Help Required with QSP

[akolthoum]

Hello everyone,
we are using the following switch models: ALE OS6900-T48C6 and ALE OS6900-T20.
Through CheckMK we can see that on some link aggregations output errors are being logged at unspecific times. Neither the link aggregation as a whole nor the individual links are utilized at those times. It is always only one link of the aggregation that is affected. If this link is disabled, the errors then appear on the next link.
Therefore, we have excluded the ports as well as the individual member links of the aggregations as the root cause.
We suspect either a firmware bug or this information:
“The system-default QSP allocates minimal egress queue depth to all traffic classes. Applications with high PPS or burst behavior may trigger tail-drops even under low bandwidth conditions.”
Is anything known about this second point regarding our switch models or the firmware versions we are running (8.9.221.R03 GA and 8.9.78.R01 GA)? Has anyone experienced similar behavior?
Many thanks!

[akolthoum]

Hello, gentil reminder.

[silvio]

I haven't seen this post because it is in the wrong forum. I look into the network part for technical support.
What errors you see in CheckMK. Health port or CRC or or or...?
Is there any special config for the ports of the LAG? Because you are asking about QSP... is your config other than the default qsp? If not than strict priority is working (SP).

I assume that the port with the error is the primary port of the LAG. If you use also for the hashing within the LAG the default config, than BC, UUC and MC are forwarded through the primary port. You can check at the switch the counters per port (show interface 1/1/1). Check if after the next error message at the primary port the counter is increasing (per direction).
There are some solutions: use flood-limit in the whole switche(s) for traffic (BC, MC, UUC). Change the hashing algorithm if you realy need a lot of MC (f.e. for TV or video).
BR Silvio

[akolthoum]

Dear Silvio,

Thank you for your response and welling to help.

As I am new to the whole Alcatel Switches and to this website, I wasn’t sure where can I post my question, however, thank you very much. ´
Below is a screenshot from the error messages we are getting from CheckMK.
For all you other questions, we have standard configuration, nothing especial: OIrts of LAG: nothing, only standard configuration
The ports that have the errors, is indeed the primary port.
Hashing is the default
Counters were reset to check them again, as suggested

You have mentioned that there are some solutions, could you please elaborate more, so I can see if it is applicable with our customer
Error:
Output: [Uplink zu SWITCH-NAME Port 1/1/12 + 2/1/12 und 1/1/14 + 2/1/14, Dynamic Aggregate Number 10 ref aggNumsize 4], (up), MAC: 78:24:59:21:15:46, Speed: 4.29 GBit/s, In: 1.15 GBit/s (26.66%), Out: 3.64 GBit/s (84.86%), Errors out: 0.174% (warn/crit at 0.01%/0.1%)(!!)
Perfdata: outqlen=0;;;; in=143141822.919355;4750000000;4875000000;0;536870911.875 out=455574935.66129;4750000000;4875000000;0;536870911.875 inerr=0;11.077729;110.77729;; indisc=0;;;; inmcast=8.306452;;;; inbcast=7.516129;;;; inucast=110761.467742;;;; innucast=15.822581;;;; outerr=533.83871;30.690495;306.904952;; outdisc=0;;;; outmcast=8.5;;;; outbcast=5.612903;;;; outucast=306357;;;; outnucast=14.112903;;;;
[Uplink zu SWITCH-NAME Port 1/1/12 + 2/1/12 und 1/1/14 + 2/1/14, Dynamic Aggregate Number 10 ref aggNumsize 4]

[silvio]

I see in the message
"outerr=533.83871";30.690495;306.904952;;
and
"inerr=0;11.077729;110.77729;"


Not clear what is this meaning.
I think 4 entries for every port.

Please check all four ports at cli with
"show interface 1/1/12"
which type of error it is.

Normaly there should no error frame (all with 0).

BR Silvio

[akolthoum]

Hi Silvio,

thank you for your help and response.

as i already reset the counters, currently I don´t have any errors. I wan waiting until the errors comes again, and will write you back.