Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[voq/systemlag] Voq system lag functionality #1605

Merged
merged 5 commits into from Mar 31, 2021

Conversation

vganesan-nokia
Copy link
Contributor

What I did

System LAG (PortChannel for chassis system) support

Why I did it

In VOQ based chassis systems, to forward across different ASICs through Portchannels configured in different asics, a new attribute called "system port aggregator id" (spa) is defined for the LAG objects. The system lag implementation in this PR allocates unique ID for Portchannel ans the id is sent to SAI to create the LAG with spa. This chassis wide unique spa is sync-ed with chassis app db so that non-owner asics will create corresponding remote LAGs with the same spa.

For more detailed information refer: Distributed VOQ LAG HLD sonic-net/SONiC#697

How I verified it

  • test_virtual_chassis.py with system lag test case passes

Details if related

This PR depends on the following PRs:

@anshuv-mfst
Copy link

@ysmanman (Arista) - could you please review this PR, thanks.

@@ -1362,9 +1363,13 @@ bool IntfsOrch::isRemoteSystemPortIntf(string alias)
Port port;
if(gPortsOrch->getPort(alias, port))
{

if (port.m_type == Port::LAG)
return(port.m_system_lag_info.switch_id != gVoqMySwitchId);
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can we put the statement in {} to be consistent with the rest of the change?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

o.k. I'll fix this.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Fixed

Comment on lines 1383 to 1398
if (port.m_type == Port::LAG)
{
if (port.m_system_lag_info.switch_id != gVoqMySwitchId)
{
return;
}
alias = port.m_system_lag_info.alias;
}
else
{
return;
if(port.m_system_port_info.type == SAI_SYSTEM_PORT_TYPE_REMOTE)
{
return;
}
alias = port.m_system_port_info.alias;
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This piece of code is duplicated in another function like voqSyncDelIntf too . Can we wrap this into a helper method? This can avoid duplicating code and also make it easy to extend it in the future.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is done intentionally. This repeating will be there in other places also. The intention of this logic is to get whether the port or the lag is remote or not. Since we need system alias which in the Port structure,, we get the Port structure first. Since the port structure is already available and it also has the type info, I check the type in the port structure itself instead of calling another API.

Comment on lines +2985 to +3021
//Already created, syncd local lag from CHASSIS_APP_DB. Skip
it = consumer.m_toSync.erase(it);
continue;
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is it always safe to simply skip here? Is it possible this is to retry a lag that was failing to add?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes. We should always skip this. This lag is the Portchannel locally created already and synced to CHASSIS_APP_DB. Since this api, which is used for both local lags and remote lags, this is invoked for entries added to CHASSIS_APP_DB also. Here we identify the local entry by the switch id. If it is local entry, since this local is already programmed, we should always skip this and erase it. So there will not be re-try at all.

@@ -3909,7 +4051,7 @@ bool PortsOrch::addLagMember(Port &lag, Port &port, bool enableForwarding)
attr.value.oid = port.m_port_id;
attrs.push_back(attr);

if (!enableForwarding)
if (!enableForwarding && port.m_type != Port::SYSTEM)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Does this mean we cannot disable forwarding for sys lag?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Setting (enable or disable) the ingress and egress is not applicable for remote LAG members. SAI returns errors when we set these attributes for system port lag members These are done at the owner ASIC (as local physical port members)

ysmanman
ysmanman previously approved these changes Feb 23, 2021
@judyjoseph
Copy link
Contributor

/AzurePipelines run

@azure-pipelines
Copy link

Azure Pipelines successfully started running 1 pipeline(s).

@judyjoseph
Copy link
Contributor

@vganesan-nokia There are some build errors can you take a look. Let me know if you are not able to check the

link.

@vganesan-nokia
Copy link
Contributor Author

@vganesan-nokia There are some build errors can you take a look. Let me know if you are not able to check the

This is expected. We need the lag id allocator PR #1603 to be merged for this to build successfully

@judyjoseph
Copy link
Contributor

/AzurePipelines run

@azure-pipelines
Copy link

Azure Pipelines successfully started running 1 pipeline(s).

system_lag_alias = gMyHostName + "|" + gMyAsicName + "|" + lag_alias;

// Allocate unique lag id
spa_id = m_lagIdAllocator->lagIdAdd(system_lag_alias, 0);
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Currently Sonic in the teammgrd we don't set a unique oper key in the LACP BPDU, but we are in the process of adding this support (#1660)

We could use this unique id allocation schema there as well. The change there would be to allocate the lagId in teammgrd when user creates a new LAG .. and here in orchagent it should give back the same id as it should be already existing ( created before in teammgrd and updated in SYSTEM LAG TABLE.)

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The need for this unique lag id allocation is to have unique lag id for all port channels across all the asics. So we let the id allocation be done in centralized location. As you mentioned, it will be interesting use case to use uniquely allocated id as key for LACP. This will work well for multi-asic and distributed chassis. For single asic, we need to set the id boundaries and load the script in local redis server. Yes, the lag id allocator gives the same id for a given portchannel, if it is already allocated,

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes in future we would need to extend this to teamd, so that we use this logic with the LUA script itself to generate a unique LACP key filled in the LACP BPDU.

@vganesan-nokia
Copy link
Contributor Author

Capturing review approval from @ysmanman before rebasing

image

vedganes added 5 commits March 11, 2021 18:22
Signed-off-by: vedganes <vedavinayagam.ganesan@nokia.com>

Changes for voq system lag implementation
(1) Portsorch changes for allocating unique lag id from chassis ap pdb
and sending the id in system port aggregator id attribute while creating
lag local LAG
(2) Portsorch changes to synd local LAG and local LAG members to
chassis app db. The sync-ing includes the allocated unique system lag id
(3) Portsorch changes to process remote system lag from chassis app db
and create lag entry in local asic db with received system lag id
(4) Interface orch changes to identify local or remote interfaces
(for both port and lag)
(5) Orchdaemon changes in orchagent intialization to get hostname and
asic_name attributes from DEVICE_METATDATA. These are used for unique
system lag name derivation
Signed-off-by: vedganes <vedavinayagam.ganesan@nokia.com>

Test case added for testing system lag
Signed-off-by: vedganes <vedavinayagam.ganesan@nokia.com>

Changes done to do the voqSyncDelLag() only for the local lag when lag
is removed.
Signed-off-by: vedganes <vedavinayagam.ganesan@nokia.com>

While adding system port as member for a system port channel, pvid is
set. Since pvid setting is not applicable for systemports in remote
asics, SAI returns faiure. This results in display of error.
setPortPvid() api is modified to skip setting the pvid for system ports
and an INFO is displayed.
Signed-off-by: vedganes <vedavinayagam.ganesan@nokia.com>

Code review comments fix changes: (1)Enclosed single line if block by {}
(2) Added system lag negative tests in vs tests and added code to log
error on system lag id release failure to help vs tests.
In addition to these review comments fixes, changes are done to remove
redundant error logs after addNeighbor() and removeNeighbor() in
system neighbor processing
@judyjoseph
Copy link
Contributor

/AzurePipelines run

@azure-pipelines
Copy link

Azure Pipelines successfully started running 1 pipeline(s).

@judyjoseph
Copy link
Contributor

/AzurePipelines run

@azure-pipelines
Copy link

Azure Pipelines successfully started running 1 pipeline(s).

@judyjoseph
Copy link
Contributor

/AzurePipelines run

@azure-pipelines
Copy link

Azure Pipelines successfully started running 1 pipeline(s).

@vganesan-nokia
Copy link
Contributor Author

Hi @judyjoseph, 4 tests of fine grained next group test suite failed. I tried to debug this. But I am unable to simulate in my machine. For me all tests (372 tests) passed. I think there may be some timing issue. Would you please re-start the Azurepipelines again?. Thanks

@judyjoseph
Copy link
Contributor

/AzurePipelines run

@azure-pipelines
Copy link

Azure Pipelines successfully started running 1 pipeline(s).

@judyjoseph
Copy link
Contributor

/AzurePipelines run

@azure-pipelines
Copy link

Azure Pipelines successfully started running 1 pipeline(s).

@judyjoseph
Copy link
Contributor

/AzurePipelines run

@azure-pipelines
Copy link

Azure Pipelines successfully started running 1 pipeline(s).

@judyjoseph judyjoseph merged commit 691bd30 into sonic-net:master Mar 31, 2021
raphaelt-nvidia pushed a commit to raphaelt-nvidia/sonic-swss that referenced this pull request Oct 5, 2021
* [voq/systemlag] VOQ System lag functionality

Signed-off-by: vedganes <vedavinayagam.ganesan@nokia.com>

Changes for voq system lag implementation
(1) Portsorch changes for allocating unique lag id from chassis ap pdb
and sending the id in system port aggregator id attribute while creating
lag local LAG
(2) Portsorch changes to synd local LAG and local LAG members to
chassis app db. The sync-ing includes the allocated unique system lag id
(3) Portsorch changes to process remote system lag from chassis app db
and create lag entry in local asic db with received system lag id
(4) Interface orch changes to identify local or remote interfaces
(for both port and lag)
(5) Orchdaemon changes in orchagent intialization to get hostname and
asic_name attributes from DEVICE_METATDATA. These are used for unique
system lag name derivation

* [vog/systemlag] VS test for system lag
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

4 participants