[fabricportsorch] Collect counters for fabric links #1944

mlorrillere · 2021-10-06T16:28:31Z

This change is enabling collections of fabric link counters as port and queue stats.

Currently only fabric asics can be collected, so collection is disabled for switch asics.
The reasons are that BCM SAI supports up to 256 logical ports, which is not enough
for ports + fabric lanes, and fabric queue counters are not supported by SAI at
this moment.

The following sonic-swss-common PR is required: PR #551
sonic-utilities PR #1860 adds the
corresponding show fabric counters CLI command.

rlhui · 2021-10-20T17:30:56Z

@skeesara-nokia , @vganesan-nokia - please help review, thanks.

orchagent/fabricportsorch.cpp

vganesan-nokia · 2021-10-27T03:27:53Z

orchagent/fabricportsorch.cpp

+        }
+        int num_queues = attr.value.u32;
+
+        if (num_queues > 0)


Will there be any chance we'll get num_queues <= 0? If yes, please add logic for error logging. If not, we do not need this check.

Definitely >= 0 as the return value of the SAI call is u32. I think we should expect a value >= 1. I'll check if that is correct and update the code accordingly.

vganesan-nokia · 2021-10-27T03:35:24Z

orchagent/main.cpp

@@ -636,6 +636,9 @@ int main(int argc, char **argv)
        if (gMySwitchType == "voq")
        {
            orchDaemon->setFabricEnabled(true);
+            // SAI doesn't fully support counters for non fabric asics
+            orchDaemon->setFabricPortStatEnabled(false);
+            orchDaemon->setFabricQueueStatEnabled(false);
        }


These are initialized as "true" in declaration (line#s 77 and 78 below). For regular switch (i.e, gMySwitchType is not "fabric" and not "voq") are we going to have these m_fabricPortStatEnabled, m_fabricQueueStatEnabled unchanged?

FabricPortsOrch is only created when gMySwitchType is "voq" or "fabric". These settings are set here to be used when FabricPortsOrch is instanciated.

Can you check the build failure?

We will need sonic-net/sonic-swss-common#551 to merge for build to pass.

vganesan-nokia · 2021-10-27T03:44:49Z

orchagent/fabricportsorch.cpp

@@ -42,7 +45,8 @@ static const vector<sai_queue_stat_t> queue_stat_ids =
    SAI_QUEUE_STAT_CURR_OCCUPANCY_LEVEL,
 };

-FabricPortsOrch::FabricPortsOrch(DBConnector *appl_db, vector<table_name_with_pri_t> &tableNames) :
+FabricPortsOrch::FabricPortsOrch(DBConnector *appl_db, vector<table_name_with_pri_t> &tableNames,
+                                 bool fabricPortStatEnabled, bool fabricQueueStatEnabled) :
        Orch(appl_db, tableNames),
        port_stat_manager(FABRIC_PORT_STAT_COUNTER_FLEX_COUNTER_GROUP, StatsMode::READ,


Do we have unit tests for fabric port flex counter groups similar to test_flex_counter.py?

We don't. Let me see if I can add FABRIC_PORT_STAT_COUNTER for test_flex_counter.py

The VirtualChassisTopology doesn't have any fabric asics defined, so it's not a simple change to test anything for switch_type=fabric in sonic-swss.

I'll work separately from this PR into updating virtual chassis topology and adding a test_fabric_counters.py test.

arlakshm · 2021-11-11T15:20:37Z

Can you check the build failure?

arlakshm · 2021-11-16T20:37:52Z

orchagent/fabricportsorch.cpp

+        }
+        port_stat_manager.setCounterIdList(port, CounterType::PORT, counter_stats);
+    }
+    m_portNamePortCounterTable->set("", portNamePortCounterMap);
 }

 void FabricPortsOrch::generateQueueStats()


Is function generateQueueStats used anywhere ?

generateQueueStats is called by FlexCounterOrch. This code was already in place.

arlakshm · 2021-11-16T20:42:01Z

/Azp run

azure-pipelines · 2021-11-16T20:42:13Z

Azure Pipelines successfully started running 1 pipeline(s).

mlorrillere · 2021-11-16T21:35:52Z

Can you check the build failure?

We will need sonic-net/sonic-swss-common#551 to merge to get a passing build.

prsunny · 2022-05-04T22:33:12Z

Please add VS tests

linux-foundation-easycla · 2022-08-02T17:42:12Z

The committers listed above are authorized under a signed CLA.

✅ login: mlorrillere / name: Maxime Lorrillere (dde035b)

mlorrillere · 2022-08-02T21:32:03Z

/easycla

abdosi · 2022-08-17T17:44:45Z

@mlorrillere please check on code coverage failures.

rlhui · 2022-09-20T06:29:50Z

@mlorrillere - any update?

mlorrillere · 2022-09-20T16:51:59Z

@mlorrillere - any update?

Hi @rlhui
I was waiting for sonic-buildimage PR #11997 to merge before updating this review.

It looks like there was a duplicate sonic-buildimage PR #12098 merged a few days ago. I'll update this PR shortly with test coverage.

mlorrillere · 2022-09-20T17:02:18Z

Tagging @jfeng-arista

mlorrillere · 2022-09-22T16:34:03Z

@rlhui - all checks have passed.

Counters are port stats and queue stats. Currently only fabric asics could be collected. J2 fabric counter collection doesn't work yet. J2 fabric port counters fail to be collected due to logical port id for fabric links is set up to 512 while SAI supports at most 256 ports. J2 fabric queue counters are not supported by SAI at this moment (BCM confirmed). Signed-off-by: Maxime Lorrillere <mlorrillere@arista.com>

kenneth-arista · 2022-10-05T21:15:40Z

@arlakshm @rlhui Can we merge this?

Counters are port stats and queue stats. Currently only fabric asics could be collected. J2 fabric counter collection doesn't work yet. J2 fabric port counters fail to be collected due to logical port id for fabric links is set up to 512 while SAI supports at most 256 ports. J2 fabric queue counters are not supported by SAI at this moment (BCM confirmed). Signed-off-by: Maxime Lorrillere <mlorrillere@arista.com> Signed-off-by: Maxime Lorrillere <mlorrillere@arista.com>

Advance sonic-swss submodule to pick up new commits: dbdf31c [counters] Improve performance by polling only configured ports buffer queue/pg counters sonic-net/sonic-swss#2473 ab4f804 [portsorch] remove port OID from saiOidToAlias map on port deletion sonic-net/sonic-swss#2483 ab29920 [QoS] Support dynamic headroom calculation for Barefoot platforms sonic-net/sonic-swss#2412 15beee4 Add support for voq counters in portsorch. sonic-net/sonic-swss#2467 c8d4905 [vlanmgr] Disable arp_evict_nocarrier for vlan host intf sonic-net/sonic-swss#2469 31c9321 [chassis][voq]Collect counters for fabric links sonic-net/sonic-swss#1944 Signed-off-by: Kebo Liu <kebol@nvidia.com>

mlorrillere requested a review from prsunny as a code owner October 6, 2021 16:28

rlhui requested review from abdosi and arlakshm October 20, 2021 17:31

vganesan-nokia reviewed Oct 27, 2021

View reviewed changes

mlorrillere force-pushed the pr-fabric-counters branch 2 times, most recently from f6a6770 to 3f08b7e Compare November 8, 2021 22:38

vganesan-nokia previously approved these changes Nov 10, 2021

View reviewed changes

arlakshm reviewed Nov 16, 2021

View reviewed changes

mlorrillere dismissed vganesan-nokia’s stale review via a486955 February 16, 2022 22:55

mlorrillere force-pushed the pr-fabric-counters branch from 3f08b7e to a486955 Compare February 16, 2022 22:55

rlhui added the chassis label Apr 12, 2022

rlhui assigned arlakshm Apr 12, 2022

mlorrillere force-pushed the pr-fabric-counters branch from a486955 to ea2a5d4 Compare April 20, 2022 16:50

mlorrillere requested a review from judyjoseph as a code owner April 20, 2022 16:50

mlorrillere force-pushed the pr-fabric-counters branch from ea2a5d4 to dde035b Compare August 2, 2022 17:42

abdosi added the Request for 202205 Branch label Aug 17, 2022

abdosi mentioned this pull request Aug 17, 2022

[voq][orchagent][fabric]"Failed to get port object for portid" error during line card boot up sonic-net/sonic-buildimage#9033

Open

mlorrillere closed this Sep 20, 2022

mlorrillere reopened this Sep 20, 2022

mlorrillere force-pushed the pr-fabric-counters branch from dde035b to d6e1291 Compare September 20, 2022 16:59

mlorrillere force-pushed the pr-fabric-counters branch from d6e1291 to 081e093 Compare September 21, 2022 16:40

mlorrillere force-pushed the pr-fabric-counters branch from 081e093 to c818a17 Compare October 3, 2022 19:04

arlakshm approved these changes Oct 5, 2022

View reviewed changes

arlakshm merged commit 31c9321 into sonic-net:master Oct 5, 2022

yxieca added the Included in 202205 Branch label Oct 25, 2022

keboliu mentioned this pull request Oct 26, 2022

[submodule] Advance sonic-swss pointer sonic-net/sonic-buildimage#12498

Merged

7 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[fabricportsorch] Collect counters for fabric links #1944

[fabricportsorch] Collect counters for fabric links #1944

mlorrillere commented Oct 6, 2021 •

edited

Loading

rlhui commented Oct 20, 2021

vganesan-nokia Oct 27, 2021

mlorrillere Nov 3, 2021

vganesan-nokia Oct 27, 2021

mlorrillere Nov 3, 2021

mlorrillere Nov 16, 2021

vganesan-nokia Oct 27, 2021

mlorrillere Nov 8, 2021

mlorrillere Nov 10, 2021

arlakshm commented Nov 11, 2021

arlakshm Nov 16, 2021

mlorrillere Nov 16, 2021

arlakshm commented Nov 16, 2021

azure-pipelines bot commented Nov 16, 2021

mlorrillere commented Nov 16, 2021

prsunny commented May 4, 2022

linux-foundation-easycla bot commented Aug 2, 2022 •

edited

Loading

mlorrillere commented Aug 2, 2022

abdosi commented Aug 17, 2022

rlhui commented Sep 20, 2022

mlorrillere commented Sep 20, 2022

mlorrillere commented Sep 20, 2022

mlorrillere commented Sep 22, 2022

kenneth-arista commented Oct 5, 2022

[fabricportsorch] Collect counters for fabric links #1944

[fabricportsorch] Collect counters for fabric links #1944

Conversation

mlorrillere commented Oct 6, 2021 • edited Loading

rlhui commented Oct 20, 2021

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

arlakshm commented Nov 11, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

arlakshm commented Nov 16, 2021

azure-pipelines bot commented Nov 16, 2021

mlorrillere commented Nov 16, 2021

prsunny commented May 4, 2022

linux-foundation-easycla bot commented Aug 2, 2022 • edited Loading

mlorrillere commented Aug 2, 2022

abdosi commented Aug 17, 2022

rlhui commented Sep 20, 2022

mlorrillere commented Sep 20, 2022

mlorrillere commented Sep 20, 2022

mlorrillere commented Sep 22, 2022

kenneth-arista commented Oct 5, 2022

mlorrillere commented Oct 6, 2021 •

edited

Loading

linux-foundation-easycla bot commented Aug 2, 2022 •

edited

Loading