Bug: Route download performance degradation with FIB Suppression feature (even if it is disabled)

[mike-dubrovsky]

Is it platform specific

generic

Importance or Severity

Critical

Description of the bug

Note: Could you please assign the bug to me ; i have fix ready for review.

SONiC route download speed degraded (from around 4K/sec to 2K/sec) after BGP prefix suppression (bgp suppress-fib-pending) was introduced. The root cause is that RouteOrch::publishRouteState() unconditionally writes to APPL_STATE_DB and sends a Redis pub/sub notification on every route add/remove — even when FIB suppression is disabled and no consumer needs the data.

Impact
orchagent is the bottleneck for route programming; the overhead directly degrades route download speed
The majority of deployments do not enable suppression, so most users pay the cost for no benefit

Steps to Reproduce

compare route download speed in 202405 and any prior release.

Actual Behavior and Expected Behavior

publishRouteState() should be no-op when FIB suppression is disabled

Relevant log output

Output of show version, show techsupport

Attach files (if any)

Note: Could you please assign the bug to me ; i am preparing PRs for the review.
sonic-net/sonic-swss#4333
#26151
sonic-net/sonic-utilities#4361
sonic-net/sonic-mgmt#22916

[tjchadaga]

@deepak-singhal0408

[deepak-singhal0408]

@mike-dubrovsky , thanks for creating this issue/follow up..
this should be applicable to master also.. right? why the issue you created specifically mentions 202405 and any prior release?

Assigning to you.

[mike-dubrovsky]

@deepak-singhal0408 I was trying to say that 202405 is the release where FIB suppression was introduced into SONiC, and performance dropped. The issue is also applicable to all post-202405 releases.

P.S. The code is ready; I am looking at a sonic-mgmt failure that I see even in the base image without my fix (ETA for the script: April 14).
P.P.S. As MSFT will have FIB suppression enabled on T1 and above, this fix has limited value, but it is beneficial for other deployments where FIB suppression is not needed.

[deepak-singhal0408]

@deepak-singhal0408 I was trying to say that 202405 is the release where FIB suppression was introduced into SONiC, and performance dropped. The issue is also applicable to all post-202405 releases.

P.S. The code is ready; I am looking at a sonic-mgmt failure that I see even in the base image without my fix (ETA for the script: April 14). P.P.S. As MSFT will have FIB suppression enabled on T1 and above, this fix has limited value, but it is beneficial for other deployments where FIB suppression is not needed.

thanks @mike-dubrovsky for clarifying. I have updated the title of the issue. We take the fix in Master, and release owners can decide if they need this fix backported.. Good to fix this for deployments where fib-suppression is not enabled..