<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
</head>
<body text="#000000" bgcolor="#FFFFFF">
<p>Hi,</p>
<p><br>
</p>
<div class="moz-cite-prefix">On 12/12/2018 10:02 AM, Anton Degtyarev
wrote:<br>
</div>
<blockquote type="cite"
cite="mid:CAOFW4Os1BKCSkGqSwobURpsciOX_E9KA9y+0+zG1_+mwcu4u0A@mail.gmail.com">
<meta http-equiv="content-type" content="text/html; charset=UTF-8">
<div dir="ltr">Hi Donald,
<div><br>
</div>
<div>Yes, Renato had contacted me about this as well. Thank you
for adding the needed tests. I will redo my tests in the lab I
had prepared for a customer use case.</div>
<div><br>
</div>
</div>
</blockquote>
<p>please also run bgp_l3vpn_to_bgp_vrf -- this actually would have
caught the breakage in the first place, but wasn't being run as
part of the original CI.</p>
<p><br>
</p>
<blockquote type="cite"
cite="mid:CAOFW4Os1BKCSkGqSwobURpsciOX_E9KA9y+0+zG1_+mwcu4u0A@mail.gmail.com">
<div dir="ltr">
<div>The fix I had submitted was needed for locally defined
subnets in a VRF which needed to be leaked into L3VPN. The
issue I had observed was that they weren't -- because zebra
thought that these subnets were not reachable. My theory
(which worked for me and the customer) was that zebra was
treating these prefixes wrong what, I thought, the patch
corrected. Apologies for a high-level response -- I do not
remember all details as this was a month ago.</div>
</div>
</blockquote>
<p>I suspect that there are types/sub-types that need to be added to
the code if you make the original change...<br>
</p>
<p><br>
</p>
<blockquote type="cite"
cite="mid:CAOFW4Os1BKCSkGqSwobURpsciOX_E9KA9y+0+zG1_+mwcu4u0A@mail.gmail.com">
<div dir="ltr">
<div><br>
</div>
<div>I will recreate the design I was testing again with the
recent FRR master code and will give the new topotests a try
as well.<br>
<br>
</div>
</div>
</blockquote>
<p>Please also run the old one (the CI problem has been corrected
and the test is now being included.)</p>
<p><br>
</p>
<blockquote type="cite"
cite="mid:CAOFW4Os1BKCSkGqSwobURpsciOX_E9KA9y+0+zG1_+mwcu4u0A@mail.gmail.com">
<div dir="ltr">
<div><br>
Best regards,</div>
<div>Anton</div>
<div><br>
<br>
<div class="gmail_quote">
<div dir="ltr">On Wed, 12 Dec 2018 at 17:30, Donald Sharp
<<a href="mailto:sharpd@cumulusnetworks.com"
target="_blank" moz-do-not-send="true">sharpd@cumulusnetworks.com</a>>
wrote:<br>
</div>
<blockquote class="gmail_quote" style="margin:0px 0px 0px
0.8ex;border-left:1px solid
rgb(204,204,204);padding-left:1ex">Anton -<br>
<br>
Recently you had e23b9ef6d2 committed into FRR. During
subsuquent<br>
testing it was noticed that this change broke various
forms of<br>
route-leaking. In order to preserve current
functionality, we have<br>
backed this commit out. My apologies for not catching
these issues<br>
earlier on initial submission. In the meantime, we've
added new tests<br>
to the topotests to catch this problem from happening in
the future(<br>
see tests/topotests/ bgp-vrf-route-leak-basic and
bgp_l3vpn_to_bgp_vrf<br>
).<br>
<br>
Lou and I believe that your initial approach was probably
the right<br>
thing to do but it needs to not break existing
functionality. What<br>
was the use case you were needing this change for?<br>
<br>
donald<br>
</blockquote>
</div>
</div>
</div>
</blockquote>
</body>
</html>