- dovecot - dovecot.org

Sdbox/mdbox purge on Public folders
by luke+dovecot＠psy.io 15 Feb '26

15 Feb '26

Hi dovecot mailing list people, After having looked through the mailing list archives, tried various search engines & asked an LLM to hallucinate an answer all to no avail, I now pose the following question to you all: For a Dovecot 2.4 system that uses mail_driver = mdbox for regular mailboxes and also has a public namespace, how might I invoke the maintenance routine of `doveadm purge` against the m.1, m.2 ... m.N files under the mail_path location? Below is the snippet of what I have in the configuration files for the public namespace: namespace public { type = public separator = / prefix = Public/ mail_path = /var/spool/mail/Public mail_index_private_path = ~/public subscriptions = yes mailbox "communal" { auto = subscribe } } Since my dovecot instance is a self-hosted matter, I am in a position to move all mail data out of there, delete any left over storage files and change the mail_driver to Maildir, if that is the easiest way to manage the risk of dbox m file sprawl/growth. Best regards, Luke

2 1

Is there a lint utility for 2.4.x configuration?
by Steve Litt 14 Feb '26

14 Feb '26

Hi all, I use a local Dovecot IMAP server to store all my emails, which I access from either Claws-Mail or Evolution. On 7/7/2025 my email stopped working, a few hours of troubleshooting led to a Dovecot problem, a few more hours led to the 2.3 to 2.4 config file changes, and the next 3 days I went without email while trying to get a working 2.4. Finally I managed to downgrade back to 2.3.x and restore the config from backup, and didn't lose too many emails. I've waited 6 months for all the dust to settle, and now I'm going to try to make a test VM (Void Linux) with 2.4.x, using phony baloney data because I don't want to lose my email capabilities again. So my first question is this: Is there a lint utility for my configuration, so that I can cut down on experimenting with 50 different variables at once and reading conflicting documentation? Here's what I know so far about lint utilities: * I know doveconf -n and dovecot -n dump non-default settings, but I don't think they can detect syntax errors. * dovecot-sysreport doesn't seem to give me info about syntax errors. * None of the doveadm commands seem to check syntax. Thanks, SteveT Steve Litt http://444domains.com

2 3

My plan for a test Dovecot 2.4.2
by Steve Litt 14 Feb '26

14 Feb '26

Hi all, I need to thoroughly test Dovecot 2.4.2 on a Virtual Machine before upgrading my real machine. Please tell me whether you think this plan is good... I install Void Linux, my Daily Driver Distro, on a Qemu VM. I also install Dovecot and Claws-Mail. Then I copy part or all of my current (2.3.21.1 (d492236fa0)) maildir to the VM, and work until I can configure a Dovecot on the VM that serves IMAP only, SSL only (ssl = required on 2.3.x), such that Claws-Mail can read, search and delete messages on the VM just the same as on 2.3.21.1 on my daily driver. I'm using few or no Dovecot features other than using it as a Maildir driven IMAP server. Maildir is a must because I write to it from Procmail, and Procmail doesn't write every kind of mail storage. Does this sound like a good, safe way to test before upgrading? Thanks, SteveT Steve Litt http://444domains.com

1 0

Clean up of dovecot.index.pvt.log files
by Sad Clouds 14 Feb '26

14 Feb '26

Hello, I'm trying to understand how to clean up old dovecot.index.pvt.log files, which are used for per user private \Seen flags. Running "doveadm purge" does not seem to remove them. I have the following public namespace configured: namespace public { type = public separator = / prefix = Public/ inbox = no hidden = no mail_driver = mdbox mail_path = /data/mailstore/%{user|domain|lower}/mail_public/mdbox mail_index_private_path = ~/mdbox/Public.index.pvt # Enable user private \Seen flags list = yes subscriptions = no # Allow users in the 'admin' group to have complete control over the namespace acl group=admin { rights = lrwstipekxa } } The problem is - if the admin deletes a public folder, the old dovecot.index.pvt.log files forever remain in all of the users' home mailbox directories, like tombstones. I see two ways around it: 1. Attempt to store private index files in the original folder's mail path, so when the folder is deleted, all files underneath it are also deleted automatically: mail_path = %{mail_directory}/Public.index.pvt/%{user|domain|lower} However there appears to be no %{mail_directory} variable which would expand to mdbox/mailboxes/<folder_name>/dbox-Mails/ and even if there was one, with public folders, for every user subscribed to this folder, there would be a user-specific directory just for the private index file. This may not scale well with thousands of users. 2. Run a cron script to list all public folders and then compare ~/mdbox/Public.index.pvt directory of each user for stale private index files which refer to non-existent folders. This can probably be done efficiently with commands like find, awk and sed, but it feels a bit cumbersome and arcane. Can anyone suggest any other or better alternatives to solve this issue? Thanks.

1 0

shared mailbox on 2.3 - will I avoid concurrency (mailbox corruption) issues if both users are on the same server
by daniel.botting＠codethink.co.uk 13 Feb '26

13 Feb '26

Hi all, Currently we are running 2.3, with a plan to move to 2.4 later this year. Our current setup is two replicating servers, I have setup on my test servers a shared name space as this is a limited requirement at this time for an assistant to view their managers inbox. I had read here: https://doc.dovecot.org/2.3/configuration_manual/shared_mailboxes/cluster_s… For two or more servers Director is required because (taken from the docs): As mentioned in Director, you can’t have multiple servers accessing the same user at the same time or it will lead into trouble. This can become problematic with shared mailboxes, because two users who are sharing a folder may run in different servers. As this is a limited requirement at this time and we'll be migrating to 2.4 where Director has been removed (I read the Docs that there is an unsupported replacement) if I have both the manager and the assistant only accessing the managers inbox via the same server will I avoid concurrency issues (ie I presume mailbox corruption). Many thanks and have a nice weekend Daniel

1 1

Rebuilding missing index/paths
by Alain Perry 13 Feb '26

13 Feb '26

Hello, First of all, I'm sorry if this isn't the best place to ask for that kind of help, in which case I'd appreciate a pointer to any adequate asynchronous channel. ------------ Some context (skip to next horizontal line if you don't care): I'm happily self-hosting a few services at home. Some months ago I decided I needed to backup my emails to my NAS, in a way that would allow me to read or search them from a web browser as easily as I did with my "cloud" mailboxes. Since I didn't find a modern yet simple webmail that could directly read a Maildir storage, I finally bit the bullet and ran a dovecot container to copy my emails to (using imapsync). Since I opted for an IMAP server, I thought I didn't need to worry too much about the storage format and used the sdbox that (I believe) was the default. Backing up worked and searching worked and I was a happy person. I used my local webmail a couple of times to search for an old email, found it, thought this was a job well done that I wouldn't have to worry too much about anymore. Then, it stopped working. I honestly don't remember what I did around that time that might have broken things or caused a data loss, but IMAP clients couldn't access my emails anymore. Since I had other more urgent matters to address, I let that problem rot. Now that I have a bit more time, I decided to try again. ------------ My problem is that, while it is clear that some data is missing, I have no clue what that data is and how (if ?) I can rebuild it. The folder structure is there in /src/vmail/username/Mail/mailboxes. I do see index(.log) and dbox-Mails/u.* in the folders, so it does look like it conforms to the doc in this regard. However, the paths returned for the mailboxes by "doveadm mailbox path" all use guid's, and do not exist on the filesystem. I don't quite understand how it got to this point, nor what I should have lying around at those paths but I wonder if it is at all possible to rebuild them so I might regain access to my archives. Is there something I can do to recover my emails ? Thanks for reaching this point, and thank you for any help and/or pointers you might be able to provide. Regards,

2 2

fts-flatcurve change index location
by grabsoon＠gmail.com 12 Feb '26

12 Feb '26

hi is there any way to change location ONLY FOR flatcurve indexes all I want is to write fts-flatcurve directories outside ~/Maildir path I do not want to change location of: dovecot.list.index.log dovecot.mailbox.log dovecot-uidvalidity Only for flatcurve indexes. when i set mail_index_path = ~/Indexes all indexes are saved to this new location, not just flatcurve i'm afraid that changing mail_index_path value on production environment is very riski Thanks

2 1

Re: 2.4: fts flatcurve remove?
by Michael Slusarz 11 Feb '26

11 Feb '26

> On 02/10/2026 3:07 PM MST Bjoern Franke via dovecot <dovecot(a)dovecot.org> wrote: > > in the pre release docs > https://doc.dovecot.org/main/core/man/doveadm-fts.1.html#fts-flatcurve-remo… > > doveadm fts flatcurve remove > > was mentioned, but in https://doc.dovecot.org/2.4.2/ it's gone. ? https://doc.dovecot.org/2.4.2/core/man/doveadm-fts.1.html#fts-flatcurve-rem…

2 1

2.4: fts flatcurve remove?
by Bjoern Franke 11 Feb '26

11 Feb '26

Hi, in the pre release docs https://doc.dovecot.org/main/core/man/doveadm-fts.1.html#fts-flatcurve-remo… doveadm fts flatcurve remove was mentioned, but in https://doc.dovecot.org/2.4.2/ it's gone. So, how to remove the flatcurve data again? Simply a find ...--delete? Regards Bjoern

1 0

doveadm quota get issue in lcx container
by Kim Haverblad 09 Feb '26

09 Feb '26

Hi, I'm having some issues getting the quota plugin to work properly. Based on the error message, it looks like the plugin doesn't fully cooperate with Proxmox mount points for LXC containers. It appears to be referencing the secondary disk directly as /dev/nvme1n1p1, rather than the actual mount point inside the container. Here's the full error message: mlgate:/usr/lib/systemd/system# doveadm quota get -u [1]some(a)example.com doveadm([2]some(a)example.com): Error: Failed to get quota resource STORAGE: quota-fs: quotactl(Q_GETQUOTA, /dev/nvme1n1p1) failed: No such file or directory Are there any parameters that need to be adjusted to make this work, or is there an alternative plugin that would handle this setup better? //Kim

2 1