What's new in 5.0rc1 (2010-Sep-17 13:58):

*) wireless nv2 - add WDS support;
*) added ability to monitor cpu usage by task (/tool profile);
*) added support for RPS (Receive Packet Steering) on multicore systems;
*) added cisco compatible gre tunnels;
*) fix simple queues on multicore (fixes intel 82576 crash/reboot);
*) fixed sstp memory leak;
*) winbox - fixed ability to configure RB1100 switches;
*) ssh - fix possible crashing;
*) snmp - fix GET request processing;
*) ipsec - added ability to specify ID_USER_FQDN peer identity;
*) ipsec - encypt last IKE phase 1 packet when in aggresive mode as initiator;
*) wireless nv2 - added nv2-cell-radius parameter to specify distance in km to farthest client on AP;
*) wireless nv2 - fix transmit stall issue for non-11n chips;
*) wireless - fix x86 nstreme multiprocessor hang;
*) made user manager web interface work again;
*) updated drivers and kernel (to linux-2.6.35);

http://www.mikrotik.com/download.html

MEMORY LEAK ISSUE

The router is a x86, Intel, dual CPU/core

memory leak on graph.gif

P.S. /system hardware set multi-cpu=no seems to have stabilised it for now. 1 core is enough for now I think. Until a more stable version is released at least. What did this setting do, remind me, it switches the kernel binary?

P.S.2 I have not written to support about this. I have more important activities at hand.

Mangle & QT from ver 4.11 aren't working on 5.0 rc1

yes leak still there for x86 looks like they just cant find it

Become a developer for a company and see for yourself that in fact the company Can find all the bugs they want. Reality is different.

So calm down and send a supout to support@mikrotik.com from the system that has the problem

Is this v5 RC1 stable enough to start trying it on real live routed (no WDS) wirless network with relative simple configs and all rb boards and radio's?

I am very tempted to start using the new wireless features on my network because I still have the issue of very heavy congested airspectrum with many hidden nodes. Also some of my backhauls could use some boost.

But after some earlier disasters and reading the many bugs around (on more advanced funcions though) I am a bit reluctant to start uploading RC1 to my devices.

Anybody already having experiances with 5RC1 in an all MT (radio's included) environment?

There is a few bugs. IMHO is to early to deploy this version in production environment. Somehow I'm dissapointed with that RC1 it should be beta7.

I am not disappointed. So far it works. I have not encountered the Mangle bug even.

On RouterBAORDs it works good. Still using the nstreme (not V2) but I like what I see, even compared to v4.11.

Later if I have time I will upload it to more PRODUCTION machines. (Yes, I am crazy).

I still can't authenticate clients via RADIUS w/ NV2... have to use local ACL.

I have deployed on two production AP's but they are very small only 3 clients on them.. NV2 seems to be working well, but it's only the 6hr mark. I had problems with all clients disassociating with the AP at random times and not reassociating. Regular NSTREME works fine.

I have not encountered the Mangle bug even.

I think that someone else got the problem with Mangle - connection mark.

I have it on about 10 routers. Memory leak on my RB1000 still seems to be there. Booted to ~460mb free ram with 127 SSTP connections, after 6 hours, its down to 440mb. I turned on an idle timeout to try and accelerate the effect.

User manager still not work, almost in x86 platform.

The direct downloads in the web page doesn't work, only the torrent is working.

I hope you can fix it shortly.

Regards

The torrent file stop on 99.2%

Firewall not working.

I have not encountered the Mangle bug even.

I think that someone else got the problem with Mangle - connection mark.

You and TKIT are two people that have seen it so far

Model RB750G
3 L2TP client without encryption.
3 openvpn client aes128 encryption.
Version 4.11 works perfectly, load average around 40 percent.
After updating to 5rc1, mikrotik even tunnels can not lift, loading 100 percent and constantly breaks the connection winbox. In the logs at the same time throws out the old and unsolved problem - you can not use the new address, as it already exists. That is, it turns out that when picked up the new tunnel, mikrotik no time to remove the old ip address.

Mangle & QT from ver 4.11 aren't working on 5.0 rc1

Please, do not make such imprecise posts anymore, i got my coffee cold while jumping to check all the routers. So far all my connection marks for routing marks and packet marks are working fine, i will keep looking, but, looks to me ether it is configuration error, or specific complex situation is necessary to repeat the problem.

Can you give us all the details?

5.0rc1 in normal Wlan 802.11n Link with MPLS/VPLS Backbone without Firewall, QoS or so works very good. Same with two R52 on RB/600 in 2km Link. Stabil Link with low Latence by >20Mbit traffic.


dingsingo

mangle and queue works fine on x86, no more problem with queue tree, found some bug with RPS - recomended to disable RPS in this version. Supout send to support.
And RC1 still has bug with torch on intel ethernet.

Anybody check if stability of 802.11n wireless is better?

1. found some bug with RPS - recomended to disable RPS in this version
2.And RC1 still has bug with torch on intel ethernet.

Could you be more specific?

I'm not going to use "connection tracking" in my configuration.
Should RPS help me or it works only for connection tracking?

RPS help on high load ethernet links, it does not irrelevant to connection tracking.
And another bug - on x86 multicore cpu and intel ethernet with many queue when run "profile" - router reboot. Already send supout

Anybody check if stability of 802.11n wireless is better?

Hi, it works very fine. Better than 5.beta5...

dingsingo

Mangle & QT from ver 4.11 aren't working on 5.0 rc1

Can you give us all the details?

RB433AH on Ver 4.11 everything works fine. On 5.0 RC1 mangle don't "catch" traffic. Traffic on WAN interface goes 100%, so I'm checking QT and there's no treffic there. I've limits for clients in forward(4 mangle entries, & 2 QT entries for each user) and QoS in prerouting. I didn't make new config. I've just upgraded from 4.11. I checked mark names in console, and everything seemed to be ok. Do I have to make new config for 5.0RC1 or post mangle & QT listenings here? I go back to 4.11 and everything is OK.

weconfig to choose to cancel at the time of login page is in a loop.

weconfig to choose to cancel at the time of login page is in a loop.

It closes the login, and shows blank webfig page.

upgraded a rb450g and can now only get ethernet link on port1, not on rest of the ports.

here is my test result:
Server: HP DL360 G6
CPU: Intel Xeon E5620 2.40 GHz Quad core
Traffic: 250Mb/s IN / 250Mb/s OUT
-----------------------
Bios setting:
Hyper threading: Enabled,Disabled
Number of cores: All,Half,One
ROS setting:
set multi-cpu=no
reboots forever, reaches login screen and reboots
-----------------------
Bios setting:
Hyper threading: Enabled
Number of cores: All
ROS setting:
set multi-cpu=yes
RPS: enabled
result:
CPU count: 8
CPU usage: %5
crashed after a minute
-----------------------
Bios setting:
Hyper threading: Disabled
Number of cores: All
ROS setting:
set multi-cpu=yes
RPS: enabled
result:
CPU count: 4
CPU usage: %8
Mark-Packet: working
Queue Tree for marked-packets not working
-----------------------
Bios setting:
Hyper threading: Disabled
Number of cores: Half
ROS setting:
set multi-cpu=yes
RPS: enabled
result:
CPU count: 2
CPU usage: %22
Mark-Packet: working
Queue Tree for marked-packets not working
Traffic flow was unstable , going up and down

upgraded a rb450g and can now only get ethernet link on port1, not on rest of the ports.

I upgraded one 450G as well. It works. I have L3 on 1, 2 and 5 and 3 &4 are switch slaves to 2. It has OpenVPN that is working as well. Mangle connection marking as well. This and that...

P.S I have not tested Queue Tree though.

Wireless on 711nH - ping better, performance the same

here is my test result:
Server: HP DL360 G6
CPU: Intel Xeon E5620 2.40 GHz Quad core
Traffic: 250Mb/s IN / 250Mb/s OUT

Do you run latest BIOS?

Firewall do not work for me too on X86. Every counter show 0 at the firewall rules. Supout.rif is sent to support.

Is there an existing fix for snmp gets not working?


SNMP Error:
Received SNMP response with error code
error status: noSuchName
index 0
SNMPv1_Session (remote host: "10.0.180.180" [10.0.180.180].161)
community: "MT-Avant"
request ID: 391607887
PDU bufsize: 8000 bytes
timeout: 2s
retries: 5

Until we figure this out, we cannot move anything to 5.x.....


Thanks!

Anybody check if stability of 802.11n wireless is better?

Not sure what you're referring to exactly..

I had no problems with stability and 802.11n on 5.0beta6.. Only stability problems I noticed were when using NV2. NV2 seems to be much more stable I have clients connected for 1day5hr+ which is about how long ago I applied the update. Before they would all drop and not reassociate at random intervals not exceeding 12hrs.

So it appears to be more stable wirelessly... Just still not able to MAC-Authenticate clients using RADIUS when using NV2.

Is there going to be the long-awaited Lua scripting support in v5.x? rc1 is getting pretty close to final stable release.

I've been waiting very patiently (and disappointed when it was dropped from v4)

Regards,

sorry but no, there were big problems with it.

For everybody that has problems with firewall in RC1, try to disable all rules that have dst-limit setting

And another bug - on x86 multicore cpu and intel ethernet with many queue when run "profile" - router reboot. Already send supout

How many queue should be?
I'm going to use 30-50 simple queue rules.

its not a shaper queue, i mean ethernet hardware queue on intel chip 82576 or 82575
Shaper queue (simple queue or queue tree) works fine on v5RC1, tested on 1300 simple queue and queue tree

GRE tunnels! Thank you mikrotik!

Can´t format sd-card.
Revert back to 4.11, format is fine.
5.0rc1 can see it after install.
But trying to format under 5.0rc1 renders it invalid and useless.
RB450G and sandisc 2gig card.

I'm using 5.0RC1 in a couple scenarios:

1. AP with 10 subscribers. All running 802.11A with Nv2. Working great, seemed to get 10-15% better actual throughput vs nstreme. WDS bridged subscribers.

2. PTP - A few links with N cards, a few with A. All seem to work very well. Better latency, higher throughput (no measurements yet - but probably in the 10-15% range)

Negative things I've noticed:

1. CPU on some units jumps to 100% and stays there. Doesn't seem to affect performance too much. We think it may be when we make a vlan interface change, and someone else has seen something similar on SNMP2 queries - trying to narrow this down. /tool profile on mine show high usage in the "management" and "unspecified" categories - about 60%/40% when this happens. Rebooting fixes it.

2. BAD things happen when jumping firmware (/system routerboard print ) versions. I've had one unit lose its IP config, another I had to netinstall. Future upgrades I'll take up to another 5.0 beta to get the 2.27 firmware before upgrading to 5.0RC1. I believe the ones that had problems started with 2.25 or lower.

3. Remote client wireless stats are missing lots of detail such as CCQ, remote power, etc in the registration table. I assume this will be fixed soon.

I think there is a problem with hotspot. I had to downgrade 5 RB450gs because they would reboot randomly and give autosupouts. 4 of them have almost identical configs. 1 of them is in my office and has a very different config. But all 5 use hotspot.

My RB1000 still has the memory leak, but hasnt rebooted yet. No hotspot, SSTP Server
My RB1100 at home is running fine, no hotspot, with SSTP client
RB433AH as an AP in my office is running fine. Not Hotspot, no SSTP client
RB433AH at customer site running fine, No Hotspot, with SSTP-client
3x RB433AH at home offices running fine, no hotspot, with SSTP client
2x RB450g running IPSEC between each other, no SSTP client, no Hotspot.

Only commonality I can find is hotspot.

I've got a tester RB1100 with several Hotspot VLANs and light usage (20 or so people at a time at the Corporate office). No memory issues so far (steady at 30MB used), no reboots. Uptime is just under a day.

Maybe it only affects the RB4xx series with hotspot. The memory leak only happens on my RB1000 w/ SSTP server (over 100mb gone in just over a day). The only RB4xx devices I have had a problem with are all running hotspot.

with nv2 enable rate is 270/270 while with nstreme 300/300
tried diferent channels, tx power.. everything. couldn't make 300/300.

ping nv2 vs nstreme = 4ms vs 0.7ms

else is ok, bandwidth about the same.

its not a shaper queue, i mean ethernet hardware queue on intel chip 82576 or 82575
Shaper queue (simple queue or queue tree) works fine on v5RC1, tested on 1300 simple queue and queue tree

how did u checked how many harware queues where been in use?

how did u checked how many harware queues where been in use?

He wrote:

Shaper queue (simple queue or queue tree) works fine on v5RC1, tested on 1300 simple queue and queue tree

Not hardware buffer queues.
To software simple queue:
/queue simple print count

is the ccq and tx signal strenth suppose to be missing in winbox

/ping is still not working under scripts in the RC1 release (except for the very first time it is used), I'm not sure that going to an RC1 and not beta7 is strictly correct if there are things that worked perfectly in a previous version that have not yet been resolved in a beta version, upgrades are not meant to break previously working functions, especially if they have been reported as bugs through beta releases.

Surely RC1 comes out once they think everything is working to specification but reading through this list of issues we don't quite seem to be there yet. I first reported the /ping issue in beta4 and they did replicate the issue.

Cheers, Dominic.

Packet sniffer at OpenVPN interface did not capture anything.

ccq and tx-signal isn't showinf currently for Nv2 links.
and currently Nv2 doesn't support short guard interval and that is why the max data-rate is 270Mbps

For everybody that has problems with firewall in RC1, try to disable all rules that have dst-limit setting

??? i don't have dst-limit sir

then you have different problem. email support and clarify what problem you have, and include a supout.rif file

is the ccq and tx signal strenth suppose to be missing in winbox

On a rb711 as 801.11 client connected (wireless) to a rb333 running v4.11 these data are to be seen in winbox of the rb711.

ccq and tx-signal isn't showinf currently for Nv2 links.
and currently Nv2 doesn't support short guard interval and that is why the max data-rate is 270Mbps

Yes, I observed this too. It's keeping me from testing nv2 as I can't get a sense of CCQ and TX power relative to normal nstreme.

Normis,
Could you please briefly summarize problem (major) that you can confirm for RC1?

im found another bug:

this script doesn't work from /system scripts:
host 172.17.255.2 is reachable but local ppp secrets is still enabled when im runnig this script

but when im running this script directly from terminal all working correctly

5RC1 should be named 5beta7...

5.0RC1, x86 platform, we have one problem on the link on 2.4GHz-onlyG, the only one, 40MHz channel width, with nstreme enabled, framer policy=best fit, framer limit=3200, frequency upper than 2472MHz ( 2494 to 2502 ), hw retries=10, too many disconnections and the router freezes very often, low bandwidth, it's not possible to access to it at all, I have to reboot it by switching off the power...

When I disable nstreme and enable nv2 nstreme, link becomes better and more stable than before, bandwidth is around 45Mb/s real tcp traffic, the same frequency, just for notice.

4.11 don't have such problems.

Any of you guys using the RB1100 having issues? Mine has been acting strange a short while after the upgrade to RC1. I'm continually losing my link on the ether13 (poe/boot) port and the routerboard randomly rebooted on me.

I power cycled the router after tolerating the problem for some time. After the power cycle the link drop isn't happening as often, about every 4-5minutes. Before power cycling it was happening once every 30-60 seconds.

I'm going to try another port to see if I can replicate it... just seems odd as its so cyclical.. Monitoring the ping tests icmp-sequence # I show the drops happening every ~280 packets. Winbox doesn't drop the running flag for a few seconds after the drops start occurring, which could just be a delayed reaction in the GUI.

And yes I have verified the physical layer connections - it works just fine with no drops/packet loss/etc when I connect my laptop directly.

Ports 11, 12 and 13 on RB1100s were giving me all kinds of grief even on 4.11 so I pretty much stopped using them completely.

Ports 11, 12 and 13 on RB1100s were giving me all kinds of grief even on 4.11 so I pretty much stopped using them completely.

bummer... OK well at least i know its not just me!

wildbill442, what device is connected to ether13 when these problems occur?

Is possible to "readd" ack timeout to winbox in new versions /rc2 / as classic us value - in legacy 11A mode..? Distance is not very helpful in problems with clients...

Does anyone know if the official 5.0 f/w will be released this year? I have a few 25 mile PtP links on SR71-15's that are not performing good, in hopes this new firmware will fix signal/throughput issues. Been waiting a long time!

- Formatting second store does not work. (!)

- Booting with PLoP on PCs that have USB boot problems works (as long as PLoP supports the USB controller on the Motherboard/slot)

- Intel Pro/100 does not work as good as Realtek 8139 (not thoroughly tested, so not 100% sure)

- Installing with Netinstall fails miserably

- Installing with CD is good

Would like to report a bug with ssh.

Since the update was made to disable ssh password auth to accounts with ssh-keys, I am now unable to use those ssh keys
to login with extended console params added to username as in : admin+c160w@<router-ip>.

I can live with the password disabling feature, but this method greatly simplifies parsing the output in our external scripting.
Please consider looking into this?

Thanks,
ReenignE

When 5.0 RC2 or finnal ?

Dear MikroTik support, developers and Q&A,

- Exporting Netwatch to .rsc and then pasting back into a router does not work. Either exporting or pasting in is broken. Or both. This applies to the syntax.

- When you alt-tab the local terminal window inside WinBox breaks in an unpleasant way.

- exporting of user-manager customers happens with a bad parent in case admin was changed, so pasting back in does not work

- exporting of user-manager users with hide-sensitive still shows passwords

- exported config with hide-sensitive should be noted on top and in the filename to prevent problems of missing passwords in the backup later

- exported config (.rsc) did not include the user-manager profiles so importing back resulted in not working setup + additional work to make the profiles again

- Total Sector Writes and Sector Writes Since Last Reboot are for main store or for all stores summed? Please separate those for each store.

- The file browser in WinBox is very slimmed. Does it waste 1 byte of code? Make it a little more. Like 2 bytes. Geez.

For everybody that has problems with firewall in RC1, try to disable all rules that have dst-limit setting

Other rules fail as well - the content matcher is at fault here.

If filter and mangle rules are failing, this surely is not RC1 code. I know this doesn't help much in resolving any problems, but I just wanted to voice my opinion.

Mikrotik, please stablize v5 before final release. I'm starting to get depressed again. v5 (4 previous versions), should be about time right?

And I tested my x86 machine with memtest, installed windows on it. ran it for a couple of days nonstop, ran a stresstest. So I did my homework before making that machine a router. Plus I added cooling. Plus it's a DELL machine so it's expected to be good and it looks good (parts, capacitors, is heavy etc). Plus its a single core/not HT.

Good thing I had a spare pre-configured router bwahaha >:)

Hello,
I think torch in 5.0rc1 have some bug, i start to download file my download speed about 73mbps but why other users speeds also increase in torch, pleas see torch tx rates and rates in interfaces meniu.

Would like to report a bug with ssh.

Since the update was made to disable ssh password auth to accounts with ssh-keys, I am now unable to use those ssh keys
to login with extended console params added to username as in : admin+c160w@<router-ip>.

I can live with the password disabling feature, but this method greatly simplifies parsing the output in our external scripting.
Please consider looking into this?

Thanks,
ReenignE

on what arch this happens, my test ppc router took "ssh test+c160w@hs -i id_dsa" command happily

Would like to report a bug with ssh.

Since the update was made to disable ssh password auth to accounts with ssh-keys, I am now unable to use those ssh keys
to login with extended console params added to username as in : admin+c160w@<router-ip>.

I can live with the password disabling feature, but this method greatly simplifies parsing the output in our external scripting.
Please consider looking into this?

Thanks,
ReenignE

on what arch this happens, my test ppc router took "ssh test+c160w@hs -i id_dsa" command happily

This is happening on mipsbe: RB433.
/system resource print output is:

uptime: 1d3h36m51s
version: "5.0beta6"
free-memory: 50332KiB
total-memory: 62252KiB
cpu: "MIPS 24Kc V7.4"
cpu-count: 1
cpu-frequency: 300MHz
cpu-load: 42%
free-hdd-space: 30888KiB
total-hdd-space: 61440KiB
write-sect-since-reboot: 62896
write-sect-total: 6841782
bad-blocks: 0%
architecture-name: "mipsbe"
board-name: "RB433"
platform: "MikroTik"

Thanks,
ReenignE

Some bug in RC1
See the attachment This link works on 5Ghz-20Mhz

Would like to report a bug with ssh.

Since the update was made to disable ssh password auth to accounts with ssh-keys, I am now unable to use those ssh keys
to login with extended console params added to username as in : admin+c160w@<router-ip>.

I can live with the password disabling feature, but this method greatly simplifies parsing the output in our external scripting.
Please consider looking into this?

Thanks,
ReenignE

on what arch this happens, my test ppc router took "ssh test+c160w@hs -i id_dsa" command happily

This is happening on mipsbe: RB433.
/system resource print output is:

uptime: 1d3h36m51s
version: "5.0beta6"
free-memory: 50332KiB
total-memory: 62252KiB
cpu: "MIPS 24Kc V7.4"
cpu-count: 1
cpu-frequency: 300MHz
cpu-load: 42%
free-hdd-space: 30888KiB
total-hdd-space: 61440KiB
write-sect-since-reboot: 62896
write-sect-total: 6841782
bad-blocks: 0%
architecture-name: "mipsbe"
board-name: "RB433"
platform: "MikroTik"

Thanks,
ReenignE

Sorry, I just realized these were at beta6 and not rc1. When I upgraded the ssh issue went away. Thanks for checking, though.
ReenignE

wildbill442, what device is connected to ether13 when these problems occur?

Sorry for the late reply -- it was a Motorola DOCSIS 3.0 Cable modem. Can't remember the model off hand. I was using ether13 as the WAN port for the router.

seems to connection mark is not working too! yesterday I upgraded from beta 6 to RC1, today can't connect to router thrue winbox. Thanks god I can operate it thrue console Fixed it, when unsetting connection mark...

Memory leak on RB450 running sstp server, board runs for about a day and runs out of memory

I can verify the mangle connection marking bug on my RB1000 v5.0rc1. Traffic is still classified under the mangle tab....yet the connection marks do not pull over to the connections tab making it useless.

it's known that marks do not work in RC1. wait for RC2

i hope the ospf problem have been fixed until 5rc2 . i had to disable ospf in our two bgp routers just put a cisco l3 switch to do the ospf DR/BDR job with a static default route to our bgp routers