On Thu, Dec 19, 2024 at 09:40:20AM +1000, David Gwynne wrote:
>
>> On 19 Dec 2024, at 08:20, David Gwynne <[email protected]> wrote:
>>
>>> On 19 Dec 2024, at 02:17, Maurice Janssen <[email protected]> wrote:
>>> kstat(1) shows me that there are (nearly) no errors on the rx side, but
>>> showed about 470k dropped packets (qdrops) on the tx side for igc2.
>>
>> wait, what?
>
>do you see oactives going up too?
The oactives counters for igc2:0:txq:[0-3] are > 0, but didn't increase
during the last 12 hours. Same for the qdrops.
But at this moment I'm getting about 1200 NTP requests per second, as I
lowered the bandwith setting in the pool a bit (to avoid getting removed
from the pool). At the moment, there seem to be no or very little packet
loss.
I will increase the setting temporarily and see what happens with the
counters.
Part of the kstat output:
igc2:0:txq:0
packets: 1225749955 packets
bytes: 122334582656 bytes
qdrops: 140005 packets
errors: 0 packets
qlen: 0 packets
maxqlen: 1023 packets
oactive: false
oactives: 9433
igc2:0:txq:1
packets: 797916877 packets
bytes: 80690434981 bytes
qdrops: 88268 packets
errors: 0 packets
qlen: 0 packets
maxqlen: 1023 packets
oactive: false
oactives: 10854
igc2:0:txq:2
packets: 800092190 packets
bytes: 80927529361 bytes
qdrops: 29505 packets
errors: 0 packets
qlen: 0 packets
maxqlen: 1023 packets
oactive: false
oactives: 7364
igc2:0:txq:3
packets: 802568095 packets
bytes: 81134090003 bytes
qdrops: 216840 packets
errors: 0 packets
qlen: 0 packets
maxqlen: 1023 packets
oactive: false
oactives: 15490
>> do you have PCIe ASPM enabled in the bios?
No idea, I can check this tomorrow morning.
Maurice
