Rare project developers

[Alexander Gall <>]

On Thu, 17 Feb 2022 15:56:38 +0100, Frédéric LOUI <> 
said:

> NAT example:
> https://bitbucket.software.geant.org/projects/RARE/repos/rare/browse/p4src/include/ig_ctl_nat.p4#106

Yes, that *is* a direct counter attached to a table, but the three
counters I mentioned are *not*. That's the point.

>> I think they are if you use "from_hw = True" when reading the
>> entry. That's what I'm doing in the packet broker.

> I’ll re-test with "from_hw = True », but IIRC (it was along time ago, when 
> we implemented our first counter) 
> it was not reporting counter consistently.

Ok, I can't guarantee that I've seen it working consistently. I'd have
to verify that.

-- 
Alex

>> Le 17 févr. 2022 à 15:45, Alexander Gall <> a écrit :
>> 
>> On Thu, 17 Feb 2022 15:28:29 +0100, Frédéric LOUI 
>> <> said:
>> 
>>> Which SDE are you using ?
>> 
>> 9.7.1
>> 
>>> I’m using SDE 9.7.0 and did not see these errors …
>> 
>>> Basically, operations_execute() sync counters tied to a table.
>> 
>> Yes, but these counters are *not* tied to tables, are they? They are
>> indirect counters (e.g. pkt_out_stats in ig_ctl.p4).
>> 
>>> If you don’t yield this primitive counters belonging to tables are not 
>>> updated at all.
>> 
>> I think they are if you use "from_hw = True" when reading the
>> entry. That's what I'm doing in the packet broker.
>> 
>>> I tried to upgrade my WEDGE to 9.8.0 but it seems that I have the issues 
>>> now …
>>> I’ll get back to you as soon ass as I can solve this upgrade issue.
>> 
>> For the record: I don't see this happening in my 9.8.0 build.
>> 
>> --
>> Alex
>> 
>>>> Le 17 févr. 2022 à 15:17, Alexander Gall <> a écrit :
>>>> 
>>>> Hi Frederic
>>>> 
>>>> The current bf-forwarder code generates these persistent errors (every
>>>> 5 seconds) in bf_switchd:
>>>> 
>>>> Feb 15 12:14:22 spare-PB1 bf_switchd[26808]: 2022-02-15 12:14:22.995012 
>>>> ERROR  (BF_BFRT:(file=null):(func=null):0) - server_op_cb:151 Unable to 
>>>> retrieve bound_program data for client with id 3
>>>> Feb 15 12:14:27 spare-PB1 bf_switchd[26808]: 2022-02-15 12:14:27.944791 
>>>> ERROR  (BF_BFRT:(file=null):(func=null):0) - 
>>>> tableOperationWaitForCompletion:3374 Timed out waiting for completion of 
>>>> operation for table -2081122291
>>>> 
>>>> It appears that this error is caused by
>>>> 
>>>> tbl_stats_in.operations_execute(self.bfgc.target, 
>>>> 'Sync',p4_name=self.bfgc.p4_name)
>>>> tbl_stats_out.operations_execute(self.bfgc.target, 
>>>> 'Sync',p4_name=self.bfgc.p4_name)
>>>> tbl_stats_pkt_out.operations_execute(self.bfgc.target, 
>>>> 'Sync',p4_name=self.bfgc.p4_name)
>>>> 
>>>> in rare/bf_subifcounter/__init__.py. I'm not familiar with
>>>> operations_execute() but I'm a bit confused by the fact that these
>>>> objects are not tables but indirect counters.  Does
>>>> operations_execute() apply to counters too?
>>>> 
>>>> Also, what would be the benefit of the Sync operation compared to
>>>> using "from_hw = True" when reading the counters?
>>>> 
>>>> -- 
>>>> Alex
>>>>