we’ve installed v3.0rc5 on one of our rb532s (32MB RAM!). the traffic is better (packet loss,etc) but after two hours of operating, we couldn’t login, we got only ping.. after several hours we saw the memory usage and it was on 100%. Has v3.0 mikrotikos memory requirement increased? and if this the case, how can we decrease it? this RB operates just as an AP, doesn’t have any special firewall or queue rows in it… (this AP operates just as a transmitter, QoS is configured in our ‘central’ mikrotik)
yes it is.
I am reporting this leak since first rc.
Produced several supout after the reboot and after a while to compare the memory usage.
But they still were not be able to reproduce it.
Hope if more people start to report then they will find a solution.
Just kidding, but you could really use that as a workaround…
Happened in 3.0 up to rc5, to me also, on rb133 and rb112 too. Had no 532. But i also had 100% cpu too… so hard to tell what it was. Now… 2.9.46 is with me…
that would be nice if all of you contacted support with latest ROS version available (in time of writing RC5) supout.rif files.
if it is possible to connect to those boards we would appreciate if we could access these boards (for precise ip addresses/usernames/passwords/access times we will contact you, just state that if we need we can ask you for permission to connect there)
describe configuration, describe how fast the memory is full.
if it is possible create supout.rif file when memory is almost full.
I just put rc6 on a rb133 and it seems fine. At least i got rid of 100% cpu.
I will have by the end of the week a rb133c and a rb112 available and i’ll put them for a few days to see what happens.
rb133 works since this afternoon, and doesn’t show strange behaviour.
Rock solid till now.
[admin@#######] > /sys pack pri
Flags: X - disabled
# NAME VERSION SCHEDULED
0 routerboard 3.0rc6
1 system 3.0rc6
2 wireless 3.0rc6
3 ppp 3.0rc6
4 X ipv6 3.0rc6
5 routeros-rb500 3.0rc6
6 advanced-tools 3.0rc6
7 hotspot 3.0rc6
8 security 3.0rc6
9 dhcp 3.0rc6
10 routing 3.0rc6
[admin@#######] >
The board is nat-ing a single public ip to 2 private subnets (1 wireless, 1 wired. Uplink wired.). 20-30 simultaneous wireless clients connected all the time.
The firewall is the one created with the web login to the router.
[Wich by the way, since 2.9.2xxxxxx don’t remember wich, needs to corect comment to the rule no. 1 to say Allow established connections instead of Allow esatblished connections ]
CPU around 20-27% all the time. Wireless clients connect with wep 104/128 bit.
(Now that you asked, i see that i have to disable ppp, hotspot, routing and dhcp since they are not used.)
Had to return.
Since yesterday, just talked, and it got… weird.
The system is routing and everything, winbox does not stay connected more than 3-4 minutes.
And before that, the cpu usage goes crazy. 2 - 5 - 2 - 25 - 7 - 100 - 100 -100 - 30 - 5 - 45 - 100 and so on.
I’ll chech to see if else is wrong, but till then I’ll just move that board to my home, and put a pc in there.
seeing the same thing on some 532’s running rc6… also dropping packets on the ethernet link. rc5 definititly has a memory leak, we have 3 test APs on rc5 and one one rc6, in the last 12 hours I’ve had to reboot two of the 3 rc5 APs… I’m not sure if I want to load rc6 yet on them with the disconnect and packet loss issues we’re seeing…
Well, actually, rc6 seemed better than rc5. In rc5, could not see proc lower than 100%. In rc6 … it acts at it’s own will. I am righ now assambeling a link with a 133 and 133c, and i’ll try both bridge and route, in the same room. So, i’ll just have to see what’s the point of this. And yes, i will net install them before anybody tells me to do so and it will fix my problem. The traffic should be no more than 1 mbit, as i see it. Hope i’ll be back in a few hours to tell you.
( to domain admins: if this should be moved in some topic about rc6 in rb1xx series, please do so. )
