On 1/31/23 17:30, Mike Pattrick wrote:
On Tue, Jan 31, 2023 at 10:45 AM Phelan, Michael
<[email protected]> wrote:
-----Original Message-----
From: Ilya Maximets <[email protected]>
Sent: Tuesday 31 January 2023 12:23
To: Phelan, Michael <[email protected]>; Mike Pattrick
<[email protected]>; Adrián Moreno <[email protected]>
Cc: [email protected]; [email protected]; Stokes, Ian
<[email protected]>
Subject: Re: [ovs-dev] [PATCH v2 1/2] ofproto-ipfix: use per-domain template
timeouts
On 1/31/23 13:15, Phelan, Michael wrote:
-----Original Message-----
From: Ilya Maximets <[email protected]>
Sent: Friday 27 January 2023 20:03
To: Mike Pattrick <[email protected]>; Adrián Moreno
<[email protected]>; Phelan, Michael <[email protected]>
Cc: [email protected]; [email protected]; Stokes, Ian
<[email protected]>
Subject: Re: [ovs-dev] [PATCH v2 1/2] ofproto-ipfix: use per-domain
template timeouts
On 1/25/23 06:34, Mike Pattrick wrote:
On Tue, Jan 24, 2023 at 2:21 PM Adrián Moreno <[email protected]>
wrote:
From: Adrian Moreno <[email protected]>
<snip>
+dnl Check templates (Flow ID = 0x02) where sent to ObservationDomain
1.
+AT_CHECK([test $(tcpdump -r ipfix.pcap --count "udp port 5555 and
+udp[[20:4]] = 0x00000001 and udp[[24:2]] = 0x02" 2>/dev/null | tr
+-d "packets") -gt 0])
These still seem to be getting "test: -gt: unary operator expected"
errors on Intel CI, even after the modification. I'm at a loss as to
why, the error message makes it seem like tcpdump is choking on the
pcap and not writing anything to stdout. But the pcap should be
present and readable.
Yeah, I can't reproduce the failure happening on Intel CI as well.
Michael, do you know what is happening with this test?
Hi Ilya,
I have reran the test and received the same failure as reported on the mailing
list, I'm not sure what exactly is going on with this test, maybe there is some
prerequisite missing from our machine?
Thanks for checking.
It doesn't look like something is missing...
Could you provide the content of testsuite directory after the test failure?
i.e. the ipfix.pcap file and other stuff. This may shed some light.
Best regards, Ilya Maximets.
Sure, I've attached a zipped folder containing the contents of the testsuite
directory after the test failure.
Let me know if you have any more questions.
It looks like --count is a recent feature, only added in 2020. So any
copy of tcpdump preceding that will error out. This could be why the
test fails on CI?
Thanks Mike, Ilya and Michael for following up on this while I was on a sick
leave.
I think we can follow Ilya's suggestion and remove the unit testing from the
first patch so it can be backported and then add a small program that parses
IPFIX packets and prints some information. Maybe we can even reuse something
from test-netflow.c. But this will be a bigger patch (currently header
definitions are private) so it's better to do it separately.
Thanks.
--
Adrián
-M
Kind regards,
Michael.
Let me know if you need any more information.
Thanks,
Michael.
OTOH, it might be better to take the road similar to how we test
sflow and netflow. I just found specific test binaries test-sflow
and test-netflow which are just listening and printing out packtes in
human-readable format. If we had something like this for ipfix, we
could write similar unit tests that do not require any external tools or root
privileges, as we do for sflow and netflow.
Thoughts?
We might split this bug fix from the implementation of the unit test
though, since it will require a bit more code.
Best regards, Ilya Maximets.
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
