|
> I haven't heard anyone talking about mon for ages. Most people
> who are checking their servers are using ldirectord, which
> comes with the ipvs patch kit and is part of the Linux-HA project
> and part of Ultramonkey
>
> http://ultramonkey.sourceforge.net/
>
> ldirectord is an http.alert
>
> I don't know what the technical reasons for changing to ldirectord
> is. I'm sure it's a lot simpler to keep track of one service than
> all of them
Well, because my Linux director has to load balance quite a few things
(HTTP, HTTPS, FTP, NAMED (yeah yeah... not RFC), SMTP, POP3, IMAP, ....) mon
works out alright. I wrote my own monitor which checks every service and
returns a global yea or nay, and I wrote my own alert which brings the 1024
virtual servers up and down per each real-server as necessary. And the
configuration for all of this is kept in a couple of lines of something
which resembles a redhat ifcfg file...
Really a lot of my setup has been pretty unconventional. For example, rather
than using heartbeat, I wrote my own "etherbeat" which handles the dual
redundant LVS's.
mon+etherbeat for me works out pretty nicely -- like clockwork. I can bring
machines down and up at random with pretty good confidence that service will
be maintained. If there is an outage, it won't be any longer than about 61
seconds.
Well -- if anybody DOES decide to use mon like li'l ol' me, it has that bug
where startupalerts aren't actually upalerts. Just fix the script so that it
checks both flags (i.e. & ( $FL1 | $FL2 ) ) and you'll be fine.
:) All the best --
Ted
|
