MoM of today's OCP SONiC call 5/17/2022

Skip to first unread message

MS Reddy

May 19, 2022, 12:11:28 AM5/19/22
to, sonicproject, Michael Schill, Xin Liu (CLOUD)
  • MoM of today's OCP SONiC call  5/17/2022

Topics discussed.

SystemD bootchart Integration - Nvidia (Stepan)
  • What is the major use case/motivation for Nvidia to bring this tool ? >> Nvidia identified performance degradation during sonic boot time, boot carts help to find out the root cause for it.
  • What are the use cases of this tool? Can this tool be used during production?
  • What are the minimum requirements (Mem, CPU, Disk space)to support SystemD boot charts? >> disk space required - 128 KB
  • How big are the generated SVG files?  >> ~10 MB 
  • Can this feature support only boot time? Can we analyse features runtime? >> Binary must be installed during compile time and CLI can be used to use it for run time.
  • How about the use case image upgrade w.r.t boot charts configurations ? 
  • Does this feature measure dockers/micro services ? >> yes
  • Does this design publish performance numbers ? >> Not yet
  • Is this feature part of built time flags? >> No, can be added.
  • Is there any evaluation report for tool selection? 
  • PR is out for review -
Syslog Source IP configuration HLD - Nvidia (Nazaril)
  • What is the use case of having source IP configured? 
  • why can't the existing design where it selects an interface for syslog to connect is not sufficient?
  • PR is out for review - 

  • 202205 release branch not yet cut - watch out for the communication
  • 202211 release plan is on;

On Tue, Apr 26, 2022 at 9:48 AM MS Reddy <> wrote:
  • MoM of today's OCP SONiC call  4/26/2022

Topics discussed.

  • How is this DASH proposal different from existing PTF frameworks Ex: Spy Test framework? >> this proposal for data plane testing based out of DASH (Disaggregated API for SONiC HOST)
  • Where do you get the DASH proposal ? Is there any workgroup discussing it?
  • What is the scope of the DASH SAI PTF proposal? Is this proposal for functionality testing or does it cover scale as well? 
  • How about underlay testing with DASH ? Is this proposal considered the underlay - >> Guohan suggested working with Prince on this item.
  • Pdf is out for feedback -
  • Do you have a plan to raise PR to review this proposal? 

On Tue, Apr 19, 2022 at 8:56 PM MS Reddy via <> wrote:
  • MoM of today's OCP SONiC call  4/19/2022

Topics discussed.

Fast-reboot Flow enhancements - NVIDIA 
  • How is this feature different from existing warm/fast reboot sequences  ?  What is the performance/downtime improvement here? 
  • How does this design measure the control plane's downtime ? Ex: is it considered ports, lags, routes, VRF's etc? 
  • Does this feature work with existing control plane assist with warm reboot? Reference:
  • Do you have benchmarks to control/data planes or downtime w.r.t configurations? 
  • How does the design handles image upgrade use case w.r.t schema updates? >> Existing db_migrator does take care of config_db & app_db for image upgrade schema changes 
  • What is the restoration logic to be handled in orchagent? >> The logic makes sure all the do_tasks are completed without any items in the queue.
  • Does the new fast fast-reboot design support bulk-api ASIC ? >> yes
  • Is the bulk-api support from ASIC mandatory or choice to use fast fast reboot feature ? >> not mandatory that ASIC must support bulk API to get fast fast reboot feature
Project announcements:

On Tue, Mar 29, 2022 at 8:51 AM MS Reddy via <> wrote:
MoM of today's OCP SONiC call  3/29/2022

Topics discussed.

DSCP/TC Remapping for Tunnel traffic HLD - MSFT 
  • Yang models - the table names not used anymore, the yang model definitions need to update the HLD and commit. 
  • Is the design DSCP/TC remapping is vendor specific, as some vendors don't need to remap as the inner header will be used for mapping?  Yes.
  • HLD should be updated with various vendor support for remapping DSCP/TC remapping. 
  • PR is out for review -

On Wed, Mar 23, 2022 at 7:59 PM MS Reddy <> wrote:

MoM of today's OCP SONiC call  3/22/2022

Topics discussed.

  • Is the SymCrypt library FIPS compliant or certified? >>> it is certified, MSFT & security team submitted the validation, certification number is not received yet.
  • Do you have performance numbers with the SymCrypt SSL engine? 
  • Does this feature support the build time option? yes 
  • PR is out for review -
  • What are the real use cases that will address this HLD? Please list it down in HLD.
  • How does the existing "cfg reload" design differ from it? How does it benefit the admin? 
  • SONIC Currently doesn't allow two "cfg reload" commands parallely? How does this design benefit the admins? What is the problem? Does this design address it ?
  • PR is out for review -

On Tue, Mar 15, 2022 at 8:36 AM MS Reddy <> wrote:
MoM of today's OCP SONiC call  3/15/2022

Topics discussed.
  • How much performance gain/ saving time using batched requests API - 8%
  • How does the batch write API handles in case of a process crashed ?   In case a consumer crashes, Redis queues up the requests and the process should be able to consume it.
  • How does the batch process API handles the scenarios priority inversion Ex: Orchagent is single threaded, can be blocked due to a high priority task, (Link up/down/ PFC storms need to react quickly, popping routes) - Wu will look into it.
  • Is there a limitation to the size of the batch? 
  • How does it handle the writes failures in a batch? Explain redis transaction vs how producers and consumers behave in case of failure scenarios?
  • What use case does this batch API can be used for ? Please list out the use cases in HLD.
  • How did you test this API? Do you have any performance numbers? 
  • PR is out for review -
  • General: Community must focus on 202205 SONiC May release for feature delivery.


On Tue, Mar 8, 2022 at 8:29 AM MS Reddy <> wrote:

MoM of today's OCP SONiC call  3/08/2022

Topics discussed.
  • Discussed on 202205 release feature list and did house cleaning activities.
  • We have Identified the owners to migrate docker images to BullsEye.
  • Watch out for updated xls from Ying/Zhang and contribute for missing docker packages and update the ownership and reviewers list.


On Tue, Mar 1, 2022 at 9:38 PM MS Reddy <> wrote:

MoM of today's OCP SONiC call  3/01/2022

Topics discussed.
On Tue, Feb 22, 2022 at 8:53 AM MS Reddy via <> wrote:

MoM of today's OCP SONiC call  2/22/2022

Topics discussed.
  • Is the CLI to get current memory usage of the container ? yes docker stats provides that
  • If memory condition persists in monit , do u generate tech dump again again? - No generate once by special instruction from Monit
  • Are these memory threshold users configurable ? yes and the 200 MB is default available memory 
  • what kind of report is this tech support? Is there a way to provide a summary report? 
  • could we support multiple thresholds ? start at 60%, jump at 80% collect one more tech support?
  • Is this memory leak or memory thresholds for a container being reported by. syslog today? yes by the Monit process.
  •  PR is out for review -

On Tue, Jan 11, 2022 at 8:49 AM MS Reddy via <> wrote:
MoM of today's OCP SONiC call  1/11/2022

Topics discussed.
  • Are these passwd rules/policies mandated for REST/HTTP users? sure, will be added to the design.
  • Is the passw hardening supported for remote users? No, only local users.
  • Is pam_cracklib FIPS compliant? Is it using the open ssl for encrypt/decrypt?
  • Is this code part  of sonic_mgmt repo ? yes 
  • Is the feature enabled default? No will be included by compile time, and choose by run time
  • Is the passwords can be rotated ? not part of this design, can be thought through the use cases
  • Is it possible to provide an informational log for the users/applications about the passwd expiry? will be included..
  • How does it handle switch image upgrades w.r.t passwd hardening? 
  • PR out for review-

On Thu, Dec 2, 2021 at 5:39 PM MS Reddy <> wrote:
MoM of today's OCP SONiC call  11/29/2021.

Topics discussed.
  • Discussed on 202111 community release fork date (11/30/2021 PST)
  • Release planning what is in and what will be moved to the next release
  • There is a Xls from Zhang for more details marked Yellow & Red.
  • Features marked with Yellow need be handled by today with priority
  • Features marked with Red will be moved to next release


On Thu, Nov 4, 2021 at 9:49 AM MS Reddy via <> wrote:
MoM of today's OCP SONiC call  11/02/2021.

Topics discussed.
  • SONiC SAI Challenger - SAI Testing by PLVision. 
  • Slides will be published soon by PLVision.


On Tue, Oct 19, 2021 at 11:26 AM MS Reddy via <> wrote:
MoM of today's OCP SONiC call  10/19/2021.

Topics discussed.
Had 202111 release features review, stay tuned for the update.


On Tue, Oct 12, 2021 at 8:30 AM MS Reddy <> wrote:
MoM of today's OCP SONiC call  10/12/2021.

Topics discussed.
NVGRE - by Vadym/Nvidia
  • Does the design provide capability checks for tunnel resources from vendors (no.of NVGRE supported)? Ans>> No.
  • How about vNet routing support on NVGRE tunnels? Not supported, this feature does encap/decap tunnelled packets.
  • PR is out for review -
Dynamic  policy based Hashing- by Nvidia
  • Not ready, will be postponed to the next release.
Note: Today was the last HLD discussion for the 2106 release. 


On Tue, Oct 5, 2021 at 6:10 PM MS Reddy via <> wrote:
  • MoM of today's OCP SONiC call  10/05/2021.

Topics discussed.
 CMIS Diagnostics - by Dante Su 
  • There is SFP Refactoring, how does this design different from that effort? >> Debate: This solution will coexist with sfp refactoring efforts, however later will merge into SFP refactoring.
  • Are there any impacts with current SFPUtil show commands with new additions ? No impact, there will be new application advertisement, pl refer the CLI section for review 
  • PR is out for review -  
System Ready Enhancements - by Senthil Kumar Guruswamy 
  • How is this different from the current Monit feature ? Ans>> The Monit summary provides platform status such as LED etc.. it shows the running status of the container not the application readiness
  • Does this design consider Application readiness vs liveness ? Readiness means all the dependent modules up and application ready to serve the traffic, how about the application hogs on memory cycle or runs out of threads not able to service requests? will it be possible to include liveness capability into the design?
Miscellaneous - by XIn 
  • OCP date - Nov 9th- 10th 
  • OCP Schedule will be published  on OCP website 
  • What is the mode of OCP workshop - virtual 
  • SONiC/ SAI Workshop - Tech Talk / Contribution / Proposal / Innovations - Let's plan for immediate after /next day OCP Nov 9th-10th


On Tue, Sep 28, 2021 at 9:01 AM MS Reddy <> wrote:
Topics discussed.
Host Interface counters -  MLNX/Chen
  • Can this design support packet drop counters due to DDOS attacks? a few options: 1. You can attach a policer to the drop counter. You can get flow counters from policer stats>> Chen will look into it.
  •  PR is out for review -   
Guidelines for reference proprietary code  -  John/Metaswitch

On Tue, Sep 21, 2021 at 8:58 AM MS Reddy <> wrote:
  • MoM of today's OCP SONiC call  9/21/2021.

Topics discussed.
  • Is the design supported to show the list of authorised commands ? >> No, the list of commands managed by tacacs server.
  • Is the audit support 
  • How does the design work when a remote TACACS+ failover? >>> as the local server don't know the list of commands? >> this is an issue, in SONiC, users can login and run using local permissions.
  • How do I block the commands using bash(/bin/sh)/python etc? 
  • PR is out for review -
MPLS TC_to_TC_map HLD -  Alexander (Metaswitch)


On Tue, Sep 14, 2021 at 6:36 PM MS Reddy <> wrote:
  • MoM of today's OCP SONiC call  9/14/2021.

Topics discussed.
ECMP Overlay BFD support -  Prince / MSFT
  1. Does this feature support BUM traffic ? No - Since Vnet routes are based on unicast routes, it supports Unicast only.
  2. Does the feature support control plan BFD / FRR BFD ? No, this BFD offload to ASIC
  3. What is the motivation/use case endpoint monitor ip is different from the actual endpoint running BFD ?  The use case is - There are devices support  data vs control plane on different ports for monitor purpose
  4. Can the design support query ASIC BFD capabilities before writing  the BFD session ? could be possible, will add into the HLD
  5. Can the control plan & hardware offload BFD coexists on the same device ? yes/No -  Need to brainstorm complexity - Will split the HLD into two, one dedicated to BFD to describe all the scenarios.
  6. What are the default BFD timers used for offload ? WIll be included into HLD
  7. Is there any global session bfd table for default values? No
  8. Community suggested to have BFD into a separate HLD ? Ex: coexistence may cause issues, need to discuss more in details
  9. How about end user CLI to control BFD session either hardware offload or FRR BFD ? Next phase
  10. Do you have a BFD state db schema mapped to the Rfc BFD schema ? Will be included in HLD
  11. How do handle or notify BFD sessions from HW? 
  12. Is it possible to remove BFD routes from the ECMP group ? yes, vNetOrch
  13. PR is out, please leave comments here -
General Comments
  • SUggest that use HLD PR as tracking PRs for related Code PRs. REference -  -
  • 202111 feature release Deadline - Oct 1st  
  • Some features will be delayed to next release - No list yet identified
  • Feature owners with HLD ready -  if you want to schedule the review - reach out 

    Yanzhao Zhang


On Thu, Sep 2, 2021 at 7:25 AM MS Reddy <> wrote:
  • MoM of today's OCP SONiC call  8/31/2021.

Topics discussed.
SHow Running Command Enhancement -  - EdgeCore/MaxChen
  1. What are the use cases of this feature ? Customers familiar with CISCO like CLI, goal is to make it more convenient or trivial to understand the current running command . 
  2. How does the design handle the maintainability of these improvements? 
  3. How does Yang show the running command ? 
  4. It looks like manual work, shouldn't it be duplicated to the sonic-mgmt-framework which is yang driven/auto generated? >> Not really!
  5. PR is out, please leave comments here -
Routed Subinterfaces Enhancement -  Preetham/BRCM
  1. What are the use cases of shorter sub interface naming ? 
  2. Why does it required to bring in the short names? >> Kernel naming limited to 15.
  3. Where do you store the mapping short to long name child to parent ?
  4.  Suggestion - for consistency reason keep vlan interface mandate.
  5. Is this config change only for routed subinterfaces ? How do they differentiate ? 
  6. There must be a short name convention in SONiC already, is this design considered that? 
  7. Why can't this short vs long name conversion hide in intrfaceMgrD? >> it looks like the changes are widespread 
  8. PR is out for review - please leave comments here -

On Tue, Aug 24, 2021 at 8:31 PM MS Reddy via <> wrote:
  • MoM of today's OCP SONiC call  8/24/2021.

Topics discussed.
SAG - Static Anycast Gateway - EdgeCore/MaxChen

  1. Is the SAG feature enabled by default? >> No, the feature is default disabled.
  2. Why do I need a knob for SAG? >> Please list down in HLD sections, is there any implications for taking care in data path handling of SAG vs macvlan interfaces.
  3. Unless it's absolutely necessary, it's not required to have a global knob. A global knob will introduce a lot of complexity and cases to handle. >> So is it necessary to have a knob for SAG?? can the gateway & ip address list can't be sufficient ? >> It seems to be true. 
  4. SagMgrD is not required when we plan to use SVI instead of maxvlan interfaces. 
  5. Is there my hardware resource limit on SAG interfaces ? can it be referred to in CRM?
  6. Please list down what are the complexities when SAG is enabled along with SVI/macvlan interfaces in data path routing?
  7. CLI SAG command can be part of interface command.
  8. PR is out -
Show running enhancement - EdgeCore
General comments
1. Feature owners speed up as the deadline approaches 
2. Test quality is highest important for Community Features 


On Tue, Aug 17, 2021 at 9:45 AM MS Reddy via <> wrote:
  • MoM of today's OCP SONiC call  8/17/2021.

Topics discussed.
PINS - P4 Integrated Network Stack  - Google/Intel/ONF
  1. Why don't the design leverage the existing Error DB framework for feedback loop? >> It seems the PINS team is closely working on the Error framework team to address the gaps.
  2. Is this design different from FlexSAI ? Using PINS, you can model the entire SAI pipeline, not the case with FlexSAI
  3. What are the advantages of exposing the entire SAI pipeline using PINS? >>> we can do Fuzzing, Automation testing the entire pipeline being exposed.
  4. What kind of intelligence does the design provide to the applications in terms of network/application/resource errors?  >> there is new HLD work in progress.
  5. What is the plan to support vendor SAI extensions? Can vendors SAI extension be added without recompiling libSAI? >> yes, HLD described it
  6. How does the design handle the missing redis pub/sub response path / notifications ? 
  7. How about the PINS migration plan in terms of software upgrades vs ASIC upgrades? >>>Please add a section in HLD.
  8. Can this design work on packet I/O performance improvement? >> So far, the numbers are promising, and will be looked into.
  9. Can admins run SONiC without P4RT? yes.
  10. PR is out for review, please provide comments offline -

On Sun, Aug 15, 2021 at 6:33 PM MS Reddy via <> wrote:
  • MoM of today's OCP SONiC call  8/10/2021.

Topics discussed.
SONiC_SFP_refactoring HLD - Arista/MSFT
  1. How does the design support backward compatibility of existing sfp modules ? 
  2. What are the guidelines for Vendors to implement the common sfp refactor packages? Please list out few examples for vendors to embrace it
  3. Can the design provide sfp data as Dict including all the sfp fields? 
  4. PR is out for review please leave comments here -
What next: 



On Sat, Aug 7, 2021 at 1:19 PM MS Reddy <> wrote:
  • MoM of today's OCP SONiC call  8/03/2021.

Topics discussed.
  1. How does any SONiC user consume this feature? What is the plan? How to enable this feature 
  2. Is the regularSONiC release not enabled by default? yes Tom please confirm
  3. It doesn't provide any CLI and asks users directly to play around with app_db tables. What is the guidance? Please list down the instructions in the HLD section
  4. Shouldn't the design be limited to the number of FC ? platform specific 
  5. Can it be the DSCP values more than FC values right? Yes, it is 
  6. Can the design expose the DSCP/EXP values to applications? yes
  7. Is there any plan to introduce click commands / CLI ? >>> Tom 
  8. A separate table for DSCP_to_FC, and refer to those table names in CLASS_BASED_NEXT_HOP_GROUP_TABLE ?

On Sun, Aug 1, 2021 at 5:02 PM MS Reddy <> wrote:
  • MoM of today's OCP SONiC call  7/27/2021.

Topics discussed.
  1. Tech support dump improvements - Nvidia (Vivek) - Please share the HLD here
  2.  Q & A - How does it handle if the device end up with continuous coredumps? Is there way to ship the core files external?
  3. 202111 release plan (HLD & Code PR's reviewers)- will be posted shortly by Yanzhao Zhang
  4. July 23 - call for Paper OCP / OCP website to submit abstract - Select Networking Track -
  5. OCP Updates for call for papers.

On Tue, Jul 13, 2021 at 8:58 AM MS Reddy <> wrote:
  • MoM of today's OCP SONiC call  7/13/2021.

Topics discussed.
  1. 202111 release plan (HLD & Code PR's reviewers)- will be posted shortly by Yanzhao Zhang
  2. July 23 - call for Paper OCP / OCP website to submit abstract - Select Networking Track -
  3. Paper Selection - Aug 

On Tue, Jun 22, 2021 at 8:59 AM MS Reddy via <> wrote:
  • MoM of today's OCP SONiC call  6/22/2021.

Topics discussed.
  1. 202106 release status & discussed code & PR status 
  2. Wiki will be posted shortly.
  3. 202106 release cut - 06/30

On Tue, Jun 15, 2021 at 8:55 AM MS Reddy <> wrote:

  • MoM of today's OCP SONiC call  6/15/2021.

Topics discussed.
CMIS-C-CMIS [Coherent - Common Management Interface]  - Chuan Qin/MSFT
  1. What is the goal of CMIS? >>>It is to configure and monitor optics/ transceivers.
  2. Why do need a special daemon, shouldn't be used xrvd or transceiver? .>> It is extending the xrvd
  3. How does SONiC consume these interfaces? Is it REst interface or CLI? will be discussed
  4. What are the plans to integrate with SONIC, need more detailed steps w.r.t CMIS interfaces as well as CMD firmware upgrades?
  5. It would be very helpful if there would be a list of APIs which need to be implemented by vendors and how these are used. We are looking to understand which daemons are using it, CLI, etc.

General updates

202106 release updates

202111 release planning 

Few timelines

  • 202111 feature contribution submission end by 6/25/2021
  • 202111 feature roadmap review in community on 7/6/2021
  • 202111 release roadmap finalization 7/15/2021


On Tue, Jun 1, 2021 at 8:42 AM MS Reddy <> wrote:
  • MoM of today's OCP SONiC call  6/01/2021.

Topics discussed.
SAI Failure Handling - Shi-su

On Tue, May 18, 2021 at 8:41 AM MS Reddy via <> wrote:
  • MoM of today's OCP SONiC call  5/18/2021.

Topics discussed.
Sonic Dump Utility - Vivek from NVDA

On Tue, May 11, 2021 at 5:41 PM MS Reddy <> wrote:
  • MoM of today's OCP SONiC call  5/11/2021.

Topics discussed.
SRv6 HLD - Houdi from Alibaba + Intel
  • How did you arrive at this requirement ex: SRv6 sidList per Policy = 4 ? Is this specific to Alibaba deployment? 

  • How deep can transit nodes be ? Why is this limited to 3 in HLD? - Again, deployment specific - more headers results in SRH header compression.

  • How does the design support tying up the policy with routes? right now it supports policy tie up with prefixes. can be enhanced.

  • The workflow diagram is a little confusing, needs an update and should discuss it once again - pl focus on w.r.t ownership of the route data and who does what modifications + in terms of consumers/producers.

  • Alibaba/Intel will share the PR for review...

On Fri, Apr 30, 2021 at 6:52 PM Madhu Pal <> wrote:
Hi Srinadh, 
Please find answers line:


On Fri, Apr 30, 2021 at 1:27 PM <> wrote:
I would like to understand more about "Don't overload stateDB for events & alarms? It should be advised to use separate redis DB for events & alarms?"  comment. Appreciate some responses from the community. I hope I am using the right channel.

Eventd is planning to use stateDB to house event history table, alarm table and stats tables. They get updated every time an event/alarm is raised.
These tables are of fixed size: event history table size is customizable with maximum size being 40k or time limit of 30 days - at which time, eventd deletes older records. Stats table is of fixed size with a handful of records. Alarm table only contains a record when an alarm is raised and record is removed when the alarm is cleared.

How does stateDB gets overloaded? 
Madhu>> As the operational data stored in state db today, by adding events, alarms and stats (frequent data) into it makes state db easily out of limits. In addition to that a software (eventD open source)/design flaw makes it worse.  Unlike APP_DB, State DB is more read friendly, more writes/updates due to events/alarms/stats become performane issues? Btw, do you get a chance to estiamte on  state DB with new data / what is plan to mesure performance of state db with ne data writes?
Is it because of db writes? OR 40k for history table is too much in a DB? 
Madhu>> Today each redis instansce setup with multile redis db's ex: APP_DB, STATE_DB etc.. Not 100% sure REDIS memory limits and how does it set for each redis db  or any redis profile?? MSFT team can help on this.
Using serparate redis DB means, I need to create a new redis instance and create a DB with that insance? OR create a DB off existing redis, redis2, redis3 instances? 
Madhu>>I'd suggest to use new redis db similar to state db ex: event db or altogether new redis instance (redis1, redis2 - ) ex: 

On Tuesday, April 27, 2021, 09:15:18 PM PDT, MS Reddy <> wrote:

  • MoM of today's OCP SONiC call  4/27/2021.

Topics discussed.
Event & Alarm Framework HLD- Srinath - Dell
Generalizing config.bcm support all brcm platforms - BRCM


On Tue, Apr 20, 2021 at 6:42 PM MS Reddy <> wrote:
  • MoM of today's OCP SONiC call  4/20/2021.

Topics discussed.
Event & Alarm Framework HLD- Srinath - Dell
  • Is this HLD supports SONiC CLi? No, It is supported by mgmt-framework.
  • Why can't current syslog be utilized or enhanced? 
  • What is the motivation choosing eventD?
  • How does this HLD integrate with Thermal design HLD which has similar eventing.
  • How does this design handle event re-ordering? What suggestions if the events are out of order? 
  • Don't overload stateDB for events & alarms? It should be advised to use separate redis DB for events & alarms?
  • How does the design handle the life cycle of an event/alarm? What are the eviction policies enforced on to the DB? 
  • Advised to use dynamic json event profile instead of using a static map? 
  • PR is out for review -
  • Review will be continued..

On Tue, Apr 13, 2021 at 9:15 AM MS Reddy via <> wrote:
  • MoM of today's OCP SONiC call  4/13/2021.

Topics discussed.
Policy based Hashing - Nvidia
  • How does the design calculate hash resources ? There is no SAI API to calculate this. Please add a comment in the HLD. 
  • How does CRM resources handled with PbH? 
  • When port is part of the LAG, how does PbH rules PhB table - User/Orch agent should pass LAG to ASIC. - Validation should be taken care of in application.
  • Add data flow sequence diagram explains the precedence or out of order handling of hash vs rule ?
  • What is the behaviour if PBH table/resources is full ? there is no API - currently raise syslog/error to user and fallback to no PBH - hash will not be created
  • Does SONiC track any thresholds ? Currently the ACL thresholds are being tracked but not the scarce resource like ALU / mirror sessions / no thresholds being tracked
  • Hashing will be calculated only on inner frames . based on type ipv4 or ipv6 or vxlan user defined 
  • what are the fields expected to be configured for NVGRE? There is a reference  example in HLD
  • The PbH data model should be yang complaint? yes there is a section below. 
  • Is there any way to track ASIC hash resources today? No SOniC Infra support yet.
  • PR is out for review -

On Tue, Apr 6, 2021 at 9:09 AM MS Reddy via <> wrote:
  • MoM of today's OCP SONiC call  4/06/2021.

Topics discussed.
SONiC BUM control support - Mohan S
  • How does the design calculate CBR, as there is no support  for user to configure? >> It is internally calculated by the application - BRCM
  • How do you handle unknown multicast & unknown unicast storm control as there are no separate SAI policers? - Mohan will look into it
  • Are there any statistics for drop counters due to storm control for each category - Not supported 
  • How does user stop storm control - By delete storm control config.
  • Do you query strom control capability on asic ? How does the application know ASIC has this capability ? mgmt framework would support this feature ?? Mohan will check it 
  • Is there a sonic_yang support for storm control ? yes 
  • Share code PRs and Sonic Yang to the community for review - Mohan
  • PR out for Review -
SONiC Mgmt framework - ''show techsupport dump" - Kerry
  • Can the design support flexibility of what content needs to be added to the tech support dump ? - yes
  • can the design support download the tech support tarBall to clients? - Not yet, will be supported
  • PR is out for review -

On Tue, Mar 30, 2021 at 9:18 AM MS Reddy via <> wrote:
  • MoM of today's OCP SONiC call  3/30/2021.

Topics discussed.
Recycle Port support for VOQ Chassis - Eshwaran
  • Are these special ports supported by SAI? 
  • How does the recycle port be discovered and enabled for forwarding? 
  • Do you have any ASIC capability check? No 
  • Does it require a reboot ? currently yes. 
  • Does HLD support CLI for recycle ports - Not yet.
  • Is there any strong reason behind the port naming ? Why do we choose the ReCycle port ? Recycle ports vs Recirculation ports?  Ex: Recirculation ports already being referenced in p4 
  • Where do these ports exist? If it is in port-config.ini or platform.json? Pl update the HLD and add these config into platform.json as port_config.init may not be appropriate going forward.
  • Are these recycle ports per core or per ASIC? 
  • DO you define a schema for these ports? 
  • Are these Recycle ports supported by configDB schema? If yes, should we develop a yang model for it? Yes, I will support it.
  • PR is out -
DHCPV6 Relay support in SONiC -Shlomi Bitton
  • How does the design handle dhcpv6 relay w.r.t dynamic port/route changes ?
  • Is there any limitation with isp-dhcp package that it requires static configuration? which is not the case with ipv4
  • How does the design handle port down/up events wrt dhcp relay packets? 
  • Scale recommendation - 32 dhcp relay servers
  • The Nvidia team will syncUp with Tamar on DHCPV6 Relay for any overlap. 
  • Nvidia will evaluate isp-dhcp vs dhcp-relay-6 package to find out a way to simplify the configuration.
  • PR is out for review-

On Tue, Mar 23, 2021 at 9:24 AM MS Reddy via <> wrote:
  • MoM of today's OCP SONiC call  3/16/2021.

Topics discussed.

SONiC Generic Update and Rollback- Mohamad

On Tue, Mar 16, 2021 at 8:43 AM MS Reddy via <> wrote:
  • MoM of today's OCP SONiC call  3/16/2021.

Topics discussed.

SONiC In_band mgmt_vrf - Venkat/Dell
On Tue, Mar 16, 2021 at 8:35 AM MS Reddy <> wrote:
  • MoM of today's OCP SONiC call  3/9/2021.

Topics discussed.

MPLS HLD- Ann Pokora/Juniper
  • Can you explain how Juniper crpD coexists with FRR routing stack in SONiC? >> using build time flags 
  • How does this feature can be unit tested without crpd ? Ex: Few thoughts - user can push directly to redisDB 
  • What is the use case are we targeting? 
  • How does the community can leverage this feature? 
  • For MPLS,  why does it only works withJuniper stack?  Is there any MPLS functionality missing in FRR? >> yes there are few fixes from Juniper need to push to FRR community
  • Can Juniper open source crpd ? >>> There is internal discussion at Juniper.
  • What is the reason fpm syncd in a bgp container instead of switch? >> Can be discussed!!
  • PR is out for review -

On Tue, Mar 2, 2021 at 9:07 AM MS Reddy via <> wrote:
  • MoM of today's OCP SONiC call  3/2/2021.

Topics discussed.

Weighted ECMP HLD- ZhenCAI/Juniper
  • How does warm boot work with code versions with EMCP comes up Weighted ECMP on warm boot?
  • How does applications handle if SAI doesn't support weighted ECMP? 
  • Can this design support static next hop groups? 
  • How does this design work in conjunction with Metaswitch nexthop group proposal?
  • Is the design query SAI capability before programming SAI about the weight ECMP ? 
  • Is there any config knob to enable weighted ECMP feature? 
  • What is the default weight nextHop ?
  • Is this feature tightly coupled with the routing stack? 
  • Is the design providing error handling to handle the weight mismatch ?
  • How does the design treat If all the nexthop entries have no weight ?- Ex: treat them as ECMP 
  • How do know SAI supports weighted ECMP or not? If SAI doesn't support how does ASIC handle it?
  • PR is out for review -
On Tue, Feb 23, 2021 at 9:10 AM MS Reddy <> wrote:
  • MoM of today's OCP SONiC call  2/23/2021.

Topics discussed.

SONiC Yang Models- Praveen/LNKD
  • Do you enforce any yang model prerequisites before HLD approvals? yes, yang models should be discussed in sub group
  • Is there any demo on how to run the yang tests? >> Not yet, can be provided check with Praveen
  • Are there any alternatives to validate the yang models except build time? 
  • Did we think about the backward compatibility of sonic ynag models? >> will be discussed in yang model subgroup
  • Do you have any guidelines for sonic yang model release migration ? >> will be discussed in the yang model subgroup
  • Where should approval for the sonic models be ? >> sub group 
  • What about sonic yang models for FRR/bgp routing stack ? >>will be discussed in yang model subgroup
  • PR is out for review -


On Tue, Feb 16, 2021 at 9:28 AM MS Reddy <> wrote:
  • MoM of today's OCP SONiC call  2/16/2021.

Topics discussed.

 CPU queue stats- Prabhu/BRCM
  • Don't SONiC support the CPU queue stats today? No, front panel ports have this support today, extending it for CPU ports.
  • Can it be extended to show port counters & show port pfc instead of new CLI? yes
  • SNMP/MIB OID support be considered? right now the design skips CPU ports, but we can consider it 
  • Telemetry use case supported - yes (counters written into flex db)
  • Do support debug counters ? yes
  • HLD listed only MC CPU queues, don't it support unicast packet counters? Yes supports, however BRCM all CPU queues are Multicast queues- based on ASIC support
  • How does this feature work with Multi ASIC platform feature? will be added into HLD
  • Can this feature be included in a system test? yes
  • PR is out for review :
  • Release notes 202012 branch - review in progress
  • 202106 release planning - wiki will be published soon 

On Tue, Feb 9, 2021 at 9:04 AM MS Reddy <> wrote:
  • MoM of today's OCP SONiC call  2/09/2021.

Topics discussed.

Fabric Port Support SONiC/BGP Setup for VoQ system - Arista

_._,_._,_ Links:

You receive all messages sent to this group.

View/Reply Online (#486) | Reply To Group | Reply To Sender | Mute This Topic | New Topic
Your Subscription | Contact Group Owner | Unsubscribe []


MS Reddy

May 24, 2022, 12:35:24 PM5/24/22
to, sonicproject, Michael Schill, Xin Liu (CLOUD)
  • MoM of today's OCP SONiC call  5/24/2022

Topics discussed.

Active-Active dual ToR HLD - MSFT (Jing)
  • What is the convergence/down time expected to set up tunnels on server link failure? 10-20 mSec >> tunnels are ip-in-ip and being pre-setup and convergence calculated 10-20 msec measures the time taken to detect the link and re-route traffic. This depends on various scenarios, and will be listed down in the test section.
  • How is it different from the Y-cable solution? Can you add a section to list down use cases for active-standby vs active-active 
  • Can you add a section to describe the gRPC high level design? Is it a gRPC  open implementation or does it need to understand any nuances?
  • How are these ip-in-ip tunnels being established ? Is it a full mesh or selective ? >> it is selective between two TOR's, a kind of alternative to ICL
  • Why choose a Link Prober? why can't a BFD >> BFD seems to have challenges with server side.
  • Please have section in HLD describes server requirements for Dual TOR design
  • PR is out for review -
  • Next week will be continued for 10-20 mins discussion on this 
  • For Active-Standby design please check OCP Summit 2021 session Dual TOR use case for Single NIC servers - YouTube (Slideshow)
Release Update: 
  • 202205 release not cut yet due to few issues.
Reply all
Reply to author
0 new messages