VRF-HLD document #242

shine4chen · 2018-09-04T07:54:08Z

VRF high level design for SONIC

msftclas · 2018-09-04T07:54:19Z

All CLA requirements met.

nikos-github · 2018-09-17T20:45:18Z

doc/sonic-vrf-hld-v0.2.md

+
+blackhole = BIT ; Set to 1 if this route is a blackhole (or null0)
+
+Since global vrf name is null, global vrf key will becomes ROUTE_TABLE:prefix.


I suggest you use ROUTE_TABLE:DEFAULT:prefix or ROUTE_TABLE:GLOBAL:prefix or even better, the tableid.

nikos-github · 2018-09-17T20:52:00Z

doc/sonic-vrf-hld-v0.2.md

+Adding the following logic:
+
+-   the Key of NextHop now is changed from only ipaddress to a pair of
+    (ipaddress, interface_name)


Should be (ipaddress, tableid) and not (ipaddress, interface_name). Expect to be doing a lot of conversions and lookups if you leave as (ipaddress, interface_name).

can we point a route in vrf red to a nexthop (neighbor) in vrf blue? @prsunny , do we need to change the nexthop format in app db?

@lguohan Yes you can have routes in vrf red with nexthops in vrf blue.

@nikos-github , yes it is possible but the format in APP_DB is same for route since every route has a nexthop IP with the interface name. Currently orchagent doesn't look for the interface which requires a fix.

@nikos-github When set neigh to SAI layer (ipaddress, interface_name) would be more comfortable. And relatedAPP_DB/ASIC_DB needn't be changed.

nikos-github · 2018-09-17T20:59:06Z

doc/sonic-vrf-hld-v0.2.md

+>   routerCnt  ,neighCnt and rifCnt  is decreased to zero.
+
+>   When device binds to specified VRF, the ip address of slave device will be
+>   removed and kernel will delete all neigh associated slave device.


I believe on enslavement, connected and local routes are automatically moved to the table associated with the VRF device. Is that not the case with the latest kernel? Can you explain how you plan to achieve the above removal/behavior you mention?

jipanyang

Could you update the MD file and make it more structured and hierarchical? It is difficult to read in current format.

lguohan · 2018-09-19T05:40:55Z

@jipanyang , I have re-formatted the doc, please comment.

shine4chen · 2018-09-20T06:37:28Z

@lguohan thanks for help to format md file.

lguohan · 2018-09-20T07:17:55Z

why close this PR?

shine4chen · 2018-09-20T07:23:51Z

@lguohan sorry It is missoperation. reopen again.

nikos-github · 2018-10-02T22:55:10Z

doc/sonic-vrf-hld-v0.2.md

+
+```
+"VRF": {
+    "VRF-blue": {


Any reason why you are using VRF-blue instead of vrf-blue?

Any reason why you are using VRF-blue instead of vrf-blue?

I think vrf-blue is okay. I will revise it @nikos-github

nikos-github · 2018-10-02T23:13:39Z

doc/sonic-vrf-hld-v0.2.md

+```
+"VRF": {
+    "VRF-blue": {
+        "fall_through":"1" //enable global fib lookup while vrf fib lookup missed


Please change fall_through to fallback. The term fall_through tends to be used for routemaps but for vrfs the industry established term is fallback.

There should be no restriction that the fallback be the global/main table even if that may be the only one we plan to support (and I don't see why we plan to support only global/main as fallback).

Your document needs to specify how the fallback is specified. I assume the table id will be used since linux doesn't have a name for the global/main table.

@nikos-github Fallback term is better.I will revise it. The fallback global feature is very useful for specify VRF user access internet through global/main route which is defined by RFC4364. Some enterprise user still use this to access internet on vpn environment. Different vrf route leak is beyond this document scope.

nikos-github · 2018-10-02T23:34:44Z

doc/sonic-vrf-hld-v0.2.md

+list of 0 or more ip prefixes;
+
+;Status: stable
+key = INTF_TABLE:ifname:IPprefix ; an instance of this key will be repeated for


What is the advantage of having multiple keys?

Please provide how the json looks for this.

@nikos-github There is two reason why break up app-intf-table into app-intf-table and app-intf-prefix-table.

Multiple ip addresses can be configured on one interface.So I put interface common attribute into app-intf-table and keep ip-prefix specific attribute on app-intf-prefix-table.

Interface can be put to specific VRF before ip address is configured on it.

jason file looks like below:
"Ethernet1":{
"mtu":"1500”,
"vrf":"vrf-red"
},
"Ethernet0|11.11.11.1/24": {},
"Ethernet1|12.12.12.1/24": {},
"Ethernet2|13.13.13.1/24": {},
"Ethernet3|14.14.14.1/24": {},

1. Support loopback interface as part of VRF interface. 2. In SAI , VR is treated same ad VRF, so in SONiC code VRF will be implemented as VR. 3. Consider the configuration dependency between VRF binding event and interface ip address configuration. The proposal is to add VRF name before the ip address configuration. 4. When cold restart, vrfMgrd will flush all vrf related setting in linux kernel. 5. In the overall diagram, make the color scheme consistent for all modified module. 6. Add VRF show command. 7. Will not delete VRF object until all routerintfs related to VRF object have unbind from this VRF by user.

…h-pages

rebase to master

…h-pages

This reverts commit c0a86c3.

shine4chen · 2019-01-22T10:36:22Z

@lguohan @prsunny @jipanyang please help to approve this document if no more question.

shine4chen · 2019-01-23T05:58:13Z

Nephos plan to submit this feature include warm-reboot support on 201906 branch. @lguohan @prsunny @jipanyang

prsunny · 2019-01-31T03:07:08Z

Closing in favor of this commit

Add VRF-HLD document

db5b9d1

nikos-github reviewed Sep 17, 2018

View reviewed changes

jipanyang reviewed Sep 18, 2018

View reviewed changes

formating

934a4fe

shine4chen closed this Sep 20, 2018

shine4chen reopened this Sep 20, 2018

jleveque mentioned this pull request Oct 2, 2018

dhcp-relay: dhcrelay daemon is permantly blocked and not starting sonic-net/sonic-buildimage#2105

Closed

nikos-github suggested changes Oct 2, 2018

View reviewed changes

prsunny mentioned this pull request Oct 15, 2018

[intfmgrd]: Merge intfsyncd into intfmgrd sonic-net/sonic-swss#635

Merged

shine and others added 7 commits January 18, 2019 01:21

Merge branch 'gh-pages' of https://github.com/shine4chen/SONiC into g…

d876705

…h-pages

add mclag-hld image

c0a86c3

Merge pull request #1 from Azure/gh-pages

ca41afc

rebase to master

Merge branch 'gh-pages' of https://github.com/shine4chen/SONiC into g…

c4bfa7e

…h-pages

Revert "add mclag-hld image"

f3ae888

This reverts commit c0a86c3.

lguohan approved these changes Jan 31, 2019

View reviewed changes

prsunny closed this Jan 31, 2019

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

VRF-HLD document #242

VRF-HLD document #242

shine4chen commented Sep 4, 2018

msftclas commented Sep 4, 2018 •

edited

Loading

nikos-github Sep 17, 2018 •

edited

Loading

nikos-github Sep 17, 2018

lguohan Sep 19, 2018

nikos-github Sep 20, 2018

prsunny Sep 20, 2018

shine4chen Sep 20, 2018

nikos-github Sep 17, 2018

jipanyang left a comment

lguohan commented Sep 19, 2018

shine4chen commented Sep 20, 2018

lguohan commented Sep 20, 2018

shine4chen commented Sep 20, 2018

nikos-github Oct 2, 2018

shine4chen Oct 8, 2018

nikos-github Oct 2, 2018

shine4chen Oct 8, 2018

nikos-github Oct 2, 2018 •

edited

Loading

shine4chen Oct 8, 2018

shine4chen Oct 8, 2018

shine4chen commented Jan 22, 2019

shine4chen commented Jan 23, 2019

prsunny commented Jan 31, 2019


		blackhole = BIT ; Set to 1 if this route is a blackhole (or null0)

		Since global vrf name is null, global vrf key will becomes ROUTE_TABLE:prefix.

VRF-HLD document #242

VRF-HLD document #242

Conversation

shine4chen commented Sep 4, 2018

msftclas commented Sep 4, 2018 • edited Loading

nikos-github Sep 17, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jipanyang left a comment

Choose a reason for hiding this comment

lguohan commented Sep 19, 2018

shine4chen commented Sep 20, 2018

lguohan commented Sep 20, 2018

shine4chen commented Sep 20, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

nikos-github Oct 2, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

shine4chen commented Jan 22, 2019

shine4chen commented Jan 23, 2019

prsunny commented Jan 31, 2019

msftclas commented Sep 4, 2018 •

edited

Loading

nikos-github Sep 17, 2018 •

edited

Loading

nikos-github Oct 2, 2018 •

edited

Loading