Ongoing problems routing on 6200yl

[Bruce Campbell_3]

Our campus wireless environment is routed
on a Procurve 6200yl-24G running K.13.66X.
All campus wireless traffic from 700 Aruba
APs are tunneled to the central wireless
controllers, which connect to the 6200yl which
provides routing (default gateways for
3,000 active wireless users on 26 vlans,
and then ospf routing to the rest of campus,
and beyond).

Starting in April of this year, users
would occasionally lose IP connectivity.
Clearing the ARP table on the 6200yl
would restore service, as would rebooting
it. We enabled a daily reboot of the
6200yl at 5am by cron job, and have avoided
most mid day failures.

A case has been open with HP the whole time,
we've been through several firmware versions.
The daily reboot is still in place, but
every 2 weeks or so, the 6200yl now crashes
in mid afternoon. "show ip" shows most
vlans have vanished. "show ip route"
is corrupted. A reboot restores service.
(A secondary device (3500yl) is setup
with VRRP, but failover doesn't happen
unless we power the 6200yl off).

We have another 25 Procurve devices doing
routing for other applications,
in fact almost our entire campus
of 40,000 ports is Procurve at the edge,
and routed on 3500yl's and 5400zl's.
Note of those are exhibiting the same problem.

Is the 3500yl/5400zl/6200yl series
fundamentally unsuited for a major
routing function, such as campus
wireless ? The major difference
between routing for wireless, and
routing for our wired LAN, is that on
the wire, the MAC/IP bindings are generally
static (several changes per day), while on
wireless, we have a community of 25,000
MAC addresses, of which about 3,000
are active at any given time, hence
major regular turnover in the ARP table.

[Mohammed Faiz]

Hi,

We're running Aruba on a mostly HP network and I haven't seen any issues of that nature so far. Our controllers are connected to 5406s and the routing for the wireless is also performed on a 5406 (which was running K14 but is now back down to K13.71).
Admittedly we don't have close to the same number of APs (100+) as we're still migrating over a considerable number of old 'fat' APs.
I assume you've tried different hardware in place of the troublesome 6200?
The fortnightly crash sounds very strange, is anything logged to the console when that happens?

[Bruce Campbell_3]

We've changed to a 3500yl-24G running
K.13.60, and the crashes have stopped,
but users occasionally lose connectivity
when they roam across campus, until
either the ARP times out on the 3500yl,
or we manually run "clear arp" on it.
(The roam will cause the user to move
from one port to another on the 3500yl,
while keeping the same IP address). Not
all users are affected. A case is open,
we will try to RMA the 6200yl, and
try K.13.71 at a later time.

We have a Cisco 3750 standing by to
swap in if major crashes return.