT-HOP PROTO UPTIME
>* 10.0.23.0/30 0/0 e6 direct 20:10:05
192.168.3.0/24 110/1 e2 ospfv2 5d13h05m
>* 192.168.3.0/24 0/0 e2 direct 01w1d13h
Signed-off-by: Richard Alpe <richard@bit42.se>
FRR prints uptime different based on how long it is. Prior to this
commit yanger expected uptime to be on the format HH:MM:SS which it is
as long as it's shorter than a day. If it's longer the format changes,
this commit handles this.
The commit also adds unit test data triggering the different formats
(<day, <week, >week).
Side note:
cli-pretty could be made to represent the data on "FRR format" again,
which is perhaps more human friendly.
Signed-off-by: Richard Alpe <richard@bit42.se>
The ASCII art over the topology must be surrounded by .... and
is codeblock in ASCIIdoc. Since this is only used for ASCII arts
to help describe tests, border and background are white.
Fix flaky ietf-routing unit tests.
Prior to this patch, there was a unit-test race between Yanger
(sysrepo emulator) and cli-pretty, where the local system time could
change between crafting the data and printing it. This caused the CLI
output to change and the ietf-routing unit test to fail.
Fixes#668
Signed-off-by: Richard Alpe <richard@bit42.se>
In "14128047e18d statd: remove nl code and improve speed" the handling
of all interfaces where moved from statd to Yanger. This means statd no
longer tracks interfaces using Netlink, but instead relies on Yanger
to recognise and list all interfaces in the system. In the midst of
this change the ignoring of "internal" interfaces where lost.
In this commit we reintroduce this logic in Yanger and add a test case
that ensures internal interfaces isn't listed.
Fixes: 14128047e1 statd: remove nl code and improve speed
Signed-off-by: Richard Alpe <richard@bit42.se>
Yanger now has the ability to loop all interfaces (based on ip link),
this means we can run once for all interfaces.
Signed-off-by: Richard Alpe <richard@bit42.se>
Run ip link and ip addr once and handle the interface selection in
Yanger. This should be faster on systems with many ports and it
simplifies the CLI test files.
Signed-off-by: Richard Alpe <richard@bit42.se>
The test ensures that all interface types can be successfully created
and retrieved from the operational datastore. It also functions as a
regression test for issue #618 and will apply to all future interface
variants.
Fixes#622
Verify a VETH pair can be removed after a couple of dummy operations to
step the dagger generation past the initial where the pair is created.
NOTE: Infamy currenly lacks support for removing chunks of configuraion
e.g., a dut.del_config_dict(), or similar, and delete_xpath() is
not valid for configurations with dependencies like VETH pairs.
Issue #658
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Now that all routes acquired from DHCP (and ZeroConf) are installed in
the kernel FIB via Frr (staticd -> zebra -> kernel), instead of directly
to the kernel FIB, let's extend this test slightly.
Add a canary static route that is installed before the DHCP client is
even started. This canary route should be installed along with the
DHCP (classless) route(s) when the interface comes up and acquires a
DHCP lease.
We take this opportunity to also verify that the route-preference is
properly carried over from test to operational RIB status.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Latest YANG model change reverts back to using the IETF routing-protocol
type. Update tests to check for infix-routing:ospf -> ietf-ospf:ospfv2.
Also, refactor and extend infamy/route.py:_exist_route() slightly for
readability and PEP-8, as well as to use an optional model prefix also
for next-hop-list addresses.
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
- Add check for specific 169.254/16 adddress (new feature)
- Add check for disabling autoconf on interface (regression test #638)
Signed-off-by: Joachim Wiberg <troglobit@gmail.com>
Today Infix only support OSPFv2 and Static routes, this hides all other
protocols in the CLI and makes it clearar about what we support in the YANG
model.
There's a case where the first query is lost due to the network just
starting up. To handle this we introduce a retry loop in the test case
that waits for a second query message to be sent.
In this commit we also lower the query interval to speed up the case
when the first message is lost. Ideally mcd (the underlying multicast
daemon) should implement a Startup Query Interval to send it even
faster.
Signed-off-by: Richard Alpe <richard@bit42.se>
Richard Alpe