Op do 14 dec. 2017 om 22:12 schreef PiBa-NL <[email protected]> > Hi Johan, > > Op 14-12-2017 om 9:35 schreef Johan Hendriks: > > Op 13/12/2017 om 20:48 schreef PiBa-NL: > >> Hi Johan, > >> > >> Op 13-12-2017 om 17:31 schreef Johan Hendriks: > >>> When i use the show stat command I get different results? > >> Just a guess, are you using?: nbproc > 1 > > Yes indeed i use nbproc > 1, There are no old processes running on the > > machine. > > This is my global config > > > > global > > nbproc 8 > > > > # stats bind-process 1 > Oke so 'uncommenting' this 'fixed' the issue that you set a server to > maintenance state in the first haproxy process, and you are viewing the > 'show stat' of that first process.. > > stats socket /var/run/haproxy/info.sock mode 600 level admin > > log 127.0.0.1:514 local1 notice > > > > defaults > > log global > > mode http > > > > listen stats > > bind :8077 > perhaps you would like to bind this port also to the first process.? so > webgui will show the state also of the first PID.? > > mode http > > stats enable > > stats scope web-backend > > stats uri / > > stats realm Haproxy\ Statistics > > > > > > I just saw that I commented out stats bind-process 1 > > I did uncomment it and it seems to be working now for the echo "show > > stat" | socat /var/run/haproxy/info.sock stdio part, it is not switching > > to UP sometimes, but the web GUI is showing the server is UP. It did > > show it in maintenance mode first but after a while it shows UP instead > > of maintenance. > > > > > > Thank you very much for your time. > > regards > > Johan > > > However.! > > If youve understood how nbproc works, it just means haproxy gets started > several times simultaneously and each keeps its own state. That means > you need to configure 8 separate stats sockets, and 8 stats websites, > and when setting a server to maintenance you need to make sure all 8 > process are set to this new state. > > In your current config visit the stats page and press ctrl+F5 several > times in the browser(firefox?), youl see the pid on the stats page > change, and also if your 'lucky' you might see the server sometimes is > in maintenance mode when you by chance are requesting the stats page of > the same process / PID you changed that state on.. > > Anyhow.. i don't have a example ready myself but have seen some pass on > the mailinglist how to configure multiple sockets for this.. Hope above > comments gets you going in the right idea :) > > Regards, > > PiBa-NL > Thanks, I found out that each proces has its own metrics and can be controlled seperately using a unix socket. I have configured haproxy now in a way every proces has its own .sock file which i can control, and read metrics from. Reading the web I learned a lot about the proces. Also the option to chain a frontend to a proces is something i did not know, which makes separating metrics from different frontends simpler.
Thank you for pushing me in the right direction. I will reply to this thread with my config. So other people can use that as an example. I am on a mobile device now, so no acces to the box. Regards Johan
