Haproxy ingress controller 3.1.0 with haproxy 3.1.3 breaking for me

[cxfcxf]

We are running them as external mode with haproxy ingress 3.0.2 and haproxy 3.0.8 no issue

when i start to testing Harpoxy ingress controller 3.1.0 with Haproxy 3.1.3, haproxy ingress controller failed to start

the log point to a panic to start so i dig along and find the issue is haproxy can not start.

# /app/haproxy/bin/haproxy -S /var/haproxy-ingress/run/haproxy-master.sock,level,admin -f /var/haproxy-ingress/etc/haproxy.cfg
[NOTICE]   (573937) : haproxy version is 3.1.3-929bedf
[NOTICE]   (573937) : path to executable is /app/mgni-haproxy/bin/mgni-haproxy
[ALERT]    (573937) : a master CLI socket was defined, but master-worker mode (-W) is not enabled.

when i add -W in to the command

# /app/haproxy/bin/haproxy -W -S /var/haproxy-ingress/run/haproxy-master.sock,level,admin -f /var/haproxy-ingress/etc/haproxy.cfg
[NOTICE]   (573973) : Initializing new worker (573974)
[NOTICE]   (573974) : haproxy version is 3.1.3-929bedf
[NOTICE]   (573974) : path to executable is /app/mgni-haproxy/bin/mgni-haproxy
[WARNING]  (573974) : frontend 'healthz' has no 'bind' directive. Please declare it as a backend if this was intended.
[WARNING]  (573974) : frontend 'stats' has no 'bind' directive. Please declare it as a backend if this was intended.

it is able to start

and when i check back in haproxy 3.0.8, it seems it doesnt require -W to start

]# /app/haproxy/bin/haproxy -S /var/haproxy-ingress/run/haproxy-master.sock,level,admin -f /var/haproxy-ingress/etc/haproxy.cfg
[NOTICE]   (2559947) : New worker (2559948) forked

i did a brief grep on the code base of ingress controller, i dont see -W is in anywhere

is this a new feature? or the 3.1.0 ingress controller doesnt work with haproxy 3.1.3?

or i m missing some config that need to be enabled?

[ivanmatmati]

Hi @cxfcxf , Thanks for reporting, we've spotted this issue and this is fixed in master branch. It will be available in the next release. You can test it with nightly build by changing your docker image version to "nightly".

[cxfcxf]

Hi, i think the current master still breaks, the panic is from native-client v6 6.0.7 v2v3 global_convert.go, i have to use the old commit before v3 API and force it to use client-native [email protected]

[hdurand0710]

Hi @cxfcxf ,

You can pull the latest master branch by pulling the latest docker with image version = nightly.
On master, we have switched to CRD v3 that are not yet released on v3.1.

There was an issue with the convert method (CRD v1 to v3 auto conversion) in client-native library.
This is fixed with commit efa42ce.

You can also if you want use a tool to convert your CRD v1 manifests to CRD v3 manifests and then use directly CRD v3 with master ( docker image nightly).
The tool is the ingress-controller binary that you can launch with --input-file and --output-file arguments
(https://gitlab.int.haproxy.com/haproxy-controller/kubernetes-ingress/-/blob/master/documentation/controller.md?ref_type=heads#--input-file)

"--input-file=/home/helene/tmp/convert/v1/backend.yaml",
"--output-file=/home/helene/tmp/convert/v3/backend.yaml"

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.