Community discussions

MikroTik App
 
User avatar
Hammy
Forum Veteran
Forum Veteran
Topic Author
Posts: 776
Joined: Fri May 28, 2004 5:53 pm
Location: DeKalb, IL
Contact:

Bad design?

Sun Dec 14, 2014 3:50 pm

Has anyone else seen something similar?

I had an RB1200 in this location. My upstream changed from port 2 to port 3. My downstream stayed on port 1. Downstream traffic through the device was horrible, limited to single digit megs. Upstream traffic through the device was fine, limited by wireless conditions (about 60 megs). I could do simultaneous TCP BTests, one from an upstream router to this router and one from this router to a downstream router. Both tests ran simultaneously would use the full capacity of the pipe available (90 megs from the upstream and 60 megs to the downstream). The wireless conditions were obviously adequate as simultaneous TCP tests maximized throughput. The difference being these tests were ran to\from the device instead of through the device. Let's try moving the cable and IP addresses to port 2. Everything works fine.

I was on 5.26. I upgraded to 6.19 and then 6.23, making sure to upgrade the firmware each time. No improvement. Let's move the cable and IPs back to port 2. Everything works fine.

Okay, I port 3 must be bad, let's swap this out with a new RB1100AHx2 that I have so that I don't have any future port failures. The new RB1100AHx2 does the same thing in the same situation.
 
Quindor
Member
Member
Posts: 347
Joined: Tue Aug 14, 2012 2:57 am
Location: Noord-Brabant, The Netherlands
Contact:

Re: Bad design?

Sun Dec 14, 2014 8:10 pm

This absolutely makes no sense. Looking at the block diagram's of your hardware all the ports you describe are connected in the same way so there should be 0 difference between them. Blocks diagram for RB1200 and RB1100AHx2

I would try and look at different causes.
- Did you check sync rate of the specific port when a cable was connected?
- Maybe the cable isn't that good anymore, have you tried replacing the cable?
- When it slowed down so much, did you notice a lot higher CPU usage on the Routerboard?
- Did you switch all configurations correctly, making sure the port is not part of some other bridge or filter rule or something like that?
- Please only do transfer tests without wireless involved
- 'Megs' is a meaningless term which doesn't help us. Either use Megabit, Mbit or Mb/s. Or use MegaByte or MB/sec to specify clearly what kind of transfer rates you are talking about
- Depending a little bit on config but you should easily be able to get 1000Mbit or 110MB/sec on the ports you speak of. Since they are connected to the same switch internally (which has a 1000Mbit uplink to the CPU) it might be wise to distribute the ports in such a way that your traffic doesn't hit the same switch twice. With that said, the switches (And all ports) have a full-duplex connection so can utilize 1000Mbit up and down at the same time
- Your model also has dedicated CPU ports not going through the switch, these can best be used for the interfaces that get a lot of traffic (Eth6, Eth7, Eth8)

Hopefully with this method we will be able to find out what happened!
 
User avatar
Hammy
Forum Veteran
Forum Veteran
Topic Author
Posts: 776
Joined: Fri May 28, 2004 5:53 pm
Location: DeKalb, IL
Contact:

Re: Bad design?

Sun Dec 14, 2014 8:49 pm

I know the block diagrams.

Yeah, all ports were 100/Full. There were no Ethernet Rx/Tx errors. Wireless would move about 60 megabit (limit of 20 MHz channel with live traffic) on one interface and 90 megabit (limit of FastE on Rocket) on the other interface when going to or from that RB. Going through is limited to a gross of about 15 going in Eth3 and then out Eth1. Move Eth3 to Eth2 (yesterday on RB1200) or Eth6 (today on RB1100AHx2) and the through traffic returns to normal.

Profile shows nothing significant. Usage was under 10% or so.

There are no bridges involving ports (only loopback), queues, mangles or interface specific filters.

Wireless is more than sufficient to show the issue.

Megabit. Who uses megabytes when referring to communications?

It works fine when using Eth1 and Eth2, but not Eth3.

I wouldn't expect a lot of traffic in any of these scenarios given the limited wireless links.
 
User avatar
NathanA
Forum Veteran
Forum Veteran
Posts: 829
Joined: Tue Aug 03, 2004 9:01 am

Re: Bad design?

Tue Dec 16, 2014 3:21 am

I have never heard of this nor witnessed anything like this, nor does it make any sense.

Both here and on the mailing list, you kept referring to ether2 vs ether3. I was going to ask, well, for the purpose of troubleshooting/eliminating variables, have you bothered to try *other* ethernet ports, or did you just stop with ethers 2 and 3? I see that you have tried ether6 now, which is the first port on the second switch chip, and that one works, so that's interesting. How about now trying ether4 or 5 (the remaining 2 ports on the first switch), and anything in ether7-10 (the other ports on the second switch)? If the same pattern holds, you should get good performance on ether7, not on ether8, and ether4-5 and ether9-10 are unknowns.

EDIT: I was referring to 1100AHx2 port #s above, not 1200 port #s.

-- Nathan
 
User avatar
Hammy
Forum Veteran
Forum Veteran
Topic Author
Posts: 776
Joined: Fri May 28, 2004 5:53 pm
Location: DeKalb, IL
Contact:

Re: Bad design?

Tue Dec 16, 2014 4:11 am

I have never heard of this nor witnessed anything like this, nor does it make any sense.

Both here and on the mailing list, you kept referring to ether2 vs ether3. I was going to ask, well, for the purpose of troubleshooting/eliminating variables, have you bothered to try *other* ethernet ports, or did you just stop with ethers 2 and 3? I see that you have tried ether6 now, which is the first port on the second switch chip, and that one works, so that's interesting. How about now trying ether4 or 5 (the remaining 2 ports on the first switch), and anything in ether7-10 (the other ports on the second switch)? If the same pattern holds, you should get good performance on ether7, not on ether8, and ether4-5 and ether9-10 are unknowns.

EDIT: I was referring to 1100AHx2 port #s above, not 1200 port #s.

-- Nathan
The RB1100 is in production now, but I have the RB1200 here. I'll run through it and see what comes up for the other ports. Same config as now, I'll put two laptops on and give 'em hell. Sometime...

If it works, find what makes it break. If it's broken, find where it does work.

Who is online

Users browsing this forum: Fl3tch and 29 guests