this is a horrible solution to the problem. the DNS requests are essentially failing, but failing very quickly rather than timing out.
iâm trying to work on getting the docker embedded DNS proxy working. I suspect iâll need to change the IP of the docker bridge to 192.168.100.1 in order to get this working as intended. this will probably work for people (such as myself) that have spun up a VM solely for the purpose of running collabora, but would not be a solution for those running on an existing docker host.
iâm still unsure how the collabora container is meant to have its DNS configured, but that might be something for the creator of it to answer.
I just tested this, but it doesnât change the delay⌠Is CODE using the /opt/lool/systemplate/etc/resolv.conf file?
no - this is the problem. docker ensures that the /etc/resolv.conf is provided via a special mount (so the contents of this file inside the image are not actually used). the --dns-ip option allows you to override the default action which is to simply use the contents of the host /etc/resolv.conf. normally this option isnât required unless you have a nameserver bound to the loopback adapter of the host or similar weird/special config.
from what I understand the /opt/lool/systemplate/ is like a chroot jail used by the environment of the document editor, so itâs the collabora subprocess which sees this as the resolv.conf to use. this comes factory prepared from the person who packaged this image as 192.168.100.1 (and it was 192.168.65.1 in the previous release).
@guddl 's response above seems like a good solution to this issue - assuming that the chroot jail or whatever collabora uses is happy with a symbolic link.
iâm guessing that since the /etc/resolv.conf is mounted in by docker that it would probably be recommended to rename the file and create the symbolic link in the /start-libreoffice.sh or /etc/rc.local or something simiilar. iâm not sure how solid a symbolic link to a mount would be.
ok. iâve got this working ok now. i removed the /opt/look/systemplate/etc/resolv.conf file and modified the beginning of /start-libreoffice.sh so it reads like:
#!/bin/sh
# Fix lool resolv.conf problem (wizdude)
ln -s /etc/resolv.conf /opt/lool/systemplate/etc/resolv.conf
# Generate new SSL certificate instead of using the default
iâve run some tcpdumpâs and Iâm not seeing any DNS query related problems anymore. Collabora Online is working great - nice and quick and no more delays.
i think the creator of this docker should be notified of this issue. are they here reading this thread? or do I need to report this back somewhere else?
@jospoortvliet can you point this issue and the solution to the team which created the docker container? Maybe it would be good to explain a bit more how the docker container is created, or share the Dockerfile, so more people can have a look at it/improve it? Thanks!
thank you for your solution but it does not rule my house.
I have the resolv.conf I delete and I do well the symbolic link.
by cons must restart after you?
For information on symbolic link is removed if I stop the docker, you must then redo the symbolic link.
ps: I have two for a certificate for a nextcloud and collaborated.
You have an idea?
If I understand you correctly, you mean you canât find a way to open the Collabore office suite from inside Nextcloud?
For this you have to install an app: youâll need to enable experimental apps (not sure about this however) and then go to the Productivy section and enable the âCollabora Onlineâ app.