Results 1 to 2 of 2

Thread: Livestatus Errno 104

  1. #1
    Junior Member
    Join Date
    Apr 2016
    Posts
    4

    Livestatus Errno 104

    Hello everyone,

    I have installed a Shinken HA architecture with 2 nodes with Thruk and NagVis frontends with the Livestatus module. It worked for a while (it's not yet in production) until last friday when the Livestatus broker was broken simultaneaously and unexpectedly on the 2 nodes :-(

    Here are the error messages which are written in loop in the brokerd.log:
    Code:
    [1460239205] ERROR: [broker-master] LiveStatusClientError: Could not read from client: [Errno 104] Connection reset by peer
    [1460239205] WARNING: [broker-master] Error on client socket shutdown: [Errno 107] Transport endpoint is not connected
    My software versions:
    • CentOS 6.7 x86_64
    • Python 2.6.6
    • Shinken 2.4.2
    • Livestatus module 1.4.2
    • Thruk 2.02
    • NagVis 1.7.10


    Livestatus config file:
    Code:
    ## Module:      livestatus
    ## Loaded by:   Broker
    # The LIVESTATUS API makes internal Shinken data available via the network
    # using an SQL-like syntax. The API supports various access methods,
    # authentication and sophisticated performance options. The premier interface
    # to Shinken internal host and service states, historical data, performance
    # data, configuration data, comments, maintenance periods, etc.
    define module {
        module_name     livestatus
        module_type     livestatus
        host            *           ; * = listen on all configured IP addresses
        port            50000       ; port to listen
        socket          /var/lib/shinken/rw/live  ; If a Unix socket is required
        ## Available modules:
        # - logstore-sqlite: send historical logs to a local sqlite database
        # - logstore-mongodb: send historical logs to a mongodb database
        # - logstore-null : send historical logs to a black hole
        modules         logstore-sqlite
        #debug           /tmp/ls.debug   ; Enable only for debugging this module
        #debug_queries   0   ; Set to 1 to dump queries/replies too (very verbose)
    }
    Thank you for your help, I don't really have a clue of where did it come from!

    T.

  2. #2
    Junior Member
    Join Date
    Apr 2016
    Posts
    4
    I have been able to identify where it comes from. In fact, I have put a HAproxy in front of the 2 servers in order to manage failover between the 2 livestatus backends. The problem occurs when I enable the HAproxy option "tcp-check connect" which, as its name implies, does a TCP test connection on the port 50000. If I omit this option, HAproxy considers the 2 livestatus backends to be UP and load balances the request between them, which doesn't work because only one of them is started in the Shinken HA. The thing is that this same configuration worked until last friday.

    Have anyone already managed to get a highly-available Livestatus setup with HAproxy?

    T.

Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •