[c-nsp] 2960S drops/packet loss

[Andriy Bilous]

There is a bug in most of SEs for all stackable models, which makes
drop statistics unusable (see CSCso81660 for example - there are lots
of BugIDs with same diagnostics and keep in mind "Fixed-in is a lie).
Reported numbers aren't realistic and often go both ways - increase
and decrease in what appears to be a random manner. sh platform
port-asic stats drop should report right value though. So, before you
start tweaking, you may want to look further into it and just wait for
the sane IOS release. Our SE reported it's fixed in 15.2, which isn't
available for download yet.

On Thu, Dec 22, 2011 at 4:41 AM, John Elliot <johnelliot67 at hotmail.com> wrote:
>
> Hi Guys,
>
> Have a pair of 2960's in a stack, one port(trunk) connects to another DC and we are seeing ~5% packet-lossand large output drops to this DC.
>
>
> #sh interfaces gigabitEthernet 1/0/17 counters errors
>
> Port        Align-Err     FCS-Err    Xmit-Err     Rcv-Err  UnderSize  OutDiscards
> Gi1/0/17            0           0           0           0          0       182867
>
> GigabitEthernet1/0/17 is up, line protocol is up (connected)  Hardware is Gigabit Ethernet, address is a0cf.5b87.ec11 (bia a0cf.5b87.ec11)  Description: QinQ_to_DC2  MTU 1998 bytes, BW 100000 Kbit, DLY 100 usec,     reliability 255/255, txload 41/255, rxload 23/255  Encapsulation ARPA, loopback not set  Keepalive set (10 sec)  Full-duplex, 100Mb/s, media type is 10/100/1000BaseTX  input flow-control is off, output flow-control is unsupported  ARP type: ARPA, ARP Timeout 04:00:00  Last input 6d13h, output 00:00:00, output hang never  Last clearing of "show interface" counters 04:02:15  Input queue: 0/75/0/0 (size/max/drops/flushes); Total output drops: 183592  Queueing strategy: fifo  Output queue: 0/40 (size/max)  30 second input rate 9047000 bits/sec, 2075 packets/sec  30 second output rate 16324000 bits/sec, 2309 packets/sec
>
>
> As you can see, 30sec rate isnt excessive, but as the drops are outdiscards it would appear we are getting hit by the small buffers/microburst issue.
>
> Done a bit of reasearch, and as we have mls qos configured(Need to as we have to trust dscp markings), we needto look at "tweaking" the buffer allocations on the switch to hopefully mitigate these drops.
>
> There appears to be a range of recommendations when it comes to these tweaks - Hoping someone has some suggestions on what to set with "mls qos queue-set output" to alleviate the drops?(start conservative, then apply more aggressive if needed)....and also, does adjusting the buffers require an outage window?
>
> Our traffic is primarily backup(replication which is very bursty), and Internet
>
> Thanks in advance.
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/