CLOUD CORE ROUTER

Mon Jul 22, 2013 1:08 pm

Latest v6.2 build solves CCR unresponsiveness when under DDoS attack. If you don't know where to get this pre-release, email support.

Bitto · Mon Jul 22, 2013 2:52 pm

I have to disable 500M/500M Up/Down simple queue on our Ros 6.1 CCR1036 12G-4S. Average load is 150Mbit with 100000 packets voice clients started to complain about drops and jitter i tried large queues small quese nothing has helped. Enabling and disabling queues causes high cpu load and packet losses. Any ideas or fixeses?

otgooneo · Mon Jul 22, 2013 4:09 pm

I have 6 CCRs and we are using those routers for PPPoE server. I have a plan to enable BGP on that routers. Anyone tried to use BGP with Cisco using v6.2, please share your experience here. Thanks.

SwissWISP · Mon Jul 22, 2013 4:22 pm

I have 6 CCRs and we are using those routers for PPPoE server. I have a plan to enable BGP on that routers. Anyone tried to use BGP with Cisco using v6.2, please share your experience here. Thanks.

We run several sessions between our CCRs and some Ciscos and even a session to a Brocade CER. No Problems so far...
The only negative thing I can say about BGP on the CCR is that it runs only on one core, which makes it quite slow. But if you don't have to load the whole Internet Routing table on every Session, you'll be OK.

- Mat

otgooneo · Mon Jul 22, 2013 4:30 pm

I have 6 CCRs and we are using those routers for PPPoE server. I have a plan to enable BGP on that routers. Anyone tried to use BGP with Cisco using v6.2, please share your experience here. Thanks.
We run several sessions between our CCRs and some Ciscos and even a session to a Brocade CER. No Problems so far...
The only negative thing I can say about BGP on the CCR is that it runs only on one core, which makes it quite slow. But if you don't have to load the whole Internet Routing table on every Session, you'll be OK.

- Mat

Hey Mat. Thanks for your good news. Could you please tell me what ROS version you use?

SwissWISP · Mon Jul 22, 2013 6:07 pm

Hey Mat. Thanks for your good news. Could you please tell me what ROS version you use?

Currently we use 6.2.

rpingar · Tue Jul 23, 2013 8:46 am

Hi Normis,

the problems seems not be the responsivness, but the capacity to handle the packets under some attack, like the icmp by hping3.
We don't care if it is responsive but unable to pass traffic.

regards
Ros

Latest v6.2 build solves CCR unresponsiveness when under DDoS attack. If you don't know where to get this pre-release, email support.

Chupaka · Tue Jul 23, 2013 2:17 pm

that's how they called it. recheck the new build and see whether the problem has gone

Tue Jul 23, 2013 2:24 pm

Yes, that should be fixed, Rosario.

0005 · Tue Jul 23, 2013 2:28 pm

Hi Normis,

What about my requests ? Also shoud be fixed ?

krakenant · Wed Jul 24, 2013 6:11 pm

Normis, do you have any idea when all aspects of ROS will be optimized for the Tilera Chipset? We keep trying it and different parts fail. Right now the issue is throughput on a hotspot when we get above about 400 users. We have X86 devices that can do 3 times the number of users and bandwidth.

OlegM · Wed Jul 24, 2013 6:14 pm

A few months I can not deal with such a problem. In version 5, this was not. In 6.0 and 6.1 by connecting some users, a connection error does not issue, but internet does not work. It creates a dynamic connection, but it is not running. This is a problem with the version?
Noticed that this happens when the ccr not reboot more than a few days. Interfaces and ppp in winbox beginning to slow and long open and just at that momet these glitches occur.

minelli · Wed Jul 24, 2013 7:41 pm

I would like to report a problem that has been occurring in the CCR:

When working with IP + MAC interface and the option of customers ARP = reply-only after restarting the CCR band travels not normal.
It is as if the traffic is limited.
If after the restart put ARP = enable and then ARP = reply-only interface the traffic back to normal consumption.

otgooneo · Fri Jul 26, 2013 12:51 pm

I have same problem as OlegM.

rpingar · Fri Jul 26, 2013 3:17 pm

we too, on x86 and ccr, it is a 6.x problem

regards
Ros

I have same problem as OlegM.

whoknew · Fri Jul 26, 2013 4:40 pm

I have 3x CCR1036-12G-4S's and one of them has been rebooting non stop. 1 of them is on firmware 3.05 with RouterOS being v6.1, another is on v6.0 RC14 and on firmware version 3.05. The one that was rebooting was on v6.1 with firmware version 3.08. I updated it to version 3.09 and the reboots seem to have stopped.

Has anyone else run into this? I noticed some others complaining about reboots but no one had come out and said it was fixed etc. I had seen the SNMP disable etc, but I tried that and it was still rebooting.

adiputrasa · Sat Jul 27, 2013 7:29 am

BGP process only using 1 core, whenever i tried to change routing filter in BGP like prepend or anything after that I cannot connect to my winbox. SSH normal only winbox I cant use. Anyone got a similar problem with me?

morf · Sun Jul 28, 2013 12:36 am

BGP process only using 1 core, whenever i tried to change routing filter in BGP like prepend or anything after that I cannot connect to my winbox. SSH normal only winbox I cant use. Anyone got a similar problem with me?

I also have problems with BGP when loaded only one core

morf · Sun Jul 28, 2013 6:53 pm

After 8 days of work, CCR1036 began working very slowly. After the restart, everything comes back to normal.

morf · Sun Jul 28, 2013 10:52 pm

In the last 8 days, when CCR1036 began to work slowly, there was an increased memory consumption.

sonny · Mon Jul 29, 2013 1:48 pm

@morf

have you enabled SNMP on that device?
if yes try to disable and test again.

sonny · Mon Jul 29, 2013 1:56 pm

@0005

Another one issue regarding ping losses.
Equipment: CCR1036-12G-4S
Tested on: 6.1
Ports: Full Duplex, 1Gbs
Scheme: Mikrotik is transparent, no any configuration in Firewall and Queues.
sfp1 and sfp2 are in bridge mode. sfp1 is connected to CSN (WAN) sfp2 is connected to LAN.
Traffic load: avarage for DL 400Mb/s, for UL 100Mb/s
Avarage users: 4500
CPU load: max 15%
RxTx errors and drops: 0

The problem is when connecting CSN to LAN without Mikrotik no any ping losses to users and from them. But when connect Mikrotik between CSN and LAN
each 3-4 pings you get 1-2 ping losses to each user and traffic decreases.
Changing optical fiber, optical ports and modules, changing from optical to electrical ports didn't get any results.

We have the same Situation Wan=SFP und internal=SFP in ad Bridge. (ROS6.2 would like to use as shaping device)
We discoverd the same failure. If you oing through the device pings (packets) get lost.

Now we disabled SNMP and it becomes better.
Also we set connection tracking from auto to on.

morf · Mon Jul 29, 2013 2:07 pm

@morf

have you enabled SNMP on that device?
if yes try to disable and test again.

snmp off.
I sent a support mikrotik supout files and screenshots.
I use to build Jul/17/2013. In the tech support told me that I updated to the last. What is the point? In order to reproduce the situation need to wait for about 8 days. And in the 8 days still come out new builds. A support will continue to say that I have not the latest build

The idiocy!

CarulloS · Tue Jul 30, 2013 2:44 am

I posted this in another thread and I don't want to repeat it here. Not sure if it should have gone under this thread for CCR, but...

Here is the link: http://forum.mikrotik.com/viewtopic.php?f=1&t=75259

Basically, I changed CCR MTU size and it caused kernel panic and rebooted. Want to make sure its looked into.

Thanks,
Scott

0005 · Tue Jul 30, 2013 10:07 am

@0005
Another one issue regarding ping losses.
Equipment: CCR1036-12G-4S
Tested on: 6.1
Ports: Full Duplex, 1Gbs
Scheme: Mikrotik is transparent, no any configuration in Firewall and Queues.
sfp1 and sfp2 are in bridge mode. sfp1 is connected to CSN (WAN) sfp2 is connected to LAN.
Traffic load: avarage for DL 400Mb/s, for UL 100Mb/s
Avarage users: 4500
CPU load: max 15%
RxTx errors and drops: 0

The problem is when connecting CSN to LAN without Mikrotik no any ping losses to users and from them. But when connect Mikrotik between CSN and LAN
each 3-4 pings you get 1-2 ping losses to each user and traffic decreases.
Changing optical fiber, optical ports and modules, changing from optical to electrical ports didn't get any results.
We have the same Situation Wan=SFP und internal=SFP in ad Bridge. (ROS6.2 would like to use as shaping device)
We discoverd the same failure. If you oing through the device pings (packets) get lost.

Now we disabled SNMP and it becomes better.
Also we set connection tracking from auto to on.

Thank you for advise. I cheked SNMP is off. I changed connection tracking to yes and aftre tests I'll confirm whether problem is solved or not.
One question regarding Shaping. Did you try Upload and Download simultaneously. For ex. if in queue tree Max 2M for DL and Max 2M UL for 1 user does this user get 2M for DL and 2M for UL simultaneously ?

jarise · Tue Jul 30, 2013 8:04 pm

Hi,
Is it normal that In ressources IRQ, the fancon is generating about 1100 irq per second ?

6.0RC8 25/1/2012

Wondering same situation with 6.1. That looks like some trouble.. first cpu load 5-9% without network traffic.

Is this normal?

sonny · Tue Jul 30, 2013 10:54 pm

@0005

Thank you for advise. I cheked SNMP is off. I changed connection tracking to yes and aftre tests I'll confirm whether problem is solved or not.
One question regarding Shaping. Did you try Upload and Download simultaneously. For ex. if in queue tree Max 2M for DL and Max 2M UL for 1 user does this user get 2M for DL and 2M for UL simultaneously ?

bulid is from 16th July 13:48

We have real internet traffic peak arroud 160 mbit with 438 simple queuss on it.
Bridge settings use IP firewall on
If a user has 10M down and 1M up and 1M up is full the user is gettin only arround 8.5-9Mbit down.
But it seems with high load 140-160mbit some packages on bridge ar lost.

mishaM · Thu Aug 01, 2013 11:16 am

Hi normis ,

I upgrade my router 6,2 version , reset-factory router and make config manualy , after 4 day working managment died , cannot connect via winbox , telnet (i can't type username (shows only login page ) , ssh , and api ) works only mac telnet , also dies mpls/ vpn/pptp , memory usage when happens problem is 596MB .
also kernel crashes when use : torch,traffic-generator,when changed graphing ip policy access

honzam · Thu Aug 01, 2013 9:22 pm

CCR1036 - ROS 6.2
389VLAN + 286 NAT and DHCP. NO shaping. Only routing + OSPF. Real trafic
Troughput 837Mbit and CPU only 5%

Works well

Fri Aug 02, 2013 10:00 am

new build of v6.2 today, for anybody living on "the bleeding edge"

Ivoshiee · Fri Aug 02, 2013 11:31 am

new build of v6.2 today, for anybody living on "the bleeding edge"

Nice, but what is it supposed to fix or change?

Fri Aug 02, 2013 11:40 am

new build of v6.2 today, for anybody living on "the bleeding edge"
Nice, but what is it supposed to fix or change?

only bugs introduced in pre-release, so changelog will not include them (changelog only shows fixes since v6.1)

mishaM · Fri Aug 02, 2013 4:34 pm

normis i upgrade to last change and after 2 hours memory usage is ~450MB ,

mishaM · Sat Aug 03, 2013 1:40 pm

after 19 working hour happens same problem ,

Ansy · Mon Aug 05, 2013 9:57 am

after 19 working hour happens same problem ,

mishaM,
May be it's simply "DNS Amplification Attack", as I saw on my RB751U-2HnD.
Eating memory, process dns eating CPU, many DNS queries in DNS cache (my case was xcqv.de from 178.100.10[0123].*)
I've switched off IP -> DNS -> Allow Remote Requests, and no memory leaks / CPU overloads any more

thehoff · Tue Aug 06, 2013 4:31 am

There have been a lot of +1's for multicore BGP and I do agree....

I would really like to know when/if there is going to be an update where doing a /ip route print does not rely on the CLI and a whole core to view routes.

We have a pair of 1016's in production at the moment and diagnosing a route inside a BGP enabled RB that has static/policy/bgp(2feeds) is almost impossible with the current operation of the command.

Thoughts?

mishaM · Tue Aug 06, 2013 2:24 pm

after 19 working hour happens same problem ,
mishaM,
May be it's simply "DNS Amplification Attack", as I saw on my RB751U-2HnD.
Eating memory, process dns eating CPU, many DNS queries in DNS cache (my case was xcqv.de from 178.100.10[0123].*)
I've switched off IP -> DNS -> Allow Remote Requests, and no memory leaks / CPU overloads any more

same situation

armandfumal · Tue Aug 06, 2013 9:48 pm

BGP multicore become a hurry...
I use BGP with 3 full feed. It works but BGP handling is very slow...
route query take 2 mins to seach a prefixe...

morf · Wed Aug 07, 2013 4:26 pm

BGP multicore become a hurry...
I use BGP with 3 full feed. It works but BGP handling is very slow...
route query take 2 mins to seach a prefixe...

I have the same situation, but with a 2-BGP sessions.

Wed Aug 07, 2013 4:30 pm

searching for prefixes has nothing to do with multicore or routing package, this is console speed.

morf · Wed Aug 07, 2013 4:45 pm

searching for prefixes has nothing to do with multicore or routing package, this is console speed.

Normis, but you still can not deny the fact that with a BGP-FullView work is very slow?
Try to load 2 "FullView" BGP table and see how it will work CCR

Wed Aug 07, 2013 4:46 pm

searching for prefixes has nothing to do with multicore or routing package, this is console speed.
Normis, but you still can not deny the fact that with a BGP-FullView work is very slow?
Try to load 2 "FullView" BGP table and see how it will work CCR

this does not affect routing performance at all.

morf · Wed Aug 07, 2013 4:54 pm

searching for prefixes has nothing to do with multicore or routing package, this is console speed.
Normis, but you still can not deny the fact that with a BGP-FullView work is very slow?
Try to load 2 "FullView" BGP table and see how it will work CCR
this does not affect routing performance at all.

Affect the performance of the BGP. To work BGP allocated one core. The routing table updates slowly.
I still recommend you to try to get to CCR 2 FullView-table and look at the load on the cores and the speed of routing updates.

Wed Aug 07, 2013 5:18 pm

It was mentioned in several topics already.

Yes, you will get slow route updates.
Yes, route print will be slow.
No, it does not affect overall forwarding performance.
Yes, we are working on new routing implementation which will work much faster.

morf · Wed Aug 07, 2013 5:21 pm

It was mentioned in several topics already.

Yes, we are working on new routing implementation which will work much faster.

Thank you for your reply, when will the first implementations?

mudasir · Thu Aug 08, 2013 1:24 am

Hi everyone.

I am facing a very odd issue. Please check this post where I have detailed the issue http://forum.mikrotik.com/viewtopic.php?f=2&t=75556

raz · Thu Aug 08, 2013 10:11 am

Thank you for your reply, when will the first implementations?

Read in a another Topic, about the MUM with v7 in Sep...

It was mentioned in several topics already.

Yes, you will get slow route updates.
Yes, route print will be slow.
No, it does not affect overall forwarding performance.
Yes, we are working on new routing implementation which will work much faster.

I can confirm this, BGP is really slow at startup, i get over one Uplink 450k Prefixes and it takes 2-3 Minutes before
its complete and the Router is completly ready. There is no Speed Limit in deliver the Prefix Updates in the Uplink, its
only MikroTiK they work with a single core...

jmllh513 · Fri Aug 09, 2013 7:01 pm

Performance! MikroTik RouterOS like ordinary of all the V6 router. Outdoor cabinet temperature range -20 degrees to 60 degrees with a version that would be great in the tower foundation.

Asket · Fri Aug 09, 2013 8:01 pm

CCR1036-12-4S disappoint their instability.
1 router - 300 IPoE, 1 BGP FullView uplink, 3 BGP peers, about 30 EoIP links, 1GBit/s aggregated traffic - 30 % CPU, stable, all good
2 router - 1100 PPPoE sessions, 700 mbit/sec aggregated, 80000 pps . Sometimes (5-6% of connections) forget routes to client IP - client haven't internet access, no traffic in pppoe-interface. 25 % CPU without connection-tracking, with connection-tracking -100% load and hard packet loss.
3 router - 800 PPPoE sessions, 400 mbit/sec - 40% CPU, very unstable SNMP (screenshot in attach), empty in interface and queue list.

Аlas, CCR1036-12-4S does not meet the manufacturer's claims.

miahac · Sat Aug 10, 2013 6:07 pm

I have had snmp disabled on my bgp routers for years because of slowness, It seems SNMP is a huge stack but most people only want interface and processing statistics. What would be nice is a SNMP-compact package that has just that.

Thoughts

armandfumal · Sat Aug 10, 2013 8:49 pm

Normis. Making a query to search for a prefixe that take more than 2 or 3 mins is not acceptable

For network engineer is a day to day task. When you have to solve a routing issue with ccr it is a pitty

On my old cisco these tasks are in real time.

Envoyé depuis mon GT-I9505 avec Tapatalk

jmllh513 · Sun Aug 11, 2013 2:17 am

Childish I know, but we are taking about the router of the year.....

jmllh513 · Sun Aug 11, 2013 2:21 am

Looks like product naming gives lots of specs:

CCR10<ab>-<c>G-<d>S

<ab> - number of cores
<c> - number of Gigabit Ethernets
<d> - number of SFP cages

-BU -board unit (without Rackmount case), for installations in your own boxes

Mon Aug 12, 2013 9:45 am

-BU -board unit (without Rackmount case), for installations in your own boxes

We aldready sell that http://routerboard.com/CCR1016-12G-BU

every · Tue Aug 13, 2013 10:40 pm

This is what? Thank you!

morf · Wed Aug 14, 2013 11:03 pm

This is what? Thank you!

LOL

Try to connect by ip address.

every · Thu Aug 15, 2013 8:50 am

This is what? Thank you!
LOL
Try to connect by ip address.

I'm trying to IP and MAC addresses, but is not connected to the program.

morf · Thu Aug 15, 2013 11:34 am

This is what? Thank you!
LOL
Try to connect by ip address.
I'm trying to IP and MAC addresses, but is not connected to the program.

Try to change the cable into a different port. (eth1 -> eth2).
Just make sure that you have a registered IP address from the subnet mikrotik.
I think your problem is not related to ROS 6.2

raz · Sat Aug 17, 2013 5:07 pm

Press … and select the MAC Address directly, then connect.

armandfumal · Thu Aug 22, 2013 12:08 am

Press … and select the MAC Address directly, then connect.

ROS 6.2 and CCR1036

same for me. I can't connect with winbox and mac, even mac server configuration correctly configured

Thu Aug 22, 2013 11:11 am

Press … and select the MAC Address directly, then connect.
ROS 6.2 and CCR1036

same for me. I can't connect with winbox and mac, even mac server configuration correctly configured

+1 I have 8x CCR 1036 I am unable to connect to with MAC-Winbox. Have tried 6.0rc5 (default) and 6.2

longerCZ · Thu Aug 22, 2013 2:31 pm

Hi all,
we have some strange issue with CCR-1036-12G-4S, every time we turn on cold device (after 15 mins and more of powerless state) it gets stuck forever on "loading kernel from nand" notice shown on embedded display. When I then (after "heating the device" with kernel stuck) reconnect power cable it boots normally. Any ideas?
This happenned in 6.1 and 6.2 the same.

Thanks

lmatys · Fri Aug 23, 2013 1:45 pm

Hi. I'm planning the change HP ML 150 Quad Xeon 1.6 Ghz which runs many pppoe servers + htb.
I'm thinking about buying MikroTik Cloud Core Router CCR1016-12G.

The question is:
- does the MikroTik Cloud Core Router CCR1016-12G share CPU load to all available cores, when using many pppoe servers at many interfaces/vlans ?
- does it share the load to all cores when keeping traffic from all PPPOE sessions?
- does it share cpu load to all cores, when using simple queue?

Best regards.

L.M.

infused · Sat Aug 24, 2013 3:43 pm

Hi. I'm planning the change HP ML 150 Quad Xeon 1.6 Ghz which runs many pppoe servers + htb.
I'm thinking about buying MikroTik Cloud Core Router CCR1016-12G.

The question is:
- does the MikroTik Cloud Core Router CCR1016-12G share CPU load to all available cores, when using many pppoe servers at many interfaces/vlans ?
- does it share the load to all cores when keeping traffic from all PPPOE sessions?
- does it share cpu load to all cores, when using simple queue?

Best regards.

L.M.

From what I've seen so far, no. Someone can correct me though.

Rom1 · Mon Aug 26, 2013 12:18 pm

Do you plan to release the device only with the SFP-slots or combo?

raz · Tue Aug 27, 2013 10:47 am

Mhm im getting since 8 Days about my Ticket with slow TCP Performance: 2013081766000271, on my Question Sent 24 Hours ago, there was also no answer.

armandfumal · Fri Aug 30, 2013 1:08 am

Have tested 6.3 29/08/2013 12:06 Build on CCR1036

I see a gain in BGP loading prefix. Now I have loaded 458000 prefixes in 50 secondes... much better

still on one core but better performances.

good job.

pochbba · Sat Aug 31, 2013 3:09 am

I'm noticing is impossible in a real esenario work more traffic through an interface that is I show below, I want to clarify that and replaced eh SFP fiber modules which are certainly cisco

infused · Sun Sep 01, 2013 1:08 pm

Press … and select the MAC Address directly, then connect.
ROS 6.2 and CCR1036

same for me. I can't connect with winbox and mac, even mac server configuration correctly configured
+1 I have 8x CCR 1036 I am unable to connect to with MAC-Winbox. Have tried 6.0rc5 (default) and 6.2

I know that Windows8 throws a wobbly with connecting to macs...

drpater · Fri Sep 06, 2013 4:56 pm

Hello all,
CCR 1036 RoS 6.3,
/ip hotspot cookie list contain only 4096 entries, this is bug or feature?

ericsooter · Sun Sep 08, 2013 6:15 pm

I have a CCR1036 and want to swap it out with my 1100AHx2. With 6.3, is it ready for live production with BGP, OSPF and some rate-limiting queues? Backhaul bandwidth is around 300MB. Just wondering if anyone is using this as a core router with a similar setup? If not, I can wait a few revisions for the software to stabilize.

Sun Sep 08, 2013 6:36 pm

I have a CCR1036 and want to swap it out with my 1100AHx2. With 6.3, is it ready for live production with BGP, OSPF and some rate-limiting queues? Backhaul bandwidth is around 300MB. Just wondering if anyone is using this as a core router with a similar setup? If not, I can wait a few revisions for the software to stabilize.

As CCR SW is not paralellized I guess the 1100AHx2 is faster for the moment.

ropeba · Sun Sep 08, 2013 6:37 pm

Hi, I use CCR with BGP, OSPF, QUEUE... 800 Mbps, everything works ok.

             uptime: 6w3d7h4m30s
            version: 6.1
         build-time: Jun/12/2013 11:50:54
        free-memory: 3291.1MiB
       total-memory: 3969.0MiB
                cpu: tilegx
          cpu-count: 36
      cpu-frequency: 1200MHz
           cpu-load: 13%
     free-hdd-space: 901.5MiB
    total-hdd-space: 1024.0MiB
  architecture-name: tile
         board-name: CCR1036-12G-4S
           platform: MikroTik

janisk · Mon Sep 09, 2013 1:06 pm

As CCR SW is not paralellized I guess the 1100AHx2 is faster for the moment.

actually some important parts are parallelized. You can get a lot more throughput that with RB1100AHx2

Dashka · Tue Sep 10, 2013 5:32 am

Hello

I am plan cloud core router to use create less then 10 or more meta router, is that possible?
before i ask about this solution on Mikrotik RB1100 but this router dual core CPU it's not possible.

Thanks
Dashdorj

ericsooter · Tue Sep 10, 2013 6:11 am

Hi, I use CCR with BGP, OSPF, QUEUE... 800 Mbps, everything works ok.

Thanks guys for the info. I'll move forward.

manson · Tue Sep 10, 2013 9:31 am

Hello

I am plan cloud core router to use create less then 10 or more meta router, is that possible?
before i ask about this solution on Mikrotik RB1100 but this router dual core CPU it's not possible.

Thanks
Dashdorj

CCR doesn't support meta routers yet.

infused · Tue Sep 10, 2013 2:32 pm

As CCR SW is not paralellized I guess the 1100AHx2 is faster for the moment.
actually some important parts are parallelized. You can get a lot more throughput that with RB1100AHx2

Can you expand on what is? Also curious to how the router, on 6.3, is now carrying load over the cores.

Reason I ask is that during the RCs, you were just getting the OS stable on CCR, core improvement was going to come later. Just wondering where that is now, and when we will realize the full potential of this router.

Cheers.

Tue Sep 10, 2013 2:33 pm

Nearly everything utilizes the cores now. Just upgrade to v6.3 and check the "/tool profile" menu to see how load is distributed.

manson · Tue Sep 10, 2013 3:25 pm

One thing that's very important maybe for me only (DNS forwarder) still doesn't utilize more then one core ;(

brianlewis · Tue Sep 10, 2013 9:49 pm

Anyone have a status update for the CCR-1036-8G-2S+ availability? It was posted as being available in JUNE......

http://forum.mikrotik.com/viewtopic.php?t=70761

CCR1036-8G-2S+ - 10Gigabit CCR, 2x SFP+, 8x Gigabit Ethernet, 36 Core Tilera, 16GB RAM, 28Gbit throughput, 42 Million PPS, Available June

infused · Wed Sep 11, 2013 1:25 am

Nearly everything utilizes the cores now. Just upgrade to v6.3 and check the "/tool profile" menu to see how load is distributed.

Thanks for that. Does that mean that say firewall will use the resources of all cores, or just a bunch of cores?

m3gaman · Wed Sep 11, 2013 2:36 am

Looks like they're well balanced!! Nice work!!

n3c8 · Wed Sep 11, 2013 5:47 pm

So are we saying that BGP now runs on multiple cores. Is this from 6.3 production or a RC?

doush · Wed Sep 11, 2013 7:49 pm

Nearly everything utilizes the cores now. Just upgrade to v6.3 and check the "/tool profile" menu to see how load is distributed.

What about ip>firewall>mangle ?
And also PPPoE Server ?

lavv17 · Fri Sep 13, 2013 8:13 am

pcq on ccr seems to work incorrectly for large number (>500) of concurrent clients. Are there any internal limitations?

dbodeenisudm · Wed Sep 18, 2013 7:06 pm

Mhm im getting since 8 Days about my Ticket with slow TCP Performance: 2013081766000271, on my Question Sent 24 Hours ago, there was also no answer.

what is your tcp performance problem? I am having a problem; when I run a single connection tcp bandwidth test out the wan/fiber port I get 5M download and 60M upload. if I run udp I get 90M each way. Its either the router or the upstream provider. I am leaning toward router; on version 6.2.

raz · Wed Sep 18, 2013 7:51 pm

Mhm im getting since 8 Days about my Ticket with slow TCP Performance: 2013081766000271, on my Question Sent 24 Hours ago, there was also no answer.

what is your tcp performance problem? I am having a problem; when I run a single connection tcp bandwidth test out the wan/fiber port I get 5M download and 60M upload. if I run udp I get 90M each way. Its either the router or the upstream provider. I am leaning toward router; on version 6.2.

This is the same Problem like yours! UDP works as hell, i can get nearby 1GBit over but TCP sucks. Still no fix from MikroTiK. The last what i got was:

Hello,

I have tried to test bandwidth-test by running it to your specified address and my
results and speeds were very different from ones that are visible in picture that
you sent (they were all right).

Most interesting thing is that while I was typing this message, I tried to run
test again and it was running exactly how picture shows it.

I also looked through your configuration and seems that there is nothing that
could block this traffic.

I suppose there is something that blocks this traffic. Could you try to run test
and if speed is slow then reboot it and test again?

Also you could try to install RouterOS version 6.3 and that maybe could fix the
issue. RouterOS is accessible at www.mikrotik.com.

Regards,
Martins

Yes, i made an Upgrade to 6.3 and 6.3 but this is occours always...

onnoossendrijver · Wed Sep 18, 2013 8:03 pm

Come on guys, this is said 1000 times before! NEVER do a TCP test on a router, only THROUGH a router.
The CPU in 99% of the routers is not fast enough to do all the tasks necessary for a TCP test.
If you want to do a decent TCP test, use a PC on both sides of the device or link you want to test and run IPerf between those PCs.

I have tested a chain of 5 Mikrotik devices using this method, including RB600A, RB450G, RB1100x2 and x86 180-200 mbit/s TCP.
One of the links between these devices is a NV2 wireless link.

raz · Wed Sep 18, 2013 8:19 pm

Ofcourse i tested FROM Servers behind the CCR to me, and 3-4 MBit/s are not OK.
Im own a 100 MBit Home ADSL, and there is no Reason why i only get 3-4 MBit TCP but
nearby 1 GBit UDP (Even i connect in DC directly).

onnoossendrijver · Wed Sep 18, 2013 8:40 pm

Ok, ignore my previous post then. Sorry!
This smells a bit like an autonegotiation problem.
Are you sure all links negotiate at full-duplex?

dbodeenisudm · Wed Sep 18, 2013 8:49 pm

Ok, ignore my previous post then. Sorry!
This smells a bit like an autonegotiation problem.
Are you sure all links negotiate at full-duplex?

hi thanks for assuming we are idiots. yes I have tested from the router and through the router and everywhere. when I run the speed test the cpu does not go above 10% for a one tcp connection count test. I am running from a ccr to a x86 mikrotik machine. links are negotiating at full duplex. it is a problem with the ccr fiber interface I think.

onnoossendrijver · Wed Sep 18, 2013 10:27 pm

Ok, ignore my previous post then. Sorry!
This smells a bit like an autonegotiation problem.
Are you sure all links negotiate at full-duplex?
hi thanks for assuming we are idiots. yes I have tested from the router and through the router and everywhere. when I run the speed test the cpu does not go above 10% for a one tcp connection count test. I am running from a ccr to a x86 mikrotik machine. links are negotiating at full duplex. it is a problem with the ccr fiber interface I think.

Aren't you maxing out 1 core of the CCR when running a TCP test?

raz · Wed Sep 18, 2013 11:08 pm

Aren't you maxing out 1 core of the CCR when running a TCP test?

Nope, there is no huge load, not more as 4-6% per Core. BGP uses btw always 100%, but this is just Wayne, because
RouterOS did it since im started using BGP...

Edit:

See Uploaded File, im trying to Move a File through FTP (TCP Port 21)

dbodeenisudm · Thu Sep 19, 2013 7:46 pm

I am running a tcp bandwidth test where connection count=1. If that maxes out whatever then man this router really stinks. But now nothing is getting maxed out. Its either our upstream provider or the fiber interface on the mikrotik.

onnoossendrijver · Thu Sep 19, 2013 11:17 pm

I am running a tcp bandwidth test where connection count=1. If that maxes out whatever then man this router really stinks. But now nothing is getting maxed out. Its either our upstream provider or the fiber interface on the mikrotik.

You have to understand that initiating/sending/receiving TCP traffic is MUCH MUCH more demanding for the CPU than just routing that same traffic.

raz · Sun Sep 22, 2013 8:44 pm

Guys, after i played around with Downgrade to 6.1 and some queues, i found a way to fix it...

Interface ethernet full duplex = no and speed to 100Mbps. Its unbeliveable!

With Fullduplex = NO and 100 Mbps!

root@none:~# wget http://10.0.0.2/XenCenter.iso
--2013-09-22 19:39:32--  http://10.0.0.2/XenCenter.iso
Verbindungsaufbau zu 10.0.0.2:80... verbunden.
HTTP-Anforderung gesendet, warte auf Antwort... 200 OK
LÃ¤nge: 53635072 (51M) [application/octet-stream]
In Â»XenCenter.iso.4Â« speichern.

100%[======================================>] 53.635.072  6,78M/s   in 7,5s

2013-09-22 19:39:40 (6,80 MB/s) - Â»XenCenter.iso.4Â« gespeichert [53635072/53635072]

with ACTIVATED FullDUplex!

root@none:~# wget http://10.0.0.2/XenCenter.iso
--2013-09-22 19:40:46--  http://10.0.0.2/XenCenter.iso
Verbindungsaufbau zu 10.0.0.2:80... verbunden.
HTTP-Anforderung gesendet, warte auf Antwort... 200 OK
LÃ¤nge: 53635072 (51M) [application/octet-stream]
In Â»XenCenter.iso.6Â« speichern.

 1% [                                       ] 923.824     84,5K/s  ETA 9m 32s  ^C
root@none:~#

1 Gbps without FullDuplex wont work...

Test Setup:
Port 1: 10.0.0.1/24 (forced to 100 MBits, halfduplex)
Port 2+3 (LACP): 10.10.0.1/24 (Gbit, auto negotiated)

Wow, this is an heavy Bug @MikroTiK, now its time to answer my Ticket and pls fix this Bug.
[Ticket#2013081766000271] {Disarmed} Re: TCP slow in 6.2

jhaa · Sun Sep 22, 2013 10:16 pm

I'm wondering if there will be a ccr1036 version with 8xSFP 2xSFP+

manson · Tue Sep 24, 2013 9:29 am

What for do you need sfp+ in this crap? You want route 10 Gbps with CCR? Are you serious? I wonder if it can just forward so much bandwidth without any routing, firewall or queues. No chance to make any use with sfp+ in any routing purposes.

I have idea. Make old CCR to work good and then think about making new hardware. My CCR's are almost year old and they still are useless. Great thanks to Mikrotik for selling "beta" hardware with "beta" software as a "operator class router". Lol. You didn't see any real operator class router if you are willing to name this board like that.

lavv17 · Tue Sep 24, 2013 9:53 am

Lol. You didn't see any real operator class router if you are willing to name this board like that.

"real operator class router" costs 10-20 times more. For me, CCR delivers good enough performance. Especially considering it's cost.

lose · Tue Sep 24, 2013 11:44 am

hello . I have a question . What is the total number of concurrent users ( PPPOE ) in CCR1036 router?
If 4000 users simultaneously connected to router With PPPOE connection, the router will not be interrupted?

thank you.

dbodeenisudm · Tue Sep 24, 2013 10:56 pm

so pretty much the only way to get good speeds for TCP is to go 100m half? ouch.

please confirm.

dbodeenisudm · Tue Sep 24, 2013 10:57 pm

were gonna throw in a 2011 and see if that fixes it. we do not need any cpu power so we should be fine.

mhviper · Wed Sep 25, 2013 5:01 am

We are able to push around line rate (1gbit) through a ccr currently without any problems, don't know why he can't reach such result.

raz · Wed Sep 25, 2013 12:35 pm

so pretty much the only way to get good speeds for TCP is to go 100m half? ouch.

please confirm.

Please test it, if its the same Result like my Setup, we found a critical Bug, and we should report it to MikroTiK.
After i saw a lot of this Threads, its not a new Bug.

Kreacher · Wed Sep 25, 2013 5:16 pm

Hello,

What for do you need sfp+ in this crap?

Why this should be named crap? Because you are not having the need of SFP+?

You want route 10 Gbps with CCR? Are you serious?

It is in my opinion a better thing to have one 10 GBit/s uplink from the router
to the DMZ Switch and another one to the LAN Switch, even better then setting up LAGs (LACP).

I wonder if it can just forward so much bandwidth without any routing, firewall or queues.
No chance to make any use with sfp+ in any routing purposes.

Other Routers with SFP & LAG (LACP) capabilities are in the same price range.

I have idea. Make old CCR to work good and then think about making new hardware.

It is the same Hardware only with another port allocation.

Nuke · Wed Sep 25, 2013 11:32 pm

Lol. You didn't see any real operator class router if you are willing to name this board like that.
"real operator class router" costs 10-20 times more. For me, CCR delivers good enough performance. Especially considering it's cost.

I hear what you say, but I don't understand the problem. If you pay for a Toyota or Audi, you get what you pay for. The same goes if you buy a Tata, Chana or other whatever little car. Don't expect the same reliability, safety or performance. If you expect it, laughs are on you, don't care how its marketed by the manufacturer.

Carrier IMO mean things like 5 nines uptime, etc(think IOS-XR, where you can upgrade the software without rebooting). Mikrotik is good for budget links, but its not for a core network. I'll continue to use Mikrotik for outlying links, there is nothing in its class that beats it there.

I have used the CCR, one of our clients has one he brought himself. It works for him, but then again, it only does around 70mbps peak and does basic routing. It only crashed once(I think) and he is happy. So whatever floats your boat.

manson · Thu Sep 26, 2013 10:56 am

If CCR finally will work like it was intended then i'll say "it's cheap, it's fast, it's reliable". But many months past and i still don't know when or even if ever i'll use it because there is no official "road map" to make it usable. I've gave 1600 USD for "routers" that are beated by Raspberry PI in DNS forwarder performance.

But despite of that, CCR was sold to me with beta operating system as a "operator class router", they should be giving those for testing first, not making money on not tested hardware with buggy software. Sure i'm not stupid to place it in my core network, for routing purposes i have ASR 9006 for that. But i'm angry that Mikrotik is making fool of people. And i have to shine with my eyes in front of my boss, asking me why this hardware is making absolutly nothing. How many months or years i have to wait untill this flawless performance will be usable?

lavv17 · Thu Sep 26, 2013 1:18 pm

CCRs work here flawlessly for weeks serving around 900 pppoe users each. Surely there are problems with the software, but they are not global and happen only in certain configurations (e.g. I discovered a problem with PCQ and had to use dynamic simple queues).

Chupaka · Thu Sep 26, 2013 4:25 pm

We are able to push around line rate (1gbit) through a ccr currently without any problems, don't know why he can't reach such result.

what facilities of RouterOS do you use? any queues?

n3c8 · Thu Sep 26, 2013 5:29 pm

For all those complaining about the production grade of this router. This is still pre-production grade piece of kit. It's never been sold otherwise. If your distributor told you otherwise, take it up with them not Mikrotik who made this fact very clear.

If you look bad in front of your bosses or companies, its because you promised the world without checking your promises. Let's not forget that routers with this many cores at this price is new ground to even long standing established vendors. Juniper for example have had serious issues with the last 2 years of software.

Sooo, you bought a experimental piece of kit. You want to put pre-production in production, great. Have at it. You might get lucky, but its your ass.

PS: Can anyone confirm the /tool/profile cpu load of BGP?

manson · Fri Sep 27, 2013 1:34 pm

For all those complaining about the production grade of this router. This is still pre-production grade piece of kit. It's never been sold otherwise. If your distributor told you otherwise, take it up with them not Mikrotik who made this fact very clear.

If you look bad in front of your bosses or companies, its because you promised the world without checking your promises. Let's not forget that routers with this many cores at this price is new ground to even long standing established vendors. Juniper for example have had serious issues with the last 2 years of software.

Sooo, you bought a experimental piece of kit. You want to put pre-production in production, great. Have at it. You might get lucky, but its your ass.

PS: Can anyone confirm the /tool/profile cpu load of BGP?

there was no word in product specification about "pre-production hardware". It was sold as a fully functional hardware, no word even about beta software.

Kreacher · Fri Sep 27, 2013 4:54 pm

@manson

there was no word in product specification about "pre-production hardware". It was sold as a fully functional hardware, no word even about beta software.

I think it was! But that dosen´t matter at all I think, MikroTik was hiring over the last 2 years 20 new employees
so that we could imagine that the work on the CCR is going on and at one day we will have a shining device!
If you can´t wait or as you was mention by your self, preventing your ass away from your boss, I would sell
the CCR, so your boss have not a so high of cost mistakes made by yours! Perhaps for $1200 selling as used
CCR, but hey if in some month this will be a rocket machine don´t speak to me about this point please!

rpingar · Fri Sep 27, 2013 6:00 pm

an update about [Ticket#2013090666000521] vrrp backup problem - 6.x

all the ccr eth interfaces, in rare cases, get flapping all togheter.
this happen on the router where we have vrrp interface in backup mode.

thanks
Ros

Sat Sep 28, 2013 11:15 pm

an update about [Ticket#2013090666000521] vrrp backup problem - 6.x

all the ccr eth interfaces, in rare cases, get flapping all togheter.
this happen on the router where we have vrrp interface in backup mode.

thanks
Ros

So this is a current bug?

rpingar · Sun Sep 29, 2013 2:04 am

an update about [Ticket#2013090666000521] vrrp backup problem - 6.x

all the ccr eth interfaces, in rare cases, get flapping all togheter.
this happen on the router where we have vrrp interface in backup mode.

thanks
Ros
So this is a current bug?

yes i am asking about an update!

regards
Ros

infused · Sun Sep 29, 2013 4:49 am

6.3 is pretty solid on the CCR for us anyway. We have been using it in production since RC7, then RC13, then 6.3 (since December 30th 2012)

We only have one 100mbit uplink, about 20 queues, lots of firewall rules for about 10 different subnets.

StubArea51 · Sun Sep 29, 2013 5:13 pm

Has anyone used the CCR for hotspot deployments? I'm curious about max user counts and CPU load.

WirelessDSL · Mon Sep 30, 2013 2:35 pm

an update about [Ticket#2013090666000521] vrrp backup problem - 6.x

all the ccr eth interfaces, in rare cases, get flapping all togheter.
this happen on the router where we have vrrp interface in backup mode.

thanks
Ros
So this is a current bug?

Same here.
CCR with OSPF,MPLS,VPLS and TE.
RouterOS v6.4 all physical ports flap after 9 minutes. A few seconds later they come back and after another 9 minutes the same situation.
Made a ticket [Ticket#2013092466000274] about CCR Port flapping.

Installed RouterOS 6.5RC1 and this flapping doesn´t occur again up to now. But in the changelog for 6.5RC1 nothing is mentioned about this.

I hope this bug is resolved.

Attila · Tue Oct 01, 2013 12:31 am

Dear Normis.!

Multicast routing (PIM-SM) same source to different interface still doesn't work in CCR1036.

Ticket#2013020566000411

The device cannot be used for multicast routing since available.
So we can not use in IPTV environment.
Please fix this bug!

Sincerely,

Attila

drpater · Fri Oct 04, 2013 1:01 pm

Has anyone used the CCR for hotspot deployments? I'm curious about max user counts and CPU load.

Have test hotspot installation with 1036 as hotspot and 27 AP (SxT-G2HnD).
Work fine, 70-150 online users (more than 4k in radius database) with simple queues.
CPU usage less than 1-2%, peak traffic 30Mbit.

Wireless clients is mobile devices - handies, laptops, pads etc.

Seems cookie table cannot contain more than 4096 entries, but may be no.

ners · Mon Oct 07, 2013 12:18 pm

Can anyone tell if CCR1036-8G-2S+ will accept SFP modules in its SFP+ slots? I couldn't find any relevant info in the brochure.

PhilB · Tue Oct 08, 2013 4:02 pm

I remain astonished that Mikrotik are *still* pumping out more and more hardware revisions instead of making the kit they have work properly, but I guess otherwise they would have hardware guys sat around with nothing to do.

The SFP+ unit would definitely be interesting and disruptive.. if you could rely on it being useful. I'll check back in another year, again.

pcunite · Tue Oct 08, 2013 4:06 pm

I remain astonished that Mikrotik are *still* pumping out more and more hardware revisions instead of making the kit they have work properly, but I guess otherwise they would have hardware guys sitting around with nothing to do.

The SFP+ unit would definitely be interesting and disruptive.. if you could rely on it being useful. I'll check back in another year, again.

RouterOS v6.5 is taking a bit longer to get here and I'm hoping for a more formal release. They rushed the v6 series IMHO.

shayanjameel08 · Tue Oct 15, 2013 8:05 am

The board unit with 16 cores will happily run at those temperatures if installed in a waterproof case.

cse2012 · Fri Oct 18, 2013 11:34 am

balanser · Fri Oct 18, 2013 10:52 pm

would like to know the answer to a question about BGP:

So are we saying that BGP now runs on multiple cores. Is this from 6.3 production or a RC?

shayanjameel08 · Sat Oct 19, 2013 10:12 am

Got to say the price is awesome!

doush · Sat Oct 19, 2013 8:13 pm

CCR All ports FLAP s suddenly with no reason !

Ticket#2013101966000343

rpingar · Sun Oct 20, 2013 9:59 pm

CCR All ports FLAP s suddenly with no reason !

Ticket#2013101966000343

I opened a ticket more then one month ago witout any real answer [Ticket#2013090666000521]

I am still at 6.2, do you experience port flap also on 6.5???

If yes they are not yet fixed the issue.

regards
Ros

doush · Mon Oct 21, 2013 11:02 am

v6.3 in my case.
But no port flap bug fix in the changelog for 6.4 and 6.5 exists.

So I assume upgrading wont fix it.

struck · Tue Oct 22, 2013 10:11 pm

Hello

In my case I'm using the ports ether1 and ether6. After a few minutes of operation, the port ether1 link flaps. I went back to the old 1100AHx2 and the problem doesn't happen.

It's impossible to use.

Struck

an update about [Ticket#2013090666000521] vrrp backup problem - 6.x

all the ccr eth interfaces, in rare cases, get flapping all togheter.
this happen on the router where we have vrrp interface in backup mode.

thanks
Ros
So this is a current bug?

Same here.
CCR with OSPF,MPLS,VPLS and TE.
RouterOS v6.4 all physical ports flap after 9 minutes. A few seconds later they come back and after another 9 minutes the same situation.
Made a ticket [Ticket#2013092466000274] about CCR Port flapping.

Installed RouterOS 6.5RC1 and this flapping doesn´t occur again up to now. But in the changelog for 6.5RC1 nothing is mentioned about this.

I hope this bug is resolved.

Adolfo · Wed Oct 23, 2013 3:06 pm

CCR All ports FLAP s suddenly with no reason !

Ticket#2013101966000343

I opened a ticket more then one month ago witout any real answer [Ticket#2013090666000521]

I am still at 6.2, do you experience port flap also on 6.5???

If yes they are not yet fixed the issue.

regards
Ros

Hi I have the same problem with the port flap and ROS 6.5 didn´t solve the issue.
I opened a ticket too but without answer after I make a test with ROS 6.5. [Ticket#2013100766000642]

Wed Oct 23, 2013 3:25 pm

Can anyone tell if CCR1036-8G-2S+ will accept SFP modules in its SFP+ slots? I couldn't find any relevant info in the brochure.

Yes, it will also accept regular SFP modules.

rpingar · Wed Oct 23, 2013 3:33 pm

Can anyone tell if CCR1036-8G-2S+ will accept SFP modules in its SFP+ slots? I couldn't find any relevant info in the brochure.
Yes, it will also accept regular SFP modules.

Normis,

may you update all of us about port flap issue on CCR?????

regards
Ros

Wed Oct 23, 2013 3:34 pm

Can anyone tell if CCR1036-8G-2S+ will accept SFP modules in its SFP+ slots? I couldn't find any relevant info in the brochure.
Yes, it will also accept regular SFP modules.

Normis,

may you update all of us about port flap issue on CCR?????

regards
Ros

Do you mean the situation where a specific type of data triggers one time port up/down? You should try v6.6rc1

rpingar · Wed Oct 23, 2013 3:42 pm

Do you mean the situation where a specific type of data triggers one time port up/down? You should try v6.6rc1

Yes if in your case all ports goes down/up all toghether.

may you specify if it is the case?

regards
ros

Wed Oct 23, 2013 3:43 pm

Do you mean the situation where a specific type of data triggers one time port up/down? You should try v6.6rc1
Yes if in your case all ports goes down/up all toghether.

may you specify if it is the case?

regards
ros

yes, it should be fixed in v6.6rc1, you probably know the download link

Neilson · Thu Oct 24, 2013 8:56 am

@Normis

Wonderful to hear you think you have a fix for this.

Can you please detail some info about what data triggers the lockup? we got no feedback from our ticket nor any updates and we want to test in the lab before we return these to production. If you can confirm what data triggers it (we were unable to isolate any in particular) then we will generate that traffic to confirm we are freed from it.

Regards
Alexander

rpingar · Thu Oct 24, 2013 9:18 am

@Normis

Wonderful to hear you think you have a fix for this.

Can you please detail some info about what data triggers the lockup? we got no feedback from our ticket nor any updates and we want to test in the lab before we return these to production. If you can confirm what data triggers it (we were unable to isolate any in particular) then we will generate that traffic to confirm we are freed from it.

Regards
Alexander

+1

Regards
ros

doush · Thu Oct 24, 2013 1:35 pm

What kind of data triggers the ethernet flap ?
Normis ?

spoiler · Thu Oct 24, 2013 2:55 pm

We also had the flapping problem (only one eth was 'safe'). We solved it avoiding ethernet11 usage on a bridge (eoip<->eth) using sfp2 instead for that bridge (eoip<->sfp2). OSPF enabled in this ccr.

Also had another 'flapper' ccr which stopped magically flapping (or maybe due to MPLS package disabling) whick only used 3 eth devices. OSPF and BGP enabled on this ccr. No bridges.

raz · Thu Oct 24, 2013 4:40 pm

Normis,

what about my Ticket with TCP issues? #2013092666000332
and why im not getting any Answer from MikroTiK??????

Duduhandelman · Thu Oct 24, 2013 9:19 pm

Hi All.
I have CCR with 6.1 that had rebooted by watchdog.
What version should I upgrade? Flapping port 6.5

Thanks.

selric · Sat Oct 26, 2013 7:23 pm

Is the device still classed as a pre-production device and should not be used in production?

I want to run two of them with VRRP and each of them will have a full BGP-feed and in the future I would probably also add another transit to one of the routers.

Anyone doing this with the CCR?

How's the SFP+ device working?

Semir · Sun Oct 27, 2013 11:05 pm

Hi All.
I have CCR with 6.1 that had rebooted by watchdog.
What version should I upgrade? Flapping port 6.5

Thanks.

So Im not alone...
6.5 restarts,too.

rpingar · Mon Oct 28, 2013 7:58 am

Normis,

may you update all of us about port flap issue on CCR?????

regards
Ros

Do you mean the situation where a specific type of data triggers one time port up/down? You should try v6.6rc1

Hi Normis,
we tested this morning 6.6rc1 and the router started to reboot as soon as the traffic goes through (without making supout).
May be something about vrrp interfaces.

I sent the config that let the router to reboot as update about ticket#2013090666000521

please update when it is going to be fixed.
regards
Ros

Wed Oct 30, 2013 6:51 am

From our experience it seems the port flap issue is specific to the CCR1016 our 1036's seem to be fine.

Is this your experience ?

rpingar · Wed Oct 30, 2013 7:16 am

From our experience it seems the port flap issue is specific to the CCR1016 our 1036's seem to be fine.

Is this your experience ?

nop,
we have it on ccr1036.

From direct statement of Janis: it is related to a particular combination of cpu usage and traffic.

so it is more common on 1016, due to a easier high cpu usage, but not restricted to it.

regards
ros

ununium · Wed Oct 30, 2013 5:56 pm

I'm also experiencing the port flapping issue on the CCR1016.

In my case the traffic never goes over 60Mbps and the total cpu usage is never over 5%.

My setup consist of mangle rules, some filter rules and pcq's for traffic shaping. I'm using a bridge between two interfaces, one facing the lan and the other the wan.

Everything else which is not needed is disabled in the packages and the other services (SNMP, BT Server etc.) are disabled as well.

When I remove the CCR1016 from the network and connect to it directly the problem stops.

**Edit**

Tried 6.6rc1 which normis mentioned fixed the flapping, but the router crashes while booting. Back to 6.5 and waiting.

saaremaa · Wed Oct 30, 2013 10:18 pm

1. When I check the Bandwidth test between the two CCR-1016 is loaded only 1 core out of 16. This is a bug or feature? Router OS 6.5
2. The CPU temperature CCR-1016 with no load 55 * C. What is the maximum safe temperature for this processor?

Thu Oct 31, 2013 12:27 pm

Anybody with this port flapping on the CCR, can you try disabling the LCD screen, and see if anything changes at all ?

rpingar · Thu Oct 31, 2013 2:20 pm

Anybody with this port flapping on the CCR, can you try disabling the LCD screen, and see if anything changes at all ?

lcd disabled!

we will see

regards
Ros

saaremaa · Thu Oct 31, 2013 4:34 pm

fixed

wulfgard · Thu Oct 31, 2013 6:20 pm

Anybody with this port flapping on the CCR, can you try disabling the LCD screen, and see if anything changes at all ?

hello

we have this problem on x86 with v6.6rc1 28/10/2013 without package lcd

a+
Thierry

Fri Nov 01, 2013 4:01 pm

Anybody with this port flapping on the CCR, can you try disabling the LCD screen, and see if anything changes at all ?
hello

we have this problem on x86 with v6.6rc1 28/10/2013 without package lcd

a+
Thierry

This has nothing to do with x86, and only affects CCR. You probably have some unrelated issue. Email support with your supout.rif file.

Fri Nov 01, 2013 4:02 pm

We have fixed the CCR port "flapping" issue in latest v6.6rc1 build.

Please, (re-)download and (re-)install the latest build:
http://www.mikrotik.com/download/share/ ... le-6.6.npk
http://www.mikrotik.com/download/share/ ... es_6.6.zip

Changelog (where you also can see build time) is here:
http://www.mikrotik.com/download/share/ ... evnote.txt

Report back with your observation and even IF everything looks OK,
send us supout.rif file after a day or 2 of operation
so that we can check results of our fix.

ununium · Fri Nov 01, 2013 5:08 pm

We have fixed the CCR port "flapping" issue in latest v6.6rc1 build.

Please, (re-)download and (re-)install the latest build:
http://www.mikrotik.com/download/share/ ... le-6.6.npk
http://www.mikrotik.com/download/share/ ... es_6.6.zip

Changelog (where you also can see build time) is here:
http://www.mikrotik.com/download/share/ ... evnote.txt

Report back with your observation and even IF everything looks OK,
send us supout.rif file after a day or 2 of operation
so that we can check results of our fix.

normis,

Thank you for the update.

Yesterday I was trying to install 6.6 rc1 and it was crashing while booting. This only happened when the previous configuration was loaded. Once I reset the equipment to the defaults it booted fine, but again trying to load my config did crash the equipment.

I noticed it was as soon I loaded the PCQ tree config.

Could you please confirm if this problem was fixed?

I sent an e-mail of the console dump to my open ticket [#2013082866000107] .

wispwest · Fri Nov 01, 2013 8:34 pm

We are an ISP, and just finally switched off our OC3, so that now both of our upstream providers are 100% ethernet-based IP handoffs. Both 300mb. Question, our Sr Engineer that manages our core, says we can move to one of these "Cloud Core Router" by Mikrotik. We are currently using rack-mounted Cisco 7600 series. I know this new Mikrotik router will handle a lot more PPS, throughput, etc., but is it as reliable? Anyone been having issues with BGP or OSPF, or anything like that? I cannot risk ANYTHING not as reliable as our Cisco. I heard UBNT has something as well, but it looks like Mikrotik has been in the "router" game a lot longer, so I'm willing to maybe give it a shot.

Thanks for any feedback / advice.

Also, if there are any Mikrotik "experts" out there that do remote work, please contact me. I would like to know how to understand how to use this in our core database. Load-balancing and bonding are things we might be interested in.

manson · Sat Nov 02, 2013 2:30 am

Really, don't compare Cisco 7600 to CCR.

I don't want to say anything bad about Mikrotik but I had 7600 as a edge router with about bandwidth 2 Gbps usage with BGP via 5x1Gbit uplinks to internet. Almost no problems in many years. Now we have ASR 9006 as a edge router. 1,5 year with no problems. My two CCRs are from beginning till now hole plugs in rack, waiting for full multicore support in routeros. RouterOS controlled hardware is cheap and easy but really it is not for serious and bigger projects. No redundant hardware components, no redundant power supply. Only one thing that you can do to have any redundancy is to double your hardware and configuration.

pcarrillom · Sun Nov 03, 2013 4:34 am

Wispwest:

We run an ISP in Caracas Venezuela. We have CCR running flawlessly in NOTA (Nap of The Americas) in Miami, running BGP with Cogent, HE and Level3 (1Gig port each). No issues.
We also have multiple CCR running in many POP in Caracas.
For running multiple BGP feeds, CCR runs like a Champ. However, if getting full routes, route updates for full tables may be a little slow. That's the only issue we had found.
If you need remote assistance I may be able to help you.
You may contact me at pedroj_carrillo@yahoo.es

385esantic · Sun Nov 03, 2013 11:04 am

I have one CCR 1036-12G-4S and works fine until I turn on hotspot. But when I turn on hotsspot it crashes (after reboot log says "router was rebootet without proper shutdown,propably kernel failure") it reboots sometimes 10-15 minutes ahter turning on hotspot,sometimes work up to 2 hours... tried all avaible versions... but other router with same config works fine...
What should i do?

OlegM · Sun Nov 03, 2013 1:15 pm

After the upgrade to version 6.6 from 01.11.2013, constantly reboots after the connection of users PPTP. Disabled access to the CCR for users to reboot stopped. But now I can not roll back the firmware. It just does not roll back.

ununium · Sun Nov 03, 2013 4:29 pm

After the upgrade to version 6.6 from 01.11.2013, constantly reboots after the connection of users PPTP. Disabled access to the CCR for users to reboot stopped. But now I can not roll back the firmware. It just does not roll back.

Try setting the router to factory defaults and try the downgrade. That's how I had to do it.

Begetan · Mon Nov 04, 2013 12:02 am

Wispwest:
For running multiple BGP feeds, CCR runs like a Champ. However, if getting full routes, route updates for full tables may be a little slow. That's the only issue we had found.

Could you tell us please how many full BGP tables you have and how long you update full table? For RC software it was about 3-4 minute for 1-2 full BGP tables.

rpingar · Mon Nov 04, 2013 7:31 am

Anybody with this port flapping on the CCR, can you try disabling the LCD screen, and see if anything changes at all ?

I disabled LCD, and after 3 days I had only one flap port issue.
Only one porte went donw and then up in 3 secs.

Before I had all the ports going down/up at the same time.

Running 6.5

The latest 6.6 or 6.6rc1 is not usable because the router reboots as soon as it get traffic

regards
Ros

saaremaa · Mon Nov 04, 2013 10:07 am

We use the official version 6.5 for CCR-1016-12 (test bench) for OSPF and MPLS / VPLS. Problems with falling ports and reboots are not experiencing.Now test MPLS cloud loaded 1.5 Gbps aggregate bandwidth.

infused · Wed Nov 06, 2013 11:19 am

Anyone having drop outs on the CCR after a month or so of up time? We are... reboot fixes it instantly.

cybercoder · Wed Nov 06, 2013 12:42 pm

I saw this model :CCR1036-12G-4S-EM in routerboard.com

The question is, can i use it to 1 SFP traffic in and 3 stm-4 traffic out? and is this device stable enough to replace a CISCO Catalyst switch with that?
Isn't any SFP+ port support model of CCR?
Regards

Wed Nov 06, 2013 3:17 pm

v6.6 todays build should fix any remaining port reset issues. please re-download to apply the latest fix.

doush · Wed Nov 06, 2013 4:59 pm

Normis;
Latest RC1 put our CCR into a loop and it is a known issue I guess. Is it fixed in this build ?? Can we safely upgrade ?

rpingar · Wed Nov 06, 2013 5:08 pm

v6.6 todays build should fix any remaining port reset issues. please re-download to apply the latest fix.

tonight will test in production to see if it will not continuosly reboot like the rc1s.

I will update you Tomorrow morning.

regards
ros

mje · Wed Nov 06, 2013 7:34 pm

Where can I get a copy of 6.6 (Fixed)?
I'm suffering from ports that keep resetting.
Loosing my hair.

rpingar · Thu Nov 07, 2013 8:02 am

v6.6 todays build should fix any remaining port reset issues. please re-download to apply the latest fix.
tonight will test in production to see if it will not continuosly reboot like the rc1s.

I will update you Tomorrow morning.

regards
ros

6.6 is in production right now. No reboots on boot.
Il will update in a coupple of days about port flap issue.

regards
ros

janisk · Thu Nov 07, 2013 8:36 am

Normis;
Latest RC1 put our CCR into a loop and it is a known issue I guess. Is it fixed in this build ?? Can we safely upgrade ?

yes, problem was found, latest builds available for evaluations already have this fixed.

mje · Thu Nov 07, 2013 11:59 pm

6.6 has not solved my problems.

I even removed auto-detection and set the copper speed to 100Mb - which is all my Cisco can do at the moment.
The Optic seemed stable but the bouncing copper killed that - then the CCR started rebooting....
Left just the Optic plugged in - and now at home - I find it dead - so the CCR is now totally unreachable.

I get the feeling that no one at Mikrotik has tested a CCR1036 in a live enviroment - with live, full BGP feeds.
If anyone from Mikrotik wants to pop over to London - I can provide the experience. I am currently scheduled
to fly back to South Africa on Monday evening though.

Fri Nov 08, 2013 1:17 am

6.6 still has the problems with 100mbit/FDX ports randomly changing to gigabit.

It seems like some sort of database mismatch. Winbox shows the static speed/duplex as set, but export does not. See my last ticket updates.

How do Mikrotik get this soooo wrong !

We also have issues with SFP in CCR linking to certain devices on autonegotiation (lightpointe)

Fri Nov 08, 2013 11:43 am

6.6 still has the problems with 100mbit/FDX ports randomly changing to gigabit.

It seems like some sort of database mismatch. Winbox shows the static speed/duplex as set, but export does not. See my last ticket updates.

How do Mikrotik get this soooo wrong !

We also have issues with SFP in CCR linking to certain devices on autonegotiation (lightpointe)

Please send mail to support regarding both of these problems, preferably in separate tickets.

Fri Nov 08, 2013 12:50 pm

6.6 still has the problems with 100mbit/FDX ports randomly changing to gigabit.

It seems like some sort of database mismatch. Winbox shows the static speed/duplex as set, but export does not. See my last ticket updates.

How do Mikrotik get this soooo wrong !

We also have issues with SFP in CCR linking to certain devices on autonegotiation (lightpointe)

does it happen with a specific port, or with any port? does it happen if you plug into another port?

lavv17 · Mon Nov 11, 2013 12:07 pm

Upgrade 6.5 to 6.6 on a CCR leaves bonding interface with MAC address 00:00:00:00:00:00. Fixed by interface recreation.

Tue Nov 12, 2013 12:52 am

6.6 still has the problems with 100mbit/FDX ports randomly changing to gigabit.

It seems like some sort of database mismatch. Winbox shows the static speed/duplex as set, but export does not. See my last ticket updates.

How do Mikrotik get this soooo wrong !

We also have issues with SFP in CCR linking to certain devices on autonegotiation (lightpointe)
does it happen with a specific port, or with any port? does it happen if you plug into another port?

Hi Normis,

What we have observed is that ether1 on the CCR1016 will not link up to Cisco at 100FDX/NoAuto at all. We moved the connection to ether4 which works, but drops intermittently.

We cannot do any more testing on the units we are having the problem on as they are production networks, but I have a spare CCR and Cisco 1841 here which I will set up later today and give your team remote access to.

Feel free to contact me via Ticket#2013102666000026 if it is more efficient for you.

gigahz · Thu Nov 14, 2013 5:25 pm

From our experience the CCR isn't ready for hotspot use. The traffic on one of our sites had a negative trend where the bandwidth passed was lower when more users were online. This was contrary to what we experience with our X86 box as well as being contrary to logic.

I'm having similar issues using CCR1016-12G v6.0rc9 with hotspot and PPPoE server enabled. The CCR is connected to 100 Mbps FD fiber. Speedtests to local speedtest servers (within 200 miles and peered with my WAN provider) are consistently 92 Mbps, but speedtests to out-of-region sites are showing download of 6-12 Mbps, and upload of 1-3 Mbps. It seems that as latency to/from the remote site increases the throughput decreases. I'm seeing speed degradation of only ~10% when hotspot and PPPoE are turned off.

Is anyone else seeing this issue, and has it been resolved in latest v6.6? I've been reluctant to upgrade because newer releases don't seem to be getting good reviews.

I appreciate help and advise from any Mikrotik experts.

armandfumal · Thu Nov 14, 2013 6:17 pm

Wispwest:
For running multiple BGP feeds, CCR runs like a Champ. However, if getting full routes, route updates for full tables may be a little slow. That's the only issue we had found.
Could you tell us please how many full BGP tables you have and how long you update full table? For RC software it was about 3-4 minute for 1-2 full BGP tables.

My CCR with V6.5 load 467000 prefixes on 2 full table in less than 2 mins...and 14800 IPV6 prefixe...

cse2012 · Sat Nov 16, 2013 5:18 am

i wonder the ccr1036 can use 155mb/s gbic.
or it is not possible, the sfp module of the ccr1036 can only use gigabit gbic?

erkel · Sat Nov 16, 2013 2:45 pm

We are an ISP, and just finally switched off our OC3, so that now both of our upstream providers are 100% ethernet-based IP handoffs. Both 300mb. Question, our Sr Engineer that manages our core, says we can move to one of these "Cloud Core Router" by Mikrotik. We are currently using rack-mounted Cisco 7600 series. I know this new Mikrotik router will handle a lot more PPS, throughput, etc., but is it as reliable? Anyone been having issues with BGP or OSPF, or anything like that? I cannot risk ANYTHING not as reliable as our Cisco. I heard UBNT has something as well, but it looks like Mikrotik has been in the "router" game a lot longer, so I'm willing to maybe give it a shot.

Thanks for any feedback / advice.

Also, if there are any Mikrotik "experts" out there that do remote work, please contact me. I would like to know how to understand how to use this in our core database. Load-balancing and bonding are things we might be interested in.

Look at the Juniper MX series perhaps.

gigahz · Sat Nov 16, 2013 4:35 pm

Upgrading to v6.6 has not resolved our CCR1016 problems. Hotspot and PPPoE connections are still extremely slow. I've tried adjusting MTU, MRU, and turning MSS on/off with no noticeable improvement. Changing LAN port speed to 100 M half-duplex fixes the speed issue, but the ethernet becomes very unstable and locks up under continuous heavy traffic. Setting LAN port speed to 100 M half-duplex was only to test, and not something we would leave in place long term. It was just a way to isolate and test.

When Hotspot and PPPoE are turned off I can set LAN port speed to 1000 M full-duplex and achieve consistent Internet speeds to any LAN client (testing Internet to speeds of 100 Mbps up/down). So turning off Hotspot & PPPoE fixes the problem, but turning off just one or the other does not resolve the issue. I can live without PPPoE, but I need Hotspot.

I'd like to know if anyone else has seen these issues, and if so has anyone found a fix, or is this a firmware issue that Mikrotik needs to resolve?

I'd also like to understand why speed (throughput) was dramatically increased when LAN port speed was dialed down to 100 M half-duplex.

iuli4u2003 · Sat Nov 16, 2013 7:39 pm

Does anyone know how many switch-chips are CCR1016-12G BU?
somehow ONLY has a chip swich?

how many groups can group switch 12 gigabit ports?

Can you give me a link with information about these issues (switch-chip group ports at CCR1016-12G-BU)?

thanks

onnoossendrijver · Sun Nov 17, 2013 3:29 am

There are no switch chips in the CCR.

iuli4u2003 · Sun Nov 17, 2013 9:10 am

There are no switch chips in the CCR.

so can not be grouped in groups of switch ports?

onnoossendrijver · Sun Nov 17, 2013 11:01 pm

You can, with the bridge function in the software. This should be (near) wire-speed on the CCR.

ariffin · Tue Nov 19, 2013 4:37 pm

Hi,

I wanna ask if i want to do VPN with cloud router, how many connection this router can support?

keepcalm · Thu Nov 21, 2013 1:00 pm

Upgrading to v6.6 has not resolved our CCR1016 problems. Hotspot and PPPoE connections are still extremely slow. I've tried adjusting MTU, MRU, and turning MSS on/off with no noticeable improvement. Changing LAN port speed to 100 M half-duplex fixes the speed issue, but the ethernet becomes very unstable and locks up under continuous heavy traffic. Setting LAN port speed to 100 M half-duplex was only to test, and not something we would leave in place long term. It was just a way to isolate and test.

When Hotspot and PPPoE are turned off I can set LAN port speed to 1000 M full-duplex and achieve consistent Internet speeds to any LAN client (testing Internet to speeds of 100 Mbps up/down). So turning off Hotspot & PPPoE fixes the problem, but turning off just one or the other does not resolve the issue. I can live without PPPoE, but I need Hotspot.

I'd like to know if anyone else has seen these issues, and if so has anyone found a fix, or is this a firmware issue that Mikrotik needs to resolve?

I'd also like to understand why speed (throughput) was dramatically increased when LAN port speed was dialed down to 100 M half-duplex.

I have the same problem with my CCR 1036-12G-4s ROS 6.5
I use 3 WAN with different public IP, each with link of 100mb with my provider.
I use mangle to mark routing to the 3 WAN.
The clients connect via pppoe with various profiles (PPP profiles with Rate Limit), some 512k/7M, another 2M/10M, another 12M/16M.
All pppoe reach the CCR through varius EoIP tunnels.
What happens is that ALL connections, both pppoe and pptp, regardless of the profile, go to download between 3 and 4 mega, not more, with a profile without rate limit, and then out of queues, it go at 25x25 (downloadXupload)
When there is increased demand for bandwidth, clients go to download between 2 and 3 mega and with a profile without rate limit, it go at 5x5.
Traffic on the single WAN never exceeds 50 mbps.
Tha active clients are 800.
I guess there is a problem with queue management, but do not know what.

Someone can help me ????

Thu Nov 21, 2013 3:30 pm

Replaced a RB2011UAS-RM which suffered from doing 150 PPPOE Connections with a CCR1036.
Great! Nearly no CPU usage now.
Problem was the CPU usage for the Mangling rules to adapt MSS.

n3c8 · Thu Nov 21, 2013 4:34 pm

20Gbps.png

No whinging for once. Got one of the new 10Gbps MKs on the lab... Decided to chuck 20Gbps through it on a bonded 2 x 10Gbps interface.

ITS VERY LOUD NOW!!!

Fun to see it actually managing to chuck the traffic out the interface though. Would be nice to think it could do this in production but I doubt it.

Still promising results non the less.

Chupaka · Fri Nov 22, 2013 4:00 pm

Would be nice to think it could do this in production but I doubt it.

Still promising results non the less.

I'm currently playing with that toy too. in routing, it rulez. as soon as you add queues, it shows much lesser results

also, you don't have Rx on interfaces - loop those sfp+ interfaces and then create just one queue in queue tree with parent=global (the queue won't catch ANY traffic, but all incoming packets will go through only one core of 36

)

Fri Nov 22, 2013 4:31 pm

Would be nice to think it could do this in production but I doubt it.

Still promising results non the less.
I'm currently playing with that toy too. in routing, it rulez. as soon as you add queues, it shows much lesser results

also, you don't have Rx on interfaces - loop those sfp+ interfaces and then create just one queue in queue tree with parent=global (the queue won't catch ANY traffic, but all incoming packets will go through only one core of 36 )

Yes. You've to be carefully what features to use and be careful with updates.
We've 3 CCR1036 in production now. Doing Routing/Bridging/OSPF/PPPOE/Vlan/Basic filtering with ROS 6.5.
They work great so far. Blazing fast and no Problems.

SpaceCommander · Tue Nov 26, 2013 1:05 am

CCR1036-12G-4S

Been fighting it most of today, one customer getting slammed with a DoS Packet attack, 150,000-250,000 pps, its enough to cause this CCR to overload to 100% cpu! When the attack stops the router drops to 5-10% cpu, when the attack starts back up, back to 100% cpu. This is on 6.0rc7, 6.1 Release, 6.2rc1, tried all 3.

Absolutely same situation with the same CCR model and ROS 6.6.
Firewall contains lot of DNAT rules (port forwardings to internal services), tarpit, mark, allow/drop, address lists and so on.
But when CPU freq was increased from 600 to 1200 MHz, CPU usage was immediately reduced to 3-5%!
After that, usage is periodically increased up to 60-70%, and "/tool profile" says that firewall eats most of them.

tenenbaum · Tue Nov 26, 2013 7:36 am

Using firmware 6.6 on a CC and it's still flapping.
We disabled LCD and SNMP and it's a bit better.
What we noticed is that these conditions will cause flaps:

1. Over 20,000 pps per interface
2. Over 300mbit per interface
3. SNMP
4. OSPF changes

It's almost like there is a capacity limit that is either reached due to some storm or major ops volume.. by splitting one of our back-hauls to 2 x 250mbit we are avoiding most of the flaps.

Given this situation and complaints by customers, we are planning to switch the x86 platform. We like our CC routers but can't use them right now.

Tue Nov 26, 2013 8:29 am

CCR1036-12G-4S

Been fighting it most of today, one customer getting slammed with a DoS Packet attack, 150,000-250,000 pps, its enough to cause this CCR to overload to 100% cpu! When the attack stops the router drops to 5-10% cpu, when the attack starts back up, back to 100% cpu. This is on 6.0rc7, 6.1 Release, 6.2rc1, tried all 3.
Absolutely same situation with the same CCR model and ROS 6.6.
Firewall contains lot of DNAT rules (port forwardings to internal services), tarpit, mark, allow/drop, address lists and so on.
But when CPU freq was increased from 600 to 1200 MHz, CPU usage was immediately reduced to 3-5%!
After that, usage is periodically increased up to 60-70%, and "/tool profile" says that firewall eats most of them.

When under significant DoS attack your CPU load will go up in any way - question is how much your traffic can still get trough the device, I was living under DDoS for several weeks while my higher tier ISP was able to block it. CPU load was almost 100% all the time, but clients were still able to communicate through the device.

honzam · Tue Nov 26, 2013 4:02 pm

Using firmware 6.6 on a CC and it's still flapping.

Have you access for prerelease?
What's new in 6.7 (2013-Nov-26 09:57)
*) disabling/enabling ethernet ports did not work properly on CCRs - could cause port flapping;