Announcement

Collapse
No announcement yet.

Zabbix Server 3.2.9 - Action Auto Registration Message Flood

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

    Zabbix Server 3.2.9 - Action Auto Registration Message Flood

    Hello,
    since I updated to Zabbix Server 3.2.9 on 2017-10-27 09:42:34 exactly, two things happened:

    1. First of all, I started receiving Auto Registration Messages via SMTP and via Telegram API Script once again. I was very happy about that, because that stopped working months ago, but I was to lazy to search for it. Sadly I can't remember on which version it stopped working.

    2. Second, it never stopped working since then! I literally get flooded with Auto Registration Notifications! I have a screenshot of my Telegram History for you, as a brief example, but this is only two servers, as I get over 5-8000 messages per day.
    https://i.imgur.com/blyxHCW.png
    The Action log looks similar:
    https://i.imgur.com/QTcDCZq.png

    Any ideas, or should I open a bug report?

    #2
    Someone has any idea whats going on? This is driving me nuts.
    https://i.imgur.com/t80dw69.png

    Comment


      #3
      Autoregistration actions can send messages yet fail to register the host. If they do that, it keeps firing because the host doesn't exist.

      Are the hosts being enrolled? Are they hitting the same server or proxy that is discovering them?

      Comment


        #4
        Autoregistration actions can send messages yet fail to register the host.
        That is an interesting point, but I don't see a correlation just yet.
        Where can I see if it fails?

        If they do that, it keeps firing because the host doesn't exist.
        That doesn't fit, because all of that hosts already exist.

        Are the hosts being enrolled?
        No, it happens in a productive Environment where all Hosts/agents are already working for years. It started with the update to 3.2.9.

        Are they hitting the same server or proxy that is discovering them?
        We are not working with discovery, but with the Agent Auto Registration via HostMetaData https://www.zabbix.com/documentation...o_registration

        Comment


          #5
          Sorry, my last question should have been autoreg, not discovery.

          How many proxies do you have? Are these hosts monitored by server or proxy? Do these hosts have more than one entry in ServerActive?

          Post your autoregistration action that is generating the messages.

          We just had a message flood issue similar, but self induced. We have actions to add systems based on metadata, but different naming patterns hit different actions. One action that matched on just metadata sent the message. A new naming pattern was created and we weren't told, so no action that enrolled the host matched, but the notification action still matched.

          Comment


            #6
            No problem :-D

            >How many proxies do you have?
            ~25

            >Are these hosts monitored by server or proxy?
            Proxy

            >Do these hosts have more than one entry in ServerActive?
            No

            >Post your autoregistration action that is generating the messages.
            https://i.imgur.com/htxZcBy.png

            But I think I found the issue. But I stick to my opinion that Zabbix 3.2.8 or 3.2.9 changed something.
            The Hostname in the "zabbix_agentd.win.conf" of the affected Hosts is different then the Hostname used in the Frontend. Example:
            zabbix_agentd.win.conf -> SERVER01.domain.local
            Frontend -> server01.domain.local

            This leads to your suggestion "Autoregistration actions can send messages yet fail to register the host."

            Comment


              #7
              Yea, bet they are windows systems aren't they?

              We tried to use hostname.item and our windows systems would change the case of their hostname when patched and rebooted.

              Comment


                #8
                Yes, the affected systems are Windows OS hosts.

                We don't use the hostname item, we use a BAT script which generated the hostname. But the BAT generated the hostname UPPER case and in the Frontend we have the hosts lowercase.
                That must be the problem.

                Comment

                Working...
                X