ROS 6.9
What about effective distribution of the load between the processor cores?

just 80 Mbit/s L2TP, NAT and Queu tree with 20 rules server on CCR1036-12G-4S

Queue Tree is NOT Multi-thread/Multi-Core optimized yet so performance will be limited to a single core(tile) on the CCR routers. This is likely what you are experiencing.

Mikrotik have indicated this is the next thing that will get optimized once PPP optimizations are finished.

ROS 6.9
What about effective distribution of the load between the processor cores?

just 80 Mbit/s L2TP, NAT and Queu tree with 20 rules server on CCR1036-12G-4S

Queue Tree is NOT Multi-thread/Multi-Core optimized yet so performance will be limited to a single core(tile) on the CCR routers. This is likely what you are experiencing.

Mikrotik have indicated this is the next thing that will get optimized once PPP optimizations are finished.

Yes, Queu tree slows all perfomance of router - when i disable it, bandwidth increase dramaticaly.

what advise for optimise Queu tree?

It is evident that the majority of processor cores again do nothing while three cores perform the work. Question - for what are we paying money?

It is evident that the majority of processor cores again do nothing while three cores perform the work. Question - for what are we paying money?

Mikrotik have not hidden the fact that Queue Trees are not yet optimized on CCR. I am sure they will fix this as soon as they have resource.

CCR1036-12G-4S-EM, default config, after upgrade to 6.9, yellow LED's near empty SFP slots blinks all together in different intervals like sending morse code. Is that some new feature?
"system leds disable [find]" shut them off, but this is not resolution.
"system leds enable [find]" start all over again ...

And why those SFP LED's must be that bright? Every time, I look directly to CCR, I am blinded.

/system leds
set 0 disabled=no interface=sfp1 leds=eth0-led type=interface-activity
set 1 disabled=no interface=sfp2 leds=eth1-led type=interface-activity
set 2 disabled=no interface=sfp3 leds=eth2-led type=interface-activity
set 3 disabled=no interface=sfp4 leds=eth3-led type=interface-activity

You can fix this problem by applying configuration example from container below. This problem occurred due to rare bug that corrupted led configuration, we already made changes in ROS to prevent such thing happening in future release versions.

Code: Select all

/system leds
set 0 disabled=no interface=sfp1 leds=eth0-led type=interface-activity
set 1 disabled=no interface=sfp2 leds=eth1-led type=interface-activity
set 2 disabled=no interface=sfp3 leds=eth2-led type=interface-activity
set 3 disabled=no interface=sfp4 leds=eth3-led type=interface-activity

@netmaster
whats your firmware version? mine solved by upgrading firmware to 3.10

CCR 10-16 ver 6.9 all 100% Well done Mikrotik.

Currently CCR1036 and CCR1016, both ROS v.6.9, so far so good.

Regards,

Hi! CCR1036-12G-4S os. ver 6.7

I have problem with Serial port. When I want to set baud-rate=2400 i get message: "failure: specified port speed is not supported on this port"

Hi MT

We have updated or CCR to 6.9 and we are still seeing port flapping, any changes to a Ethernet interface will cause the router to crash and only a hard power off and on will bring it back.

Also there is still a issue with Ethernet TCP through put, we still can not use our 100mb leased line to the max due to this issue.

Adrian

what advise for optimise Queu tree?

Could you help me with SSL offload? I plans install http reverse prosy for SSL offload (I know, Tilera CPU supports it on very big speeds) on CCR with Router OS.

The fact that the traffic on my CCR1036 increased to 700-800 Mbps, and the packet loss began. I gave a detailed description of the scheme of testing.
At the same RB800 performance testing UDP packets of 64-750 bytes is much higher than the CCR

Hello!
Could you help me with SSL offload? I plans install http reverse prosy for SSL offload (I know, Tilera CPU supports it on very big speeds) on CCR with Router OS.

This is an interesting question. The Tilera CPU supports key-exchange (RSA, etc.) acceleration, which is the most expensive part of a SSL handshake

- the actual encryption of the data itself after key exchange has been done is not that CPU intensive anymore. But if you have a high session rate and lots of threads (CPUs), this could speed up things.

As far as I know, RouterOS doesn't support such functions, but - if it would be possible to boot some kind of plain Linux on the box and if a engine plugin for OpenSSL were available, you cold theoretically do such things.

Question: Has anyone already booted a plain Linux on a CCR and if yes - how?

I think that with stud, varnish and haproxy the CCR would surely make a decent HTTP load balancer and accelerator.

Could you help me with SSL offload? I plans install http reverse prosy for SSL offload (I know, Tilera CPU supports it on very big speeds) on CCR with Router OS.

This is an interesting question. The Tilera CPU supports key-exchange (RSA, etc.) acceleration, which is the most expensive part of a SSL handshake - the actual encryption of the data itself after key exchange has been done is not that CPU intensive anymore. But if you have a high session rate and lots of threads (CPUs), this could speed up things.

As far as I know, RouterOS doesn't support such functions, but - if it would be possible to boot some kind of plain Linux on the box and if a engine plugin for OpenSSL were available, you cold theoretically do such things.

Question: Has anyone already booted a plain Linux on a CCR and if yes - how? I think that with stud, varnish and haproxy the CCR would surely make a decent HTTP load balancer and accelerator.

This is not true, sorry! The Tilera platform used by MikroTik to build the CCR series is owning one or two units if the high speed encryption engine so called "MiCA" (at this point I want say tank you MikroTik for choosing this variant) and this MiCA is taking any data out directly from the RAM for doing the encrypting or decrypting work that means the entire or whole crypto work is done outside of the Tilera many Core CPU!!! But, that we are all able to use this as a function or option or so called feature, MikroTik must write code that is using this MiCA encryption engine for taking the entire load from the CPU, but if this is done the CCR series could be also used as a fine VPN concentrator as well.

Why booting a second system or pre boot system? And by the way the Linux must be compiled for the Tilera platform it selfs otherwise the code will not be executed on this machine!
[...]
No one is able to do so, or your are owning a Tilera SDK and you have an account to the over 2500 pre compiled code packets (.rpm) from the Tilera company and also perhaps a compiler that matches the Tilera platform.

I think that with stud, varnish and haproxy the CCR would surely make a decent HTTP load balancer and accelerator.

Then I wish for my self and all others an extra LAN port either SFP, SFP+ or copper based for setting up two or more routers to build a router cluster with ARP balance over CARP so that the whole load will be balanced over many routers, but this depends on the own wishes mostly.

And - AFAIK - RouterOS utilizes the engine since v6.8 (don't know if full or only in parts), it is mentioned somewhere in the changelog.

Ok, if there are only one or two engines per CPU and not one per core, that's a point.

Just got on my CCR1036 stopped working vlan. Stopped going traffic on some Vlan. Helped rebut. ROS v6.10

i cannot use 6.10 have reboots every some days. max uptime was 4 days. i went back to 6.7.

both firmwares have a memleak when creating support files. i tried to create a support-file every 10min to have one in case of an error. you can see the mem going down rapidly.

after a conversation with support i upgraded to 6.10 which is unstable. now i am down to 6.7.

can someone tell us here which version of routeros für ccr is stable to use in production environment.
i need ipsec, sstp und l2tp

Looking to understand the impact of the Firewall on high speed routing on a CCR. The benchmark's refer to 25 rules, I assume they are "passing" rules rather than blocks? Ie the traffic is being checked 25 times in a row and passing before being routed on to the end locations.

1. What impact would rule 1 being "Drop All UDP" have on the performance of the forwarding plane. Minimal? None?
2. If the 25 rules are going from most generic to most specific on DROP's and then the default is accept what would I expect to see from a impact view on the unit?

I know these are really loaded questions, just trying to understand based on the data on the web page and real world, really how the tests were completed and what an inverted usage of the firewall would have?

As you can probably imagine, this is all around DDoS and keeping the firewalls alive, especially when they have 10G interfaces.

any reason 6.7 and not 6.9 ?

I am having performance issues with 6.10...

Looking to understand the impact of the Firewall on high speed routing on a CCR. The benchmark's refer to 25 rules, I assume they are "passing" rules rather than blocks? Ie the traffic is being checked 25 times in a row and passing before being routed on to the end locations.

1. What impact would rule 1 being "Drop All UDP" have on the performance of the forwarding plane. Minimal? None?
2. If the 25 rules are going from most generic to most specific on DROP's and then the default is accept what would I expect to see from a impact view on the unit?

I know these are really loaded questions, just trying to understand based on the data on the web page and real world, really how the tests were completed and what an inverted usage of the firewall would have?

As you can probably imagine, this is all around DDoS and keeping the firewalls alive, especially when they have 10G interfaces.

25 rule each did process the packets. In normal conditions this would not happen, and actual packets would pass only few of them and packets from established and related connections just 2 to 4 rules. Adding 1 or removing 1 rule from that list would not change that much.

Also, dropping a packets does not cost (as in CPU resources) that much.

25 rule each did process the packets. In normal conditions this would not happen

Does anybody use CCR for NAT?
I need about 1G of NAT.

Does anybody use CCR for NAT?
I need about 1G of NAT.

And the CCR1036 is not able to delivery you that throughput?
Ok for SPI & NAT you should shorten the entire throughput
for something around 3% - 5% in normal cases.

I want to connect CCR CCR1036-12G-4S-EM with Cisco switch SG500 series via multimode fiber optic on SFP transceiver.
Which is better to use :
1. Both side with Mikrotik SFP module S-85DLC05D
2. Both side with Cisco SFP module MGBSX1
3. On CCR side using Mikrotik SFP module S-85DLC05D & on Cisco side using Cisco SFP module MGBSX1

Does anybody use CCR for NAT?
I need about 1G of NAT.

And the CCR1036 is not able to delivery you that throughput?
Ok for SPI & NAT you should shorten the entire throughput
for something around 3% - 5% in normal cases.

I am using x86 server with ROS 6.10 only for nat. Utilisation is 5% for 300M,
but sometimes this machine got occasionaly reboot. Support didn't find eny problem in config.

It's cheaper to buy new CCR than ty to find compatible PC box.

I just wonder if enybody running NAT on CCR at least 3 month without crash o reboot.

Does anybody use CCR for NAT?

I need about 1G of NAT.

Begetan, please email support with these details and supout.rif file. Thanks!

Hello,
I am interested in status of IPSec hardware acceleration. In the beginning of this topic you have mentioned that this will be available soon.
Today I have tested to encrypt traffic between two CCR1036 and the best result was about 300Mbps.

Regards,

ners, thank you for reply!

Could you provide please some more details:

1. What is maximum utime?
2. Current version of ROS.
3. Does NAT utilise multi cores or single core?

# CPU LOAD IRQ DISK
0 cpu0 7% 7% 0%
1 cpu1 5% 5% 0%
2 cpu2 4% 4% 0%
3 cpu3 6% 6% 0%
4 cpu4 7% 7% 0%
5 cpu5 4% 4% 0%
6 cpu6 3% 3% 0%
7 cpu7 3% 3% 0%
8 cpu8 4% 4% 0%
9 cpu9 8% 8% 0%
10 cpu10 9% 9% 0%
11 cpu11 7% 7% 0%
12 cpu12 7% 7% 0%
13 cpu13 9% 9% 0%
14 cpu14 5% 5% 0%
15 cpu15 4% 4% 0%
16 cpu16 6% 6% 0%
17 cpu17 8% 8% 0%
18 cpu18 7% 7% 0%
19 cpu19 8% 7% 0%
20 cpu20 7% 7% 0%
21 cpu21 4% 4% 0%
22 cpu22 4% 4% 0%
23 cpu23 5% 5% 0%
24 cpu24 8% 8% 0%
25 cpu25 3% 3% 0%
26 cpu26 4% 4% 0%
27 cpu27 9% 9% 0%
28 cpu28 9% 9% 0%
29 cpu29 1% 1% 0%
30 cpu30 9% 9% 0%
31 cpu31 7% 7% 0%
32 cpu32 11% 11% 0%
33 cpu33 9% 9% 0%
34 cpu34 4% 4% 0%
35 cpu35 5% 5% 0%

ners, I think this is useful information for all. Please update information after 1 month of stable work.

I had previously RB1100AHx2 with uptime of 300 days running on ROS 5.18. Hope CCR should be the same.

ners, I think this is useful information for all. Please update information after 1 month of stable work.

I had previously RB1100AHx2 with uptime of 300 days running on ROS 5.18. Hope CCR should be the same.

I have CCR1036-12G-4S. Doing 81 NATs over 19 EoIP tunnels and 62 VLANs. Total bandwidth it currently serves is a bit below 700Mbps, sometimes peaking 750Mbps, but thats not CCRs limitation, its mine. Load rarely exceeds 30%. Uptime over 2 months. Rebooted recently, because added one package. v6.5 Running more or less stable.

not if using vfr's. Router hang after puting an Interface to a vrf.

all after 6.5 is best. I have seen no big problem to the latest 6.8-9-10. 10 the best couse of partly fixes cisco non gigabit issue

My CCR1036 suddenly erupts this morning, I don't know why..

anyone tried 6.11 on ccr? i am still stuck at 6.7 because 6.9 and 6.10 run unstable. max uptime was about 4 days.

wow! I would take a guess there was a short on the board between that component and the ground. Depending on the cause. if manufacture fault then yes. if foreign object then no

My CCR1036 suddenly erupts this morning, I don't know why..

installed new firmware?

Posted in the other thread, I am on 6.3, which is mostly stable. Am I right in saying 6.7 is the most stable now at the moment?

I'm having stability issues with my CCR (CCR1016-12G) after upgrading from ROS 6.5 to 6.10 and also 6.11. The CCR will run for a few days (as little as 2 days and as long as 7 days) before it loses connectivity. I can't ping or access via Winbox. Sometimes the touchscreen and serial console are unresponsive as well. Only solution is to unplug power and plug back in. Sometimes (though less frequent) I lose the connectivity, but touchscreen and serial console work.

I noticed today when this happened and I still had serial connectivity that

Code: Select all

/system routerboard print

showed the firmware version was 3.10 with version 3.12 available. I updated the firmware to 3.12 and rebooted. Here's to hoping this solves the issue. If not, I'm going to revert back to 6.5 or possibly 6.7. Does anyone know where I can download older versions of ROS?

we are going to put 2 CCR into production with 6.11 code

primary use bonding, vlan, vrrp and bgpd

bad move?

we are going to put 2 CCR into production with 6.11 code

primary use bonding, vlan, vrrp and bgpd

bad move?

uptime: 5w5d20h25m43s
version: 6.9
build-time: Jan/31/2014 11:18:19
free-memory: 3483.3MiB
total-memory: 3969.0MiB
cpu: tilegx
cpu-count: 36
cpu-frequency: 1200MHz
cpu-load: 15%
free-hdd-space: 903.6MiB
total-hdd-space: 1024.0MiB
architecture-name: tile
board-name: CCR1036-12G-4S
platform: MikroTik

I have a CCR1036-12G-4S with the following running on it;

hotspot
PCC load balancing
2 Nat rules

I am seeing slow load times on webpages and I suspect a possible issue with DNS Caching. I am on v6.1 with firmware version 3.09. All packages installed are running v6.1 including user-manager.

Should I update to a newer version to possibly resolve the dns issue? If so what version seems to be the most stable, I see a lot of 6.5's possibly?

I have several CCR's in production now, all of them are running similar setups and seeing similar performance. all of them are on v6.1

maybe SRPF impact. try disable RP filtering if used(usual/reasonal for bras/cpe only), temporally.

Does anybody use CCR for NAT?

I need about 1G of NAT.

1Gb/sec NAT is fine for 1036.

Very useful information about CCR is available here:
http://mum.mikrotik.com/presentations/RU14/megis.pdf

Very useful information about CCR is available here:
http://mum.mikrotik.com/presentations/RU14/megis.pdf

Thank you. This is fantastic stuff. Where can we find more of this sort of thing? I highly suggest this is more visible to people like us browsing the forums. It actually answered a lot of my questions.

Very useful information about CCR is available here:
http://mum.mikrotik.com/presentations/RU14/megis.pdf

Thank you. This is fantastic stuff. Where can we find more of this sort of thing? I highly suggest this is more visible to people like us browsing the forums. It actually answered a lot of my questions.

Come to the MUM, that is what it's for

Hello all

since yesterday I am using (in production environment) a CCR1016-12G with 6.13, as BGP router connected to two peers

anybody else using a CCR as BGP router, with a recent build, can report that the initial issues had been solved? I am a bit worried... since in the first four months of 2014 we tried several 6.x builds, and with most of them BGP was not error-free (the router crashed suddendly and needed a power cycling to restart; we have two other CCRs in production, without BGP, and they all have several months uptime with no issue at all)

Leo

What did CPU utilization for BGP look like during these outages?

Just out of curiosity, could you share your watchdog settings?

we never enabled a watchdog on any MT router

should I do it, on the bgp ccr?

can I safely assume that, if I enable watchdog, next time that the ccr will have a BIG problem, it will reboot by itself, instead of crashing and needing somebody run there to power cycle it?

of course, I would be happier, if MT has sorted things out and with RC13 has already solved this issue; I am an optimistic person; I like to risk... actually I have already ordered one more ccr, a brand new CCR1016-12S-1S+, and would like to use it as our main bgp router

Hi paulsa,

What is your latest result?

keep us posted paulsa.

Thanks for the information, I am shopping for a low cost router for my new colo space, I ordered 2 100MB connections and will use BPG. I was debating using the Cloud router however I need something 100% stable. The other products I was looking at was either a Ubiquiti Edgemax Pro, A pfsence box, or getting a used Cisco or Juniper router on ebay.

So v6.13 seems stable? Anyone using a hotspot with usermanager? Anyone using PCC and needing to use allow remote dns use?

Our redundant concentrator running 4.17 on a RB1000 has been working perfectly for years.

in our case, UNTIL NOW, 6.13 seems stable, when used as BGP router; uptime 9 days

in our case, UNTIL NOW, 6.13 seems stable, when used as BGP router; uptime 9 days

@ddd

Our redundant concentrator running 4.17 on a RB1000 has been working perfectly for years.

It is even the same deal with this in my eyes!
If they (MikroTik) is not inserting many new features the half of the crowd is crying loud,
and if all this fine things find their way in RouterOS and there are some failures or issues
the rest of the crowed is crying loud! And what to do now if you don´t want to loose clients?

-Would the way UBNT was going the right thing?
Fiddle out all things and then delivering the CCR Series
-Is this way the right one? But I really thing they (MikroTik) will never be able to test out
all different situations and set ups their clients are using and having installed in the field.
So less features, options and functions should be the goal?

In my opiniion sorting out major bugs has the highest priority and comes long before adding features.

the situation now is that ccr is not usable in production because every new update that should solve some anoying bugs intruduces new bugs you have to work around. and it doesnot seem to get a lot better over time.

(selling only if the router is stable)
.

If a product is not fit for purpose then it shouldn't be sold.

If a user wants to unlock the full feature set they can do so at their own risk.

The current stability issues are going it very difficult to trust these routers again.

(selling only if the router is stable)
.

If a product is not fit for purpose then it shouldn't be sold.


In the case of the CCR it seems its only stable under certain configurations. I believe it would be best to release a cut back version of the firmware that only allows tested and stable configuration / packages.
If a user wants to unlock the full feature set they can do so at their own risk.


The current stability issues are going it very difficult to trust these routers again.

(selling only if the router is stable)
.

If a product is not fit for purpose then it shouldn't be sold.

In the case of the CCR it seems its only stable under certain configurations. I believe it would be best to release a cut back version of the firmware that only allows tested and stable configuration / packages.
If a user wants to unlock the full feature set they can do so at their own risk.

The current stability issues are going it very difficult to trust these routers again.

in my opinion the problem is the release cycle where they mix up versions with new features with bugfixes.
i would like to see a 6.7.10 for example. 6.7 was stable, compared to the next versions and they should have fixed the bugs there and released some minor-minor version until 6.7.x is stable.

so everybody would be happy and this plan would take some pressure from the development team to finally get out the "one first stable version" quickly.

And this is the part I can´t consider to you, related to the circumstance that a
first stable version with only the half activated feature set offered by RouterOS
or in other words without all from RouterOS given options and functions, only to
tell the whole public, "we have now the first reallystable release!"

And this is the part I can´t consider to you, related to the circumstance that a
first stable version with only the half activated feature set offered by RouterOS
or in other words without all from RouterOS given options and functions, only to
tell the whole public, "we have now the first reallystable release!"

But with this strategy i can decide if there is a stable Version with the Features i Need and if so i got to the shop an buy a MikroTik-router. otherwise i can always decide to take the risk of a Version with more features, but not so stable.

now i dont have the choice.

I have a very big problem with my CCR1009 with ROS 6.15, I prefix that I've bought it now so I test only with 6.12 and 6.15 but the problem persists.
The situation is this:
-ether1 with subnet 192.168.88.1/24
-ether2 with subnet 192.168.0.5/24
If a connect 2PC or router on the 2 different subnet I can't access the other subnet and I can't also ping.
The conf is very simply and with my rb951 work good with all versione of ros inclused 6.15.

I have a very big problem with my CCR1009 with ROS 6.15, I prefix that I've bought it now so I test only with 6.12 and 6.15 but the problem persists.
The situation is this:
-ether1 with subnet 192.168.88.1/24
-ether2 with subnet 192.168.0.5/24
If a connect 2PC or router on the 2 different subnet I can't access the other subnet and I can't also ping.
The conf is very simply and with my rb951 work good with all versione of ros inclused 6.15.

My CCR1009 is working great with 3 different subnets ... 1 on the bridge linking ether1-4 as switch group + ether 5 & 6 bonded for LACP + SFP+, ether 7 and 8 have their own subnets, with ether 8 actually leading to another router as backup WAN. Primary WAN (500Mbps symmetrical fibre) is on SFP.

I'm not sure how you configure your CCR1009 but perhaps you can try putting the 2nd subnet on the non-switch group (i.e. ether 5 to 8 ).

I've already tried to do this, so I don't know what's the problem.
What ROS version and firmware do you have?

I would check firewalls on both PCs: Windows Firewall blocks ICMP not from local subnet at least in WinXP, for example

you can check with Tools -> Torch on both interfaces - I'm sure, there are packets received and transmitted

This really should have been in a separate thread.

What is the default gateway on the 192.168.88.253 PC?

What is the default gateway on the 192.168.0.250 ATA?

Can the PC ping 192.168.0.5?

On the 192.168.0.251 mikrotik,

Code: Select all

/ip route
add gateway=192.168.0.5

If the 192.168.0.0/24 hosts don't know how to get back to 192.168.88.0/24, how can they respond to pings and traceroutes and web requests from devices in 192.168.88.0/24 or any other non-192.168.0.0/24 hosts?

On the 192.168.0.251 mikrotik,

Code: Select all

/ip route
add gateway=192.168.0.5

I forgotten the gateway :shock:
So now I added it and from the mikrotik 951 (192.168.0.251/24) I can reach the ccr1009 192.168.0.5 but not 192.168.88.0/24.
From the ccr1009 (precisely 192.168.88.1) I can't reach 192.168.0.251 despite I added the rule

Code: Select all

add distance=1 gateway=192.168.0.251

Perhaps is needed static routing indirect?

On the 192.168.0.251 mikrotik,

Code: Select all

/ip route
add gateway=192.168.0.5

I forgotten the gateway
So now I added it and from the mikrotik 951 (192.168.0.251/24) I can reach the ccr1009 192.168.0.5 but not 192.168.88.0/24.
From the ccr1009 (precisely 192.168.88.1) I can't reach 192.168.0.251 despite I added the rule

Code: Select all

add distance=1 gateway=192.168.0.251

Perhaps is needed static routing indirect?

Now that I'm on a real screen and not the smartphone...

According to what you wrote, you told the 951 that it was its own default gateway. Set the gateway to 192.168.0.5.

Hi everyone,
I'm looking for detailed implementation information for the Tilera architecture.

Basically, while considering CCRs as candidates to offload some tasks from Cisco and Juniper boxes, we'd like to get some deterministic scales based on the number of cores involved in each and avery bare-metal-implemented feature.

Which features are implemented on bare tiles ? How many tiles per feature's scalability (i.e. core per x thousand L2TP session, core per y Gbps of switched or sampled traffic, etc...) ? Is there a dynamic ressource allocator or is it hardcoded ? If dynamic, how to proritize when oversubscribing ?

200Mb on my WAN side.
about 300 customers connected using pppoe
not very complex firewall rules (35 filter rules, 25 nat rules and 15 mangle rules)
and 30 simple queues using pcq.

But my concern is that i will serve inet connection to my customers using JUST ONE ETHERNET on the LAN side.

Does it affect to the performance a MULTICORE router can afford?

Make sense to get a 16 or 32 core router with this topology limitation ?

As far as i can see a 9 core router can accomplish the load without issues...

Very useful information about CCR is available here:
http://mum.mikrotik.com/presentations/RU14/megis.pdf

Thank you. This is fantastic stuff. Where can we find more of this sort of thing? I highly suggest this is more visible to people like us browsing the forums. It actually answered a lot of my questions.

Come to the MUM, that is what it's for

So made the mistake to get myself a CCR1036-8G-2S+

6.22 (3.19 firmware) *all* Ethernet ports connect at 10/Half Duplex (multiple switches/servers, multiple cables), *all* Ethernet ports receives 0 frames/packets according to Interface statistics.

Upgraded to 6.23 (3.20 firmware) *exactly* the same issue.

Nice. I have a USD1K door stop...

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

Dear Normis!

The CCR-1036 support SO-DIMM ECC RAM memory ?
(...like: Kingston SO-DIMM 8 GB ECC DDR3-1333 x2)

What type of SO-DIMM memory is the best for the CCR-1036 ?
(ECC vs. NON ECC, 1333MHz vs. 1600MHz)

Please help choosing between these...

Thx.

Attila

The Tilera CPU supports it from the DDR3 controllers onboars, so it should work.