SRX

Anyone out there ever do any CX111 large scale (200+) deployments?  I have an issue with CX111s that won't stay up for over a week.  Every few days the connection bounces.   I'm trying to find out if this is normal behavior on a VZW LTE network for CX111s, or if maybe I should do some digging.

CX111 Code Version: 2.2.2

CX111 Mode: Providing DHCP to SRX

SRX Code Version:  12.1X46-D20.5

Traffic Type over LTE:  OSPF over IPSec Tunnel (in agressive mode)

The connection to the ISP dies and comes back in 30-45 seconds.  So far, there's no real pattern to the dying.  OSPF dead timers expire and it starts up again.  It's like the LTE connection just stops momentarily. 

The customer is currently testing on a non-Juniper branded Cradlepoint and everything seems to be working good after 6 days.  The Juniper branded CX111's don't seem to last that long without bouncing.

Anyone out there encounter anything like this?

No I have not seen this happening, It can be a lot of things, can you attach your config so we can have a look ?

also are others of the 200 CX111 deployments having the same problems ?

It turns out Verizon puts the session in a "dormant state" to free up tower connections after 5 seconds of inactivity and "restarts" the session after traffic flows thorugh.   So it wasn't a modem or config issue.  It wasn't a hard "disconnect", but OSPF shows it as such every few days.  When the OSPF packet fails and tries again, the LTE channel is reserved again and communication works again.

This seems to be limited to the LTE channels, as they need free bandwidth on those channels and they take up more radio resources than 3G channels.

Thanks for the help!

This was a very helpful post to me as it helped to explain why my CX111 modems have been dropping their connections every now and then.  

The trick is to periodically send a keep-alive that will force the Verizon LTE to keep the channel reserved.  For example, I use an rpm probe, that essentially sends a packet once every 5 seconds, something like this:

[edit services rpm]
admin@SRX# show 
probe Probe-Google {
    test DnsSvr {
        probe-type icmp-ping;
        target address 8.8.8.8;
        probe-count 12;
        probe-interval 5;
        test-interval 5;
        hardware-timestamp;
    }
}

It is not foolproof, as this does not always guarantee that the Verizon link will stay up. But having this probe in the configuration has cut down on the link failures by about 60% to 80%.

Clarke Morledge

College of William and Mary