Hi Nathan, Actually, if both services are hosted on the same servers, my configuration should work. And from your configuration example, this is what I understood (servers have same name in your second configuration).
Anyway, happy to have helped you :) Baptiste On Wed, Apr 23, 2014 at 12:56 AM, Nathan Reilly <[email protected]> wrote: > Hi baptiste, > > I think the requirements have been confused, the services are separate, but > must fail together, if that makes sense, don't worry, it has taken me weeks > to get the requirements clear from the project, but traffic to port 80 comes > from an external source, and traffic to port 8080 is from an internal service > bus. In any case, the config works as expected (until the project comes up > with a new requirement), I've taken your advice on the tcp-check on board and > will amend the config. > > Thanks again for your help. > > Nathan > > Sent from my iPhone > >> On 23 Apr 2014, at 2:19 am, Baptiste <[email protected]> wrote: >> >> Hi Nathan, >> >> There might be a mistake in your configuration. >> Since you configured a port on the server directive line, then HAProxy >> will forward all the traffic to this port, whatever the incoming port >> was. >> I doubt this is what you want. >> >> Second, both backends are using sitck table from the other backend... >> So you don't have the same persistence for each service. >> Better having one stick table in a backend and point to it from the >> other backend. >> >> Third, the tcp-check does not work like this. >> >> Here is how I would do your configuration: >> >> frontend servername:80 >> bind servername:80 >> bind servername:8080 >> default_backend server >> >> backend server >> option tcp-check >> tcp-check connect port 80 >> tcp-check connect port 8080 >> stick-table type ip size 1 >> stick on dst >> server server-01-ext server01 check inter 5000 >> server server-02-ext server02 check inter 5000 >> >> >> Baptiste >> >> >>> On Tue, Apr 22, 2014 at 1:22 PM, Nathan Reilly <[email protected]> wrote: >>> Hi Baptiste, >>> >>> You put me in the right track, I had over-complicated my config, here is the >>> working config for anyone who's interested: >>> >>> frontend servername:80 >>> bind servername:80 >>> bind servername:8080 >>> acl int dst_port 8080 >>> acl ext dst_port 80 >>> use_backend server-int if int >>> use_backend server-ext if ext >>> default_backend server-ext >>> >>> backend server-ext >>> option tcp-check port 80 >>> option tcp-check port 8080 >>> stick-table type ip size 1 >>> stick match dst table server-int >>> stick on dst >>> server server-01-ext server01:80 check inter 5000 >>> server server-02-ext server02:80 check inter 5000 >>> >>> backend server-int >>> option tcp-check port 80 >>> option tcp-check port 8080 >>> stick-table type ip size 1 >>> stick match dst table server-ext >>> stick on dst >>> server server-01-int server01:8080 check inter 5000 >>> server server-02-int server02:8080 check inter 5000 >>> >>> Thanks for your help, it's truly appreciated. >>> >>> Cheers, >>> Nathan >>> >>> >>>> On Mon, Apr 21, 2014 at 2:07 PM, Baptiste <[email protected]> wrote: >>>> >>>> Hi Nathan, >>>> >>>> Why not simply using a single frontend with two binds pointing to one >>>> backend? >>>> It would do the trick. >>>> Concerning checking, you can use the tcp-check feature to ensure both >>>> ports are available on each server. >>>> >>>> Baptiste >>>> >>>> On Mon, Apr 21, 2014 at 1:20 PM, Nathan Reilly <[email protected]> >>>> wrote: >>>>> Hello all, >>>>> >>>>> I have a situation that I have been trying to get working for days, yet >>>>> can't seem to get there. There are two sets of requirements, I can get >>>>> either requirements working in isolation, but not together. >>>>> First requirement is I need an active/passive failover pair, and in the >>>>> case >>>>> of failure of one server, connections must stick to the available server >>>>> until it fails, so if connections are initially to server1, then it >>>>> fails, >>>>> all connections go to server2, even when server1 comes back online, >>>>> server2 >>>>> must keep servicing requests until it fails, and so on. >>>>> >>>>> >>>>> Second requirement is that I have two backend servers, each with two >>>>> 'services'. These services are linked, so if either service goes down, >>>>> the >>>>> entire server must be marked as down. >>>>> >>>>> See config so far below, >>>>> >>>>> frontend ext servername:80 >>>>> acl server-01-down nbsrv(server-01) lt 2 >>>>> acl server-02-down nbsrv(server-02) lt 2 >>>>> use_backend server-ext-01 if server-02-down >>>>> use_backend server-ext-02 if server-01-down >>>>> default_backend server-ext-01 >>>>> >>>>> frontend int servername:8080 >>>>> stick-table type ip size 1 expire 5d >>>>> stick on dst >>>>> acl server-01-down nbsrv(server-01) lt 2 >>>>> acl server-02-down nbsrv(server-02) lt 2 >>>>> use_backend server-esb-01 if server-02-down >>>>> use_backend server-esb-02 if server-01-down >>>>> default_backend server-esb-01 >>>>> >>>>> backend server-01 >>>>> server server-01-ext server01-ext:80 check inter 5000 >>>>> server server-02-int server01-int:8080 check inter 5000 >>>>> >>>>> backend server-02 >>>>> server server-02-ext server02-ext:80 check inter 5000 >>>>> server server-02-int server02-int:8080 check inter 5000 >>>>> >>>>> backend server-ext-01 >>>>> stick-table type ip size 1 expire 5d >>>>> stick on dst >>>>> server server-01-ext server01-ext:80 check inter 5000 >>>>> >>>>> backend server-ext-02 >>>>> stick-table type ip size 1 expire 5d >>>>> stick on dst >>>>> server server-02-ext server02-ext:80 check inter 5000 >>>>> >>>>> backend server-int-01 >>>>> stick-table type ip size 1 expire 5d >>>>> stick on dst >>>>> server server-01-int server01-int:8080 check inter 5000 >>>>> >>>>> backend server-int-02 >>>>> stick-table type ip size 1 expire 5d >>>>> stick on dst >>>>> server server-02-int server02-int:8080 check inter 5000 >>>>> >>>>> Any assistance or recommendations would be greatly appreciated, >>>>> >>>>> Thanks in advance, >>>>> >>>>> Nathan >>> >>>
