I recently purchaed 6 x RB1100AHx2 that we were planning on using to replace our older VIA EPIA based core routers. We switched out one yesterday, and it appeared to work fine. The unit we switched out was our traffic shaper, it has a fair number of queues, and firewall/mangle rules, but the traffic is only 40-80 Mb/s. Average CPU load on core 0 is 35% - 65% while core 1 is mostly 0 % (we are only using one interface atm, we are awaiting a new rack and unitl then the unit is using VLANs on the single interface).
We have so far encountered 3 problems:
Both the main fan and aux fan when activated always run at 12k RPM, not sure if this is normal I thought they would only run at max speed when CPU temperature is high. CPU temp stays around 41 degrees C which to me is not high.
The CPU load on the 1st core seems normal flucuating with load, but the CPU load on the 2nd core seems odd. Since we are using only one interface I wasn’t expecting much from the 2nd core until we re-wire everything so it hovers are 0-2 % as I would expect, most of the time. Periodically the 2nd core will all of the sudden go to 75-99% sit there for 10 mins or so then drop to 0% again. When we use the profile tool it reports queue traffic causing the load. The issue is that not much changes there is no sudden burst of traffic, actually I’ve witnessed this occur on low load <20 Mb/s, while I’ve seen higher loads 75-80 Mb/s not cause this.
Finally the last problem is that the RB1100AHx2 seems to reboot randomly. We ran the unit for 2 weeks on the bench without incident but it was not under the same “real-life” workload. This has happened twice on average once a day since actually installing it. It doesn’t appear to be CPU/bandwidth load related.
Thanks for the info, I believe that the PoE port in te port currently being used. I’ll have to make a trip out to the remote site tomorrow and change that.
Even though this is in an old thread I will update what happened.
We basicly pulled all our initial RB1100AHx2 out of use. We had purchased more of them, and have been using them without incidient as control routers at each of our towers.
The initial batch we bought were put off to the side and we continued running on x86 based hardware. About a week and a half ago one of our x86 based edge routers starting acting up. It ended up being leaking capacitors causing the instability. In a pinch I grabbed one of the original RB1100AHx2’s to run in place. Same odd reboot issues. This time however the error log triggered a thought… RAM.
We swapped out the RAM and just like that the RB1100AHx2 was stable. Turns out upon investigation and testing the initial batch of RB1100AHx2 we recieved had defective ram chips installed. Whenever we build our x86 routers it has always been standard practice to stress test the hardware. I guess thinking that buying a pre-packaged product somehow negated that requirement but lesson learnt it doesn’t.
We ended up testing all the ram in the initial 6 units we purchased by installing them in x86 based computers. Of the 6 units, 4 of them had ram that immediately showed errors in Memtest86. After changing the ram on the RB1100AHx2 they have been running in production for 41d on RouterOS v5.25 without a single hickup.
