Dear,
I have a RB750GL in one of our POPs and it routes and authenticates about to 40 clients using Hotspot. We are experiencing a big delay when accessing and browsing through the web management interface of our Rocket M5 that acts as PTMP AP at that location. This happens since we replaced previous RB750 (that freezes every time with 100% CPU usage even with no traffic and acting as bridge, a piece of shit that caused us much trouble) for the current RB750GL, that used to work fine on other site. Also some clients on that location are reporting connection slowness.
First we thought it was a problem with Rocket M5, but replacing it for a new one or replacing cables, power supply, PoE injector and the port it connects on RB750GL and switching from switch chip to bridge didn’t solve it. So we realized such slowness doesn’t happen when both Rockets M5 are accessed from a computer on same network as they are (this was tested in production and in our lab). It isn’t a problem with Rocket M5 since one of them never had such problem before we replaced it for a RB912 on another POP. So we through it could be some problem on RB750GL’s Ethernet port where we connect backhaul PTP radio, but changing to a different one didn’t solve it.
Then I come to our last test: I created a VLAN between Rocket M5 and RB750GL and other VLAN between RB750GL and another RB on our core and bridged them two, then set IP address for VLAN on both ends and the slowness when accessing Rocket M5 for management is gone when done using these addresses!
RB750GL’s firmware is version 3.33, RouterOS was on 6.34.6 and we upgraded it to 6.36.4 (bugfix) and we did a test disabling all firewall filter/NAT/mangle rules. At some step of debugging process all our clients where disconnected on Hotspot and queues were gone and problem persisted, so I don’t think it’s related to Hotspot (even because this problem also happened when we used a Address List to authenticate users some time ago) or queues.
Conclusion: our RB750GL is buggy when routing between different networks.
We are about to remove this RB750GL and just transport clients in L2 directly from Rocket M5 to another RB, as that proved to solve slowness problem.
Just want to know if there is something we are missing or we have actually found a bug.