Socket permissions issues with kopano-server build 8.7.81.162
-
Hi,
looks like there is a bug ore something else between kopano-server version 8.7.81.151 and 8.7.81.162.
In version .151, /var/run/kopano/server.sock was created as kopano:kopano
Since version .162 the socket was created as root:root and without any permissionsThis breaks dagent, z-push, etc. for me … Anyone else with similar issues ?
chown kopano:kopano and chmod with right permissions fixes the issue … but the old behavior was definitely changed between this versions.
-
This is due to commit 5995dbf7cf8.
chown kopano:kopano and chmod with right permissions fixes the issue … but the old behavior was definitely changed between this versions.
Yes, but permissions will be reset on service-restart.
I’ve created a PR fixing the problems, but it will need further review by kopano.
As a first remedy, you can revert the above merge.
++umgfoin -
I got the same problems as the op - and many more:
changing owner and group of the socket does not solve all connection problems. I had to switch to network communication (localhost:236) for dagent, search, webapp…
Postfix and dagent didn’t communicate via socket either - had to switch to localhost:2003
All deamons now start in some kind of foreground and don’t fork into background. This blocks startup of the hole server. I guess it’s because they are now only compatible with systemd and no other init-systems anymore. Sadly.
kopano-search changed option for config-file from small letter “c” to capital “C”. Why this?
kopano-dagent doesn’t know option “-d” anymore…
I had to downgrade to 8.7.80 to get a working system.
-
@Fux said in Socket permissions issues with kopano-server build 8.7.81.162:
All deamons now start in some kind of foreground and don’t fork into background.
As was documented in
RELNOTES.txt
.I guess it’s because they are now only compatible with systemd and no other init-systems anymore.
You are incorrect. Even non-systemd systems have a way to background processes. The most trivial being the
&
character, which everyone using sh (a popular implementation of some init systems) should know about already.kopano-dagent doesn’t know option “-d” anymore…
It had the same function as the absence of -F, so was practically useless. As there is no more self-backgrounding, -d has no meaning and had to be removed.
-
I have got the same Problem. How to fix it? I used chmod to get it working until the next service restart.
-
hi,
prio.sock is also root:root
does this also needs to be kopano:kopano?
-
@jengelh said in Socket permissions issues with kopano-server build 8.7.81.162:
The most trivial being the
&
character, which everyone using sh (a popular implementation of some init systems) should know about already.The problem with this (and with the background-option of start-stop-daemon) is: you don’t see any (error) outputs of the started process.
Just checked some init-scripts of other popular daemons (like postfix, samba, bind…): no one uses the "&"or the background-option of start-stop-daemon. I think there are still some good reasons for daemons to self fork into background (i.e. for having the return code in the start script and working with it).
-
webapp does not function either. after login it cannot connect to kopano core.
[Sun Jun 16 21:46:08.595819 2019] [:error] [pid 255] [client 192.168.2.6:50436] PHP Fatal error: require_once(): Failed opening required '/usr/share/kopano-webapp/config.php' (include_path=':/usr/share/kopano/php') in /usr/share/kopano-webapp/server/includes/bootstrap.php on line 26
-
@Fux said in Socket permissions issues with kopano-server build 8.7.81.162:
The problem with this (and with the background-option of start-stop-daemon) is: you don’t see any (error) outputs of the started process.
That does not seem so different from systemd, where
systemctl start xyz
generally also does not output any lines from the process (just the unit as a meta entity) and you need to go to journalctl or a logfile.I think there are still some good reasons for daemons to self fork into background (i.e. for having the return code in the start script and working with it).
Except that such was never correctly implemented. kopano-server could fail after fork, but before reaching the mainloop, such as when the database schema upgrade went sour. Or when sockets/ports are already occupied. Or when mysql is down. Or your config file did not have the right incantations.
You are just making a panic for no reason.
-
@jengelh i think that that “panic” part is totaly right.
The problem is the errors trow you off and you searching to fix this first, goes in the wrong direction.Every logrotate the rights are on socket files are rest to root:root
Tested this also with : .164 today on debian.
I saw the fix, so i’ll wait for the fix and reset the right every restart/reboot/start.If one is using my get-community-packages script, with the enabled backup function, then it’s easy to revert back to .150.
remove the “amd64” folder with the latested community packages.
goto you backup folder, move the good version back and rename it back to amd64.apt update
and install it again.The remove and install blocks i used, on my Debian 9 server.
apt-get remove kopano-backup kopano-client kopano-common kopano-contacts kopano-dagent kopano-gateway kopano-ical kopano-lang kopano-monitor kopano-python-utils kopano-search kopano-server kopano-server-packages \ kopano-spamd kopano-spooler kopano-utils kopano-webapp kopano-webapp-plugin-desktopnotifications kopano-webapp-plugin-filepreviewer kopano-webapp-plugin-files kopano-webapp-plugin-filesbackend-smb \ kopano-webapp-plugin-folderwidgets kopano-webapp-plugin-htmleditor-minimal-tinymce kopano-webapp-plugin-htmleditor-quill kopano-webapp-plugin-intranet kopano-webapp-plugin-mdm \ kopano-webapp-plugin-pimfolder kopano-webapp-plugin-quickitems kopano-webapp-plugin-smime kopano-webapp-plugin-spell kopano-webapp-plugin-spell-en kopano-webapp-plugin-spell-nl \ kopano-webapp-plugin-titlecounter php7-mapi python3-kopano python3-kopano-search python3-kopano-utils python3-mapi z-push-backend-kopano z-push-kopano \ libkcarchiver0 libkcarchivercore0 libkcfreebusy0 libkcicalmapi0 libkcindex0 libkcinetmapi0 libkcpyplug0 libkcrosie0 libkcserver0 libkcsoap0 libkcutil0 libmapi1
( ! Note, if more is more is removed, just capture what is removed, so you can reinstall it again ).
apt-get install kopano-backup kopano-client kopano-common kopano-contacts kopano-dagent kopano-gateway kopano-ical kopano-lang kopano-monitor kopano-python-utils kopano-search kopano-server kopano-server-packages \ kopano-spamd kopano-spooler kopano-utils kopano-webapp kopano-webapp-plugin-desktopnotifications kopano-webapp-plugin-filepreviewer kopano-webapp-plugin-files kopano-webapp-plugin-filesbackend-smb \ kopano-webapp-plugin-folderwidgets kopano-webapp-plugin-htmleditor-minimal-tinymce kopano-webapp-plugin-htmleditor-quill kopano-webapp-plugin-intranet kopano-webapp-plugin-mdm \ kopano-webapp-plugin-pimfolder kopano-webapp-plugin-quickitems kopano-webapp-plugin-smime kopano-webapp-plugin-spell kopano-webapp-plugin-spell-en kopano-webapp-plugin-spell-nl \ kopano-webapp-plugin-titlecounter php7-mapi python3-kopano python3-kopano-search python3-kopano-utils python3-mapi z-push-backend-kopano z-push-kopano \ libkcarchiver0 libkcarchivercore0 libkcfreebusy0 libkcicalmapi0 libkcindex0 libkcinetmapi0 libkcpyplug0 libkcrosie0 libkcserver0 libkcsoap0 libkcutil0 libmapi1
-
@thctlo said in Socket permissions issues with kopano-server build 8.7.81.162:
@jengelh i think that that “panic” part is totaly right.
The problem is the errors trow you off and you searching to fix this first, goes in the wrong direction.Panicing over a lost -F option is the silliest thing I’ve heard.
-
I am running into the same issue as explained here on this thread. Yesterday I upgraded to 8.7.81.164 on Debian Stretch and the server socket was not accessible from the other Kopano services. When I
cd /var/run/kopano && chown kopano. server.sock prio.sock
after (!) and kopano-server is already running it seems that the services work fine because spooler and dagent are not reporting the usual gsoap() connection failures. But I still cannot logon through WebApp, I get the error
Cannot connect to Kopano Core
I can run
kopano-admin -l
successfully though. Running
z-push-top
also doesn’t show any activity, no device/user shown here although configuration is double-checked. Smartphones cannot sync any more. The kopano logs don’t show any related warnings or errors. Any clues where to look at?
-
anyone ? my whole Kopano mail system is down, I cannot make use of it. I really don’t want to start mess around with downgrading. Any help or hint appreciated.
EDIT: after lot of trial & error I did find following workaround. For anyone experiencing the same issue, here’s some information:
- First of all I modified “/etc/kopano/webapp/config.php” and changed from socket to https.
before ==> define(“DEFAULT_SERVER”, “file:///var/run/kopano/server.sock”);
after ==> define(“DEFAULT_SERVER”, “https://localhost:237/”);
After this change Kopano-WebApp login works successfully ! That indicated that the issue indeed lies in server socket. So …
- … I reverted this modification to the original state (using the server socket) and then I checked the file permissions on the server socket file “/var/run/kopano/server.sock” and “/var/run/kopano/prio.sock” As suggested initially in this thread I just changed owner+group to “kopano”. The files looked like that:
srwxr-xr-x 1 kopano kopano 0 Jun 25 14:35 /var/run/kopano/prio.sock srwxr-xr-x 1 kopano kopano 0 Jun 25 14:35 /var/run/kopano/server.sock
But this was not enough as WebApp requires write access to this socket file and that’s why all the thing didn’t work. After adding the write permission to others it works like a charm. The files now look like that:
srwxr-xrwx 1 kopano kopano 0 Jun 25 14:35 /var/run/kopano/prio.sock srwxr-xrwx 1 kopano kopano 0 Jun 25 14:35 /var/run/kopano/server.sock
This is just a workaround and certainly Kopano will provide a patch soon.
- First of all I modified “/etc/kopano/webapp/config.php” and changed from socket to https.
-
@micro said in Socket permissions issues with kopano-server build 8.7.81.162:
anyone ? my whole Kopano mail system is down, I cannot make use of it. I really don’t want to start mess around with downgrading. Any help or hint appreciated
Your signature still lists one of our supported versions but this issue only occurs in our nightly downloads. I am wondering why you are now using these. Did you stop with your subscription?
For production deployments it’s always recommended to use our supported packages. And if an issue occurs you have direct contact with our support.
Part of these nightly downloads (if your really want to use them instead) is that you should be doing extensive testing before upgrading your installation as breakage will occurs in these packages at any time.
-
Felix no. My signature is still valid and is showing the configuration of our company setup. My posting here refers to the community edition which I use for private use at home. That’s why I was calm even when my mail server was down for >24h ;) but I am happy that I have access to my mails/calendar/contacts again until this get fixed with a patch.
-
Hai,
Small update on this.
I’ve updated today from 8.7.81.151.4eaf3e0cf-0+249.1 to 8.7.81.203.1889fded5-0 and today i’m happy to say, i have had zero problems with this update.
i have only one suggestion.
libgsoap-kopano-2.8.84 was not removed after the install of libgsoap-kopano-2.8.86Just an suggestion here.
If you change the packaging numbers then this is also nicely upgraded.
for example from libgsoap-kopano-2.8.86 to libgsoap-kopano-2.8-86~0+1.1
Or create a meta package which does the remove of the older version. -
If you change the packaging numbers then this is also nicely upgraded.
for example from libgsoap-kopano-2.8.86 to libgsoap-kopano-2.8-86~0+1.1
Or create a meta package which does the remove of the older version.(I’ll read that as 2.8.84 -> 2.8.86)
Packaging systems have a hardremoval instruction (Obsoletes
in rpm,Replaces
in deb), but they are prone to cascading effects. Just think of all the breakage if qt5 were marked as replacing qt4, it would either lead to removal of all qt4-requiring apps, or block the installation of qt5.A weakremoval instruction is the way, but, as far as I know, only openSUSE has such.
-
Hi,
sorry for using this old thread but is this problem resolved?
I am on kopano-server 8.7.84.133.fdc228634 and have to change the socket owner after every reboot of the server.Thanks.
Greetings
Sebastian -
Hi @sbauhaus,
the issue came back a while ago, but has since been resolved.
-
Fix was included in kopanocore-8.7.84-140-g5f74297a5 .