All over sudden dovecot thinks FS is read only
Hi list
I have to dovecot instances running which host my mailboxes on Centos7
dovecot.x86_64 1:2.2.26.0-2.el7.centos installed dovecot-debuginfo.x86_64 1:2.2.26.0-2.el7.centos installed dovecot-mysql.x86_64 1:2.2.26.0-2.el7.centos installed dovecot-pigeonhole.x86_64 1:2.2.26.0-2.el7.centos installed
all over sudden both dovecot "think" that the mailbox filesystem is read only and fill the logs with
Oct 28 11:39:50 mbox1 dovecot: lmtp(tobster@brain-force.ch): Error: ssY/MGZQ9FnoLgAAC4IaJw:2: sieve: failed to open logfile (LOGGING TO STDERR): open(/home/vmail/brain-force.ch/tobster/.dovecot.sieve.log) failed: Read-only file system
But a test at cli shows that I can write to the location
$ echo "test" >/home/vmail/brain-force.ch/tobster/.dovecot.test $ cat /home/vmail/brain-force.ch/tobster/.dovecot.test test $ echo "test1" >/home/vmail/brain-force.ch/tobster/.dovecot.test $ cat /home/vmail/brain-force.ch/tobster/.dovecot.test test1
Mounts show that the FS should be rw /dev/vda2 / ext4 rw,relatime,data=ordered 0 0
I restarted dovecot several time already. As well I restarted the physical host on which the dovecot vms reside.
Filesystems are not full and are not running out of inodes.
Does anyone have an idea why dovecot suddenly behaves like that?
Thanks a lot for any hint Regards
tobi
:2017-10-28T11:51:Tobi:
Oct 28 11:39:50 mbox1 dovecot: lmtp(tobster@brain-force.ch): Error: ssY/MGZQ9FnoLgAAC4IaJw:2: sieve: failed to open logfile (LOGGING TO STDERR): open(/home/vmail/brain-force.ch/tobster/.dovecot.sieve.log) failed: Read-only file system
But a test at cli shows that I can write to the location
$ echo "test" >/home/vmail/brain-force.ch/tobster/.dovecot.test $ cat /home/vmail/brain-force.ch/tobster/.dovecot.test test $ echo "test1" >/home/vmail/brain-force.ch/tobster/.dovecot.test $ cat /home/vmail/brain-force.ch/tobster/.dovecot.test test1
Mounts show that the FS should be rw /dev/vda2 / ext4 rw,relatime,data=ordered 0 0
Check filesystem permissions on the location. The user dovecot is running under needs to be able to write there. Otherwise check dmesg as well for any "remount" messages.
-- Andraž 'ruskie' Levstik Geek/Hacker/Tinker
What use is magic if it can't save a unicorn?
On October 28, 2017 at 1:12 PM "Vlad K." <dovecot-ml@acheronmedia.hr> wrote:
On 2017-10-28 11:51, Tobi wrote:
Does anyone have an idea why dovecot suddenly behaves like that?
Since that's CentOS 7, I've heard SystemD tends to go mad (madder than usual) and mount the service with readonly paths...
-- Vlad K.
Check the systemd unit for ProtectHome, and make sure that is turned off.
Aki
Hello list
I saw the response from Vlad in my queue (as dovecot was broken there has been no delivery to mailbox) but after fix and queue flush Vlads message does not show up in my mailbox.
Check the systemd unit for ProtectHome, and make sure that is turned off.
Found these lines in unit file
# Enable this if your systemd is new enough to support it: ProtectSystem=full
and after commenting the 2nd one, dovceot could write again to the FS :-) Have no idea why this setting started to make problems just today :-)
Thanks so much for the fast help here
tobi
Am 28.10.2017 um 11:51 schrieb Tobi:
Hi list
I have to dovecot instances running which host my mailboxes on Centos7
dovecot.x86_64 1:2.2.26.0-2.el7.centos installed dovecot-debuginfo.x86_64 1:2.2.26.0-2.el7.centos installed dovecot-mysql.x86_64 1:2.2.26.0-2.el7.centos installed dovecot-pigeonhole.x86_64 1:2.2.26.0-2.el7.centos installed
all over sudden both dovecot "think" that the mailbox filesystem is read only and fill the logs with
Oct 28 11:39:50 mbox1 dovecot: lmtp(tobster@brain-force.ch): Error: ssY/MGZQ9FnoLgAAC4IaJw:2: sieve: failed to open logfile (LOGGING TO STDERR): open(/home/vmail/brain-force.ch/tobster/.dovecot.sieve.log) failed: Read-only file system
But a test at cli shows that I can write to the location
$ echo "test" >/home/vmail/brain-force.ch/tobster/.dovecot.test $ cat /home/vmail/brain-force.ch/tobster/.dovecot.test test $ echo "test1" >/home/vmail/brain-force.ch/tobster/.dovecot.test $ cat /home/vmail/brain-force.ch/tobster/.dovecot.test test1
Mounts show that the FS should be rw /dev/vda2 / ext4 rw,relatime,data=ordered 0 0
I restarted dovecot several time already. As well I restarted the physical host on which the dovecot vms reside.
Filesystems are not full and are not running out of inodes.
Does anyone have an idea why dovecot suddenly behaves like that?
Thanks a lot for any hint Regards
tobi
On October 28, 2017 at 1:41 PM Tobi <tobster@brain-force.ch> wrote:
Hello list
I saw the response from Vlad in my queue (as dovecot was broken there has been no delivery to mailbox) but after fix and queue flush Vlads message does not show up in my mailbox.
Check the systemd unit for ProtectHome, and make sure that is turned off.
Found these lines in unit file
# Enable this if your systemd is new enough to support it: ProtectSystem=full
I recommend you leave this ON and instead set
ProtectHome=false
To enable writable /home
Aki
Hi Aki
with the following in dovecot.service unit file
ProtectSystem=full ProtectHome=false
my logs still fill with RO warnings. Only if I remove the ProtectSystem (comment out) dovecot runs without RO errors for the home.
Cheers
tobi
Am 28.10.2017 um 12:47 schrieb Aki Tuomi:
On October 28, 2017 at 1:41 PM Tobi <tobster@brain-force.ch> wrote:
Hello list
I saw the response from Vlad in my queue (as dovecot was broken there has been no delivery to mailbox) but after fix and queue flush Vlads message does not show up in my mailbox.
Check the systemd unit for ProtectHome, and make sure that is turned off.
Found these lines in unit file
# Enable this if your systemd is new enough to support it: ProtectSystem=full
I recommend you leave this ON and instead set
ProtectHome=false
To enable writable /home
Aki
participants (4)
-
Aki Tuomi
-
Andraž 'ruskie' Levstik
-
Tobi
-
Vlad K.