Trac can't log in when ipv6 is enabled on nginx

[robe]

This is to log an issue that happens whenever nginx container is rebooted.

I haven't found a good work around for this.

There are two issues here:

1) Can't get it so on restart nginx doesn't start with ipv6 enabled. So after reboot I need to do this in the osgeo7 nginx container

This I have in the sysctl config to disable ipv6 so just reloading the contif works

 sysctl -p

2) Trac hates ipv6 for some reason (perhaps the newer trac won't). Gitea works fine with ipv6 enabled, but trac does not.

This last time even though nginx wasn't registering an ipv6 address (or I thought it wasn't), for some reason tracsvn still did not work.

Not sure which sequence of steps did it:

1) Did service reboot on trac, but it wouldn't start up so I had to manually start it with

lxc start tracsvn

It complained the IP was in use, so I did it again and second time it came up (might have just been on it's way up.

But then when I checked the containers with

lxc list

I could see the nginx now had an ipv6 (maybe I was imaging things before that it didn't).

So I logged into nginx

lxc exec nginx

sysctl -p
systemctl nginx reload

and that seemed to fix tracsvn session issue

[strk]

I just had to run sysctl -p once again, as trac was again broken this morning

[robe]

Strange it happened again. I didn't reboot nginx container and it usually only happens on reboot.

[robe]

I think this will become a non-issue once I upgrade trac to stretch.

[robe]

I think I fixed this on nginx finally, but I also upgraded trac to debian 10, so wouldn't be an issue anymore anyway. Will be for older not upgrade vms which I hope to get rid of early next year (and move all stuff off of them that is important)

[robe]

This is still an issue. Perhaps upgrading trac will fix.

[strk]

we upgraded Trac to 1.6, did you see this problem since then ?

[robe]

I'll have to turn back on ipv6 on the nginx container to confirm. I'll check later tonight.