haproxy fails at startup when using server name instead of IP

[Bill Waggoner]

This is similar to #689734 I believe.

When starting haproxy using a DNS name on the 'server' line haproxy fails to start, giving the message:

```
May 05 19:09:40 hyrule systemd[1]: Starting HAProxy Load Balancer...
May 05 19:09:40 hyrule haproxy[1146]: [ALERT] 124/190940 (1146) : parsing [/etc/haproxy/haproxy.cfg:157] : 'server scanmon' :
May 05 19:09:40 hyrule haproxy[1146]: [ALERT] 124/190940 (1146) : Failed to initialize server(s) addr.
May 05 19:09:40 hyrule systemd[1]: haproxy.service: Control process exited, code=exited status=1
May 05 19:09:40 hyrule systemd[1]: haproxy.service: Failed with result 'exit-code'.
May 05 19:09:40 hyrule systemd[1]: Failed to start HAProxy Load Balancer.
```

In this case the server statement was:

` server scanmon myservername.mydomain.org:8000`

Changing it to use the IP address corrected the problem.

I believe there is a missing dependency for DNS in the unit file.

--Info:
Description: Ubuntu 18.04 LTS
Release: 18.04

haproxy:
  Installed: 1.8.8-1
  Candidate: 1.8.8-1
  Version table:
 *** 1.8.8-1 500
        500 http://archive.ubuntu.com/ubuntu bionic/main amd64 Packages
        100 /var/lib/dpkg/status

[Andreas Hasenack (ahasenack)]

Thanks for filing this bug in Ubuntu.

I tried a simple case of bringing up two lxd containers, one with haproxy as the frontend, another one with just apache as the backend, and this config file on the frontend:

(defaults from the package go here)
frontend mycontainer
 bind *:9090
 mode http
 default_backend myapache

backend myapache
 mode http
 balance roundrobin
 server apache-1 bionic-backend.lxd:80 check

I then rebooted this frontend and it picked up the bionic-backend.lxd ip just fine via dns.

This sounds indeed like some sort of race condition during boot. It happens only after you reboot a machine? After the boot, if you login and issue a "systemctl restart haproxy", it then works?

Could you share some details about the networking setup of this machine? For example:
cat /etc/network/interfaces /etc/network/interfaces.d/*
cat /etc/netplan/*.yaml
cat /etc/hosts
cat /etc/resolv.conf

Also, does this machine have a wired network connection, or does it depend on wifi and someone logging in on gnome to activate network-manager and its wifi connection?

[Launchpad Janitor (janitor)]

[Expired for haproxy (Ubuntu) because there has been no activity for 60 days.]

[Gustin Johnson (gustin-meganerd)]

I can confirm that I also have this issue. If I ssh into the host and manually issue "systemctl restart haproxy" haproxy runs just fine. I am also using hostnames that resolve via DNS and not /etc/hosts.

This is a headless VM on wired network (KVM bridged network). The IP is assigned via DHCP.

There is nothing in /etc/netplan

/etc/hosts only has the default localhost entries and some IPv6 lines at the end (this network does not have IPv6, only the link local stuff).

/etc/resolv.conf points to the localhost and thus I used:
 sudo systemd-resolve --status
to get the actual DNS server(s), which are correct and functioning as expected.

Any other ideas?

[Andreas Hasenack (ahasenack)]

Do you have fully qualified hostnames in the haproxy config? Or a bare name?

[Launchpad Janitor (janitor)]

[Expired for haproxy (Ubuntu) because there has been no activity for 60 days.]