Port configuration incremental update support #2305

Junchao-Mellanox · 2022-06-02T07:07:28Z

What I did

portsyncd no longer handle port configuration change and put it to APP DB
Implement incremental configuration change in portmgr
Adjust portsorch to meet incremental configuration change requirement

Why I did it

portsyncd and portmgrd both handles port configuration change, this causes unnecessary handling in ports orch

I set the mtu to 1500 and search the log:
 
admin@sonic:~$ sudo config interface mtu Ethernet0 1500                                                                                                                                                                            
admin@sonic:~$ sudo cat /var/log/syslog | grep portsyncd
 
 
Log shows:
 
May 31 07:41:25.275792 sonic NOTICE swss#portsyncd: :- handlePortConfig: Configure port Ethernet0 attribute admin_status value to up                                                                                               
May 31 07:41:25.275792 sonic NOTICE swss#portsyncd: :- handlePortConfig: Configure port Ethernet0 attribute alias value to etp1                                                                                                    
May 31 07:41:25.275792 sonic NOTICE swss#portsyncd: :- handlePortConfig: Configure port Ethernet0 attribute index value to 1                                                                                                       
May 31 07:41:25.275792 sonic NOTICE swss#portsyncd: :- handlePortConfig: Configure port Ethernet0 attribute lanes value to 0                                                                                                       
May 31 07:41:25.275792 sonic NOTICE swss#portsyncd: :- handlePortConfig: Configure port Ethernet0 attribute mtu value to 1500                                                                                                      
May 31 07:41:25.275792 sonic NOTICE swss#portsyncd: :- handlePortConfig: Configure port Ethernet0 attribute speed value to 25000

portmgrd shall do increment configuration update

How I verified it

manual test
regression test
added new unit test case

Details if related

…ne may directly change APP DB

Conflicts: tests/mock_tests/mock_table.cpp

Conflicts: tests/mock_tests/Makefile.am

Junchao-Mellanox · 2022-06-02T09:23:36Z

/azpw run Azure.sonic-swss

mssonicbld · 2022-06-02T09:23:38Z

/AzurePipelines run Azure.sonic-swss

azure-pipelines · 2022-06-02T09:23:47Z

Azure Pipelines successfully started running 1 pipeline(s).

liushilongbuaa · 2022-06-06T05:15:26Z

/azp run Azure.sonic-swss

azure-pipelines · 2022-06-06T05:15:36Z

Azure Pipelines successfully started running 1 pipeline(s).

Junchao-Mellanox · 2022-06-07T00:38:13Z

/azpw run Azure.sonic-swss

mssonicbld · 2022-06-07T00:38:15Z

/AzurePipelines run Azure.sonic-swss

azure-pipelines · 2022-06-07T00:38:24Z

Azure Pipelines successfully started running 1 pipeline(s).

Junchao-Mellanox · 2022-06-13T06:42:00Z

/azpw run Azure.sonic-swss

mssonicbld · 2022-06-13T06:42:02Z

/AzurePipelines run Azure.sonic-swss

azure-pipelines · 2022-06-13T06:42:11Z

Azure Pipelines successfully started running 1 pipeline(s).

prsunny · 2022-06-17T01:08:12Z

cfgmgr/portmgr.cpp

+             * We should not call any ip command if portOk=false. However, it is
+             * valid to put port configuration to APP DB which will trigger port creation in kernel.
+             */
+            bool portOk = isPortStateOk(alias);


I dont think we need to change this? portsyncd does the handlePortConfigFromConfigDB . So portmgrd can still wait and keep the current flow.

Hi @prsunny , handlePortConfigFromConfigDB is only called once at init stage. It cannot handle case like dynamic port breakout. So, I suppose we need keep this change. What do you think?

ok. so the logic here retries for some part of the configuration (kernel), but write to app_db in the first place.

My concern is, developer has to make sure kernel writes cannot be attempted if portOk is false which is different from the previous logic. Secondly, it keeps re-writing app_db during the wait period. We need to revisit this logic

how about doing this in the if case and keep the original logic.

for (auto &entry : field_values) { writeConfigToAppDb(alias, fvField(entry), fvValue(entry));

First let me confirm your solution:

if (!isPortStateOk(alias)) { SWSS_LOG_INFO("Port %s is not ready, pending...", alias.c_str()); for (auto& i : kfvFieldsValues(t)) { writeConfigToAppDb(alias, fvField(i), fvValue(i)); } it++; continue; } // rest of the code keeps as original one

This solution makes code much clear and simple. But It will repeatedly put the configuration to APP DB until the port state turns "ok".

But its the same with your current implementation as well, as you are looping through all retry attributes and writing to app_db, correct?

no, for the retry process there are 3 cases:

port turns ok, mtu / admin_status will be configured into kernel, other fields will not be put into APP DB again

port is still not ok and mtu/admin_status were changed by user, it will only put the new mtu / admin_status to APP DB

port is still not ok and mtu / admin_status were not changed, nothing will be put into APP DB

How about this? You can use similar to 'configured' below. Apply it to APP_DB only first time in this else part. I would like to simplify the code. I dont think we need to worry the 2nd case (port is still not ok and mtu/admin_status changed). The new values get anyways applied once the port is ready.

prsunny

changes lgtm, one comment to address

prsunny · 2022-06-21T01:12:52Z

cfgmgr/portmgr.cpp

+            for (auto &entry : field_values)
+            {
+                writeConfigToAppDb(alias, fvField(entry), fvValue(entry));
+                SWSS_LOG_NOTICE("Configure %s %s to %s", alias.c_str(), fvField(entry).c_str(), fvValue(entry).c_str());


i'm confused by this. It seems like we are anyways writing all the fields to APP DB, so why individual writes above like mtu, tpid?

I will fix this

Left a comment above to move this section to top. Also please have this as INFO logs as we don't want to keep logging during retries.

prsunny · 2022-06-21T01:15:58Z

cfgmgr/portmgr.cpp

+             * We should not call any ip command if portOk=false. However, it is
+             * valid to put port configuration to APP DB which will trigger port creation in kernel.
+             */
+            bool portOk = isPortStateOk(alias);


My concern is, developer has to make sure kernel writes cannot be attempted if portOk is false which is different from the previous logic. Secondly, it keeps re-writing app_db during the wait period. We need to revisit this logic

azure-pipelines · 2022-06-24T03:14:49Z

Azure Pipelines successfully started running 1 pipeline(s).

Junchao-Mellanox · 2022-06-28T09:30:25Z

/azpw run Azure.sonic-swss

mssonicbld · 2022-06-28T09:30:27Z

/AzurePipelines run Azure.sonic-swss

azure-pipelines · 2022-06-28T09:30:36Z

Azure Pipelines successfully started running 1 pipeline(s).

Junchao-Mellanox · 2022-06-30T01:43:41Z

/azpw run Azure.sonic-swss

mssonicbld · 2022-06-30T01:43:43Z

/AzurePipelines run Azure.sonic-swss

azure-pipelines · 2022-06-30T01:43:53Z

Azure Pipelines successfully started running 1 pipeline(s).

Junchao-Mellanox · 2022-06-30T10:04:04Z

/azpw run Azure.sonic-swss

mssonicbld · 2022-06-30T10:04:08Z

/AzurePipelines run Azure.sonic-swss

azure-pipelines · 2022-06-30T10:04:18Z

Azure Pipelines successfully started running 1 pipeline(s).

Junchao-Mellanox · 2022-07-01T14:36:52Z

/azpw run Azure.sonic-swss

mssonicbld · 2022-07-01T14:36:54Z

/AzurePipelines run Azure.sonic-swss

azure-pipelines · 2022-07-01T14:37:04Z

Azure Pipelines successfully started running 1 pipeline(s).

cfgmgr/portmgr.cpp

prsunny

lgtm

prsunny · 2022-07-01T23:55:48Z

@Junchao-Mellanox , the VS test failure is tracke by #2365. Please retry once this is fixed.

prsunny · 2022-07-07T03:27:46Z

/azp run

azure-pipelines · 2022-07-07T03:27:55Z

Azure Pipelines successfully started running 1 pipeline(s).

Update sonic-swss submodule pointer to include the following: * Port configuration incremental update support ([sonic-net#2305](sonic-net/sonic-swss#2305)) * [VS Test] Skip failing subport tests ([sonic-net#2370](sonic-net/sonic-swss#2370)) * [teammgr]: Waiting MACsec ready before doLagMemberTask ([sonic-net#2286](sonic-net/sonic-swss#2286)) * [vnetorch] [vxlanorch] fix a set of memory usage issues ([sonic-net#2352](sonic-net/sonic-swss#2352)) * [tests] [asan] add graceful stop flag ([sonic-net#2347](sonic-net/sonic-swss#2347)) * [asan] suppress the static variable leaks ([sonic-net#2354](sonic-net/sonic-swss#2354)) * Add support for IP interface loopback action ([sonic-net#2307](sonic-net/sonic-swss#2307)) * [orchagent]: srv6orch support for uSID ([sonic-net#2335](sonic-net/sonic-swss#2335)) Signed-off-by: dprital <drorp@nvidia.com>

Update sonic-swss submodule pointer to include the following: * Port configuration incremental update support ([#2305](sonic-net/sonic-swss#2305)) * [VS Test] Skip failing subport tests ([#2370](sonic-net/sonic-swss#2370)) * [teammgr]: Waiting MACsec ready before doLagMemberTask ([#2286](sonic-net/sonic-swss#2286)) * [vnetorch] [vxlanorch] fix a set of memory usage issues ([#2352](sonic-net/sonic-swss#2352)) * [tests] [asan] add graceful stop flag ([#2347](sonic-net/sonic-swss#2347)) * [asan] suppress the static variable leaks ([#2354](sonic-net/sonic-swss#2354)) * Add support for IP interface loopback action ([#2307](sonic-net/sonic-swss#2307)) * [orchagent]: srv6orch support for uSID ([#2335](sonic-net/sonic-swss#2335)) Signed-off-by: dprital <drorp@nvidia.com>

*portsyncd no longer handle port configuration change and put it to APP DB *Implement incremental configuration change in portmgr *Adjust portsorch to meet incremental configuration change requirement

Junchao-Mellanox and others added 12 commits March 29, 2022 08:45

Implement incremental port configuration update

0273978

Improve unit test coverage

623b64a

Merge remote-tracking branch 'origin/master' into incre-port-update

400ee4b

Merge remote-tracking branch 'origin/master' into incre-port-update

24697b2

Fix vs test failure

c104b77

Merge remote-tracking branch 'origin/master' into incre-port-update

1cca0aa

Do not check configuration value since there could be case that someo…

97fae06

…ne may directly change APP DB

Merge branch 'master' into incre-port-update

9dce0ca

Merge commit '910bfd4' into incre-port-update

9c5990d

Conflicts: tests/mock_tests/mock_table.cpp

Port incremental configuration support

0b9b2e1

Merge remote-tracking branch 'origin/master' into incre-port-update-1

0b18221

Conflicts: tests/mock_tests/Makefile.am

Remove unused code

bbaadc5

Junchao-Mellanox requested a review from prsunny as a code owner June 2, 2022 07:07

prsunny mentioned this pull request Jun 14, 2022

[UT] [Portsyncd] Added Unit Tests for portsyncd #2297

Merged

prsunny reviewed Jun 17, 2022

View reviewed changes

prsunny reviewed Jun 21, 2022

View reviewed changes

Junchao-Mellanox added 2 commits June 21, 2022 09:51

Fix review comment

55a18d9

Fix review comment

ebeb091

Fix review comment

5b99253

Merge remote-tracking branch 'origin/master' into incre-port-update-1

25cb579

prsunny reviewed Jul 1, 2022

View reviewed changes

cfgmgr/portmgr.cpp Show resolved Hide resolved

prsunny approved these changes Jul 1, 2022

View reviewed changes

Junchao-Mellanox mentioned this pull request Jul 6, 2022

Port configuration incremental update support sonic-net/SONiC#985

Merged

prsunny merged commit 7126857 into sonic-net:master Jul 7, 2022

dprital mentioned this pull request Jul 7, 2022

[submodule] Advance sonic-swss pointer sonic-net/sonic-buildimage#11380

Merged

7 tasks

liat-grozovik added the Request for 202205 Branch label Aug 11, 2022

yxieca added the Included in 202205 Branch label Aug 11, 2022

Port configuration incremental update support #2305

Port configuration incremental update support #2305

Conversation

Junchao-Mellanox commented Jun 2, 2022

Junchao-Mellanox commented Jun 2, 2022

mssonicbld commented Jun 2, 2022

azure-pipelines bot commented Jun 2, 2022

liushilongbuaa commented Jun 6, 2022

azure-pipelines bot commented Jun 6, 2022

Junchao-Mellanox commented Jun 7, 2022

mssonicbld commented Jun 7, 2022

azure-pipelines bot commented Jun 7, 2022

Junchao-Mellanox commented Jun 13, 2022

mssonicbld commented Jun 13, 2022

azure-pipelines bot commented Jun 13, 2022

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

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

prsunny left a 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

azure-pipelines bot commented Jun 24, 2022

Junchao-Mellanox commented Jun 28, 2022

mssonicbld commented Jun 28, 2022

azure-pipelines bot commented Jun 28, 2022

Junchao-Mellanox commented Jun 30, 2022

mssonicbld commented Jun 30, 2022

azure-pipelines bot commented Jun 30, 2022

Junchao-Mellanox commented Jun 30, 2022

mssonicbld commented Jun 30, 2022

azure-pipelines bot commented Jun 30, 2022

Junchao-Mellanox commented Jul 1, 2022

mssonicbld commented Jul 1, 2022

azure-pipelines bot commented Jul 1, 2022

prsunny left a comment

Choose a reason for hiding this comment

prsunny commented Jul 1, 2022

prsunny commented Jul 7, 2022

azure-pipelines bot commented Jul 7, 2022