[SOLVED] LoadBalance, packets being dropped - nf_conntrack/net_ratelimit



  • Good evening,

    I have come through this candidate of assist of participants / users with more experience to me some candidate in this, that lately has taken away my sleep.

    I'm having problems losing packets on the links they have, after increasing speed, I'll show them the scenario:

    BEFORE UP [

    MODEM-01 - 120M / 10M - NET
    MODEM-02 - 60M / 6M - NET

    PFSENSE - 1 WAN / 1000 - 1 LAN / 1000 - i5 3330 - SSD 128 - 4 GB DDR3
    RB2011 - With VLANs for a WAN to do pfsense, this way to pass bridge links with direct DHCP pro pfsense

    ]

    AFTER UP [

    MODEM-01 - 120M / 10M - NET
    MODEM-02 - 240M / 20M - NET

    PFSENSE - 2 WAN / 1000 - 1 LAN / 1000 - i5 3330 - SSD 128 - 4 GB DDR3
    "I removed RB2011 and installed 1 more card in pfsense, believing that this could be a problem. But I tested with an RB2011 too, and it lost the same way"

    ]

    Before the UP, I've never had problems, there have never been complaints about a paging battery or the clues lost in online games.

    But you do not have to choose, you just need a band and request a 240 mega download pack, make a modem trocar, make a bench game and everything perfect, but it must have been replaced by a blessed modem, which has double (hgb10r-02) and make (speed) the speed and speed of the speed dial. The speed of the speed dial is shown in the figure below. humax) and to my surprise, a packet loss continues ..

    After the up and down, I discovered some things, like I'm not sure I'm right, continue below:

    1 - The packet loss is not immediate, it starts in 1 hour, it is not rule, sometimes it takes 2,3,4 hours to start it ..

    2 - Also not of packet loss, with consumption below 40 Mega, only when above 50,80,100 ..

    3 - At the modem log, remove these lines:

    kernel: nf_conntrack: full table, losing packet <- This titles English
    [kernel: net_ratelimit: 779 callbacks suppressedet] <- This represents 2, 3 times

    4 - When you change the MAC address of the WAN and restart the modem, it spends a longer time to present as packet loss.

    When I searched these lines on google, I found this article:
    https://forums.gentoo.org/viewtopic-p-4575074.html?sid=aafd12b9dce5b8bd995d748fa996043e

    But the same is not service, it is not necessary to remove from it the correct information to arrive at some solution / rule.

    And what puzzled me the most, the other modem lost the ability to make its 120 mega pack, and its limit is 120 ~ 125.

    PS: In this period, also requested a reduction of the plan to 120 Mega, which did not make much of a difference, continued to lose packages ..

    This is no longer the case, I apologize for having extended it, but I believe that as much information is possible, if you need to send an error or return some command to DEBUG find a solution for this.

    Thank you all, long life and thrive!



  • SOLVED,

    I migrated to NethServer.

    Thanks for nothing!