Incorrect stanza in haproxy.cfg after the redeployment of the related application

[Przemyslaw Lal]

Charm revision: 66 (latest/stable)
Series: focal

charm-graylog was related to haproxy:
graylog:website <-> graylog-haproxy:reverseproxy

with SSL cert and key configured on haproxy.

After the redeployment of 3 graylog units (remove-unit x3, add-unit x3), haproxy charm started generating an incorrect haproxy.cfg.

No other configs or relations have been changed in the meanwhile.

Correct haproxy.cfg before (redacted confidential information): [0]
Incorrect haproxy.cfg after (redacted confidential information): [1]
juju show-unit output: [2]

Things that we tried:
- re-creating the relation between haproxy and graylog - no luck
- wiping peer relation data on graylog-haproxy (where this bad 82 port config shows up) - no luck
- removed and re-added ssl_{cert,key} config in haproxy - no change at all
- remove unit, full redeployment of the VM, re-add unit - ended up with the same state (port 82)
- removed application, full redeployment of the VM, re-added the application - no luck, port 82 still, port 443 missing

[0] https://pastebin.ubuntu.com/p/Th4DqNTdys/
[1] https://pastebin.ubuntu.com/p/rMgPrJFQRD/
[2] https://pastebin.ubuntu.com/p/4QFBGP8H5t/

[Przemyslaw Lal (przemeklal)]

As a workaround, I just placed the correct config and locked it with `chattr +i`. The unit is blocked in the `maintenance idle` state but at least the app is reachable.

[Seyeong Kim (seyeongkim)]

I freshly deployed bundle [1]

initially it has 82 and 80 port

and doesn't have 443 for ssl

I'm analyzing code but need any hints to analyze further.

Thanks.

[1] https://pastebin.canonical.com/p/Qb5bYVznH3/

[Seyeong Kim (seyeongkim)]

This is solved when I set below configuration.

https://pastebin.ubuntu.com/p/Ct8nkXs2rT/