On Fri, Feb 5, 2021 at 4:14 PM Willy Tarreau <[email protected]> wrote:
> HAProxy 2.4-dev7 was released on 2021/02/05. It added 153 new commits
> after version 2.4-dev6.
> - Some significant lifting was done to the Prometheus exporter, including
> new fields, better descriptions and some filtering. I've seen quite a
> bunch pass in front of me but do not well understand what it does, all
> that interests me is that some users are happy with these changes so I
> guess they were long awaited :-)
about that, please note two breaking changes:
- objects' status are no longer a gauge value which you need to
translate manually; instead the state is a label with a proper string
value. The value of the metric simply informs whether the state is
active or not.
so we went from:
haproxy_server_status{proxy="be_foo",server="srv0"} 2
(which meant MAINT)
to:
haproxy_server_status{proxy="be_foo",server="srv0",state="DOWN"} 0
haproxy_server_status{proxy="be_foo",server="srv0",state="UP"} 0
haproxy_server_status{proxy="be_foo",server="srv0",state="MAINT"} 1
haproxy_server_status{proxy="be_foo",server="srv0",state="DRAIN"} 0
haproxy_server_status{proxy="be_foo",server="srv0",state="NOLB"} 0
This change is valid for frontend, backend, server with different label values.
- similar change with health checks where we put a state label:
haproxy_server_check_status{proxy="be_foo",server="srv0",state="HANA"} 0
haproxy_server_check_status{proxy="be_foo",server="srv0",state="SOCKERR"} 0
haproxy_server_check_status{proxy="be_foo",server="srv0",state="L4OK"} 0
haproxy_server_check_status{proxy="be_foo",server="srv0",state="L4TOUT"} 0
haproxy_server_check_status{proxy="be_foo",server="srv0",state="L4CON"} 1
haproxy_server_check_status{proxy="be_foo",server="srv0",state="L6OK"} 0
haproxy_server_check_status{proxy="be_foo",server="srv0",state="L6TOUT"} 0
haproxy_server_check_status{proxy="be_foo",server="srv0",state="L6RSP"} 0
haproxy_server_check_status{proxy="be_foo",server="srv0",state="L7TOUT"} 0
haproxy_server_check_status{proxy="be_foo",server="srv0",state="L7RSP"} 0
haproxy_server_check_status{proxy="be_foo",server="srv0",state="L7OK"} 0
haproxy_server_check_status{proxy="be_foo",server="srv0",state="L7OKC"} 0
haproxy_server_check_status{proxy="be_foo",server="srv0",state="L7STS"} 0
haproxy_server_check_status{proxy="be_foo",server="srv0",state="PROCERR"} 0
haproxy_server_check_status{proxy="be_foo",server="srv0",state="PROCTOUT"} 0
haproxy_server_check_status{proxy="be_foo",server="srv0",state="PROCOK"} 0
It means:
* a lot more metrics for large setup (but you can still filter as
explained in the doc)
* easier use on prometheus side: you will be able to group per state
very easily now.
Generally speaking I'm very interested in feedback regarding this
change. This was motivated by the usage I saw in several companies,
where people were struggling making use of those metrics.
Willy: it is probably a wise idea to keep it for the 2.4 final release
notes, some people might want to know that during their update; a lot
of people have their production alerts on those metrics.
Thanks,
--
William
