[j-nsp] What is your experience with the EX2200

Chris Lee chris at datachaos.com.au
Tue Dec 12 04:56:17 EST 2017


On Tue, Dec 12, 2017 at 2:29 AM, Kevin Day <toasty at dragondata.com> wrote:

> The problems with the EX2300 I've seen:
>
> * The out-of-the-box version of Junos on them has a clock bug where it
> says its NTP synced, but the system clock gets advanced to 2038 and things
> go crazy. It's sometimes difficult to keep the switch running long enough
> to upgrade it.
>

Yep hit this bug several times but was hard to pinpoint an exact cause,
sometimes it seemed to be worse if I'd powered the switch up and left it
for a while before loading the initial config. After a while I found that
manually setting the time to within a few minutes of correct local time and
rebooting the switch that it would be enough to get the new image deployed.


> * It's also painfully slow to reboot, but not quite as bad as the EX2200.
>

Yeah was equally disappointed to see that boot times hadn't improved any
over the EX2200, in my experience both platforms take a full 5 minutes to
boot and start forwarding traffic.


> * Some old half-duplex devices don't negotiate with the EX2300, but work
> with the EX2200. It's improved with later builds, but not perfect.
>

We similarly had issues with old model Gallagher cardax controllers not
properly negotiating on the EX2300, they used to negotiate themselves to
10Mbps Half-Duplex on the old Cisco 3750 switches just fine, took a bit of
trial and error on the EX2300 configs to work out to just set the port to
10Mbps Full-Duplex with no auto-negotiate and no flow-control and they show
up as working at 10Mbps Full Duplex now.

>
> Overall neither switch has had bad enough issues that I'd shy away from
> them. The 2300 is definitely an upgrade that I'm happy with.
>

Agreed, although the need to purchase a "paper" licence for the 2300 just
to enable virtual-chassis when it was otherwise standard on the EX2200
seems to be a step backwards as well, I'm having to store a whole heap of
envelopes with the paper licence to apparently prove the right to use that
feature.


More information about the juniper-nsp mailing list