Hi Aki,

Uhm well, that’s pretty hard not to do so… Mail.app only has one “Delete” option and that triggers the situation. My support application fetches the email and then deletes it (basically the same as Mail.app). My customers use a variety of MUA’s and they see the same behaviour..

I think the “notification” for the removal is not yet on the other machine, but it sees something changing and thinks “well I have this new message, that you do not have, sync it”.

Thanks, Remko

On 21 Oct 2017, at 15:06, Aki Tuomi aki.tuomi@dovecot.fi wrote:

Maybe you could not delete the message instantly, but keep track of last seen UID. UIDs increase monotonically, so you can rest assured that next arriving email has larger UID.

Then you can bulk delete mails that are older than one day.

Aki

...

On October 20, 2017 at 8:01 PM Remko Lodder remko@FreeBSD.org wrote:

Hi,

Anyone has suggestions? The situation also happends when I “delete” a message from my Phone. It seems to get replicated instantly and the just deleted email is back in the mailbox again.

I did remove HA Proxy support in the meantime to rule that out and I have enabled the default replication_max_conns.

thanks Remko

...

On 13 Oct 2017, at 11:56, Remko Lodder remko@FreeBSD.org wrote:

Dear Dovecot and community,

We run a small email service for our customers, based on two machines that are made “redundant or clustered” by using the replication feature of Dovecot. This works well, for most emails.

Sometimes the following happends:

Email to our support database arrives at the inbox. Every period a cronjob looks into that mailbox and parses the information and makes a support ticket from that message. The cronjob deletes the email afterwards and sees whether there are more new mails or not.

In between the deletion and the (next) check, the email gets “resynced” again from the “other machine”, and is seen as a new email by the cronjob and gets handled again. So basically we get two tickets for one email. After that second run the email is no longer replicated.

This behaviour can also be seen via webmail and email clients, this suggests that the replication might not be aware (enough) that the replication is occurring or that an item already had been synced from A to B (and does not need to get back from B to A after A deleted it).

Is there a way to fiddle with the acknowledgement timing or give the processes some more space/time to get on par with eachother?

Below is the configuration of machine B, they are synchronised through puppet, so only the hostname and IP addresses are different. (so for replication, A has: tcps:hostname_of_b:12346 and B has tcps:hostname_of_a:12346). Also note that HAproxy is prepared but not in use at all.

# 2.2.33.1 (e9afa7f18): /usr/local/etc/dovecot/dovecot.conf # Pigeonhole version 0.4.20 (7cd71ba) # OS: FreeBSD 11.1-RELEASE amd64 auth_mechanisms = plain login disable_plaintext_auth = no doveadm_password = # hidden, use -P to show it haproxy_trusted_networks = IPv4_Haproxy IPv6_Haproxy lda_mailbox_autocreate = yes lda_mailbox_autosubscribe = yes lmtp_save_to_detail_mailbox = yes mail_fsync = always mail_location = sdbox:~/sdbox mail_plugins = " quota notify replication" managesieve_notify_capability = mailto managesieve_sieve_capability = fileinto reject envelope encoded-character vacation subaddress comparator-i;ascii-numeric relational regex imap4flags copy include variables body enotify environment mailbox date index ihave duplicate mime foreverypart extracttext imapsieve vnd.dovecot.imapsieve namespace { inbox = yes location = mailbox Drafts { auto = subscribe special_use = \Drafts } mailbox Junk { special_use = \Junk } mailbox Sent { auto = subscribe special_use = \Sent } mailbox "Sent Messages" { special_use = \Sent } mailbox Spam { auto = subscribe special_use = \Junk } mailbox Trash { auto = subscribe special_use = \Trash } prefix = separator = . } passdb { driver = pam } plugin { imapsieve_mailbox1_before = file:/usr/local/lib/dovecot/sieve/report-spam.sieve imapsieve_mailbox1_causes = COPY imapsieve_mailbox1_name = Spam imapsieve_mailbox2_before = file:/usr/local/lib/dovecot/sieve/report-ham.sieve imapsieve_mailbox2_causes = COPY imapsieve_mailbox2_from = Spam imapsieve_mailbox2_name = * mail_log_events = delete undelete expunge copy mailbox_delete mailbox_rename mail_log_fields = uid box msgid size mail_replica = tcps:hostname_of_machine_a:12346 sieve = ~/.dovecot.sieve sieve_after = /usr/local/etc/dovecot/sieve/after/ sieve_before = /usr/local/etc/dovecot/sieve/global/ sieve_pipe_bin_dir = /usr/local/lib/dovecot/sieve sieve_plugins = sieve_imapsieve sieve_extprograms sieve_vacation_dont_check_recipient = yes } postmaster_address = postmaster@ourdomain.tld protocols = imap pop3 lmtp sieve replication_max_conns = 100 service aggregator { fifo_listener replication-notify-fifo { mode = 0666 } unix_listener replication-notify { mode = 0666 } } service auth { unix_listener /var/spool/postfix/private/auth { mode = 0666 } } service doveadm { inet_listener { port = 12346 ssl = yes } } service imap-login { inet_listener imap_haproxy { haproxy = yes port = 10143 } inet_listener imaps_haproxy { haproxy = yes port = 10144 ssl = yes } service_count = 1 } service imap { process_limit = 1024 } service lmtp { inet_listener lmtp { address = IPv4, IPv6, 127.0.0.1 ::1 port = 24 } unix_listener /var/spool/postfix/private/dovecot-lmtp { group = postfix mode = 0600 user = postfix } } service pop3 { process_limit = 1024 } service replicator { process_min_avail = 1 unix_listener replicator-doveadm { mode = 0666 } } ssl_ca =