December 2019 - dovecot

SEARCH=INTHREAD not announced in capabilities
by kael 08 Dec '19

08 Dec '19

Hello, Looking at the SEARCH=INTHREAD extension, Dovecot implements it but the server doesn't announce the extension in its capabilities (at least testing with imap://dovecot.org or looking at https://github.com/dovecot/core/blob/169b1ff42a28cdf2768391d20e8fef13aeae54… ) It does work though: a UID SEARCH INTHREAD REFS HEADER subject "Re: bash script" * SEARCH 106075 106076 106084 a OK Search completed (1.047 + 0.000 + 1.046 secs). I'd have thought that announcing THREAD=REFS would imply SEARCH=INTHREAD, but testing with Fastmail only the former is implemented. Looks like a bug in Dovecot as the client can't discover the SEARCH=INTHREAD capability. Cheers. -- kael

1 0

Accessing the -a and -d from dovecot-lda
by Aristedes Maniatis 08 Dec '19

08 Dec '19

I've got exim set up to deliver like this: command = /usr/local/libexec/dovecot/dovecot-lda -a $original_local_part@$original_domain -d $local_part@$domain -f $sender_address If an email to is bob+sales(a)ish.com.au then we should have "-a bob+sales(a)subdomain.ish.com.au -d bob(a)ish.com.au". Now in dovecot, only bob(a)ish.com.au is a real user but I want to be able to use both addresses in different places. For LDAP lookup I need to reference bob(a)ish.com.au. But in sieve, I need to see the original address with plus notation. However, dovecot appears to get the "-a" address into %u and the "-d" address can't be accessed. https://doc.dovecot.org/configuration_manual/config_file/config_variables/ Is what is the point of dovecot-lda having both -a and -d options? Can I get at both bits of data? Thanks Ari

2 1

ssl_min_protocol = TLSv1.3 does not work
by Laurens Post 08 Dec '19

08 Dec '19

Hi all, I'm trying to set up my server with support for TLS 1.3 only, but that does not seem to be supported. First off, TLS 1.3 itself does work fine, so it's not the config or ssl library, and 1.3-only works fine with Postfix. The problem is only in disabling TLS 1.2 for Dovecot. On connection, I'm getting an error that 1.3 is an "Unknown ssl_min_protocol setting". Reading the source code, it seems that `openssl_min_protocol_to_options` in `src/lib-ssl-iostream/iostream-openssl-common.c` is simply missing an entry like { SSL_TXT_TLSV1_3, TLS1_3_VERSION, SSL_OP_NO_SSLv3 | SSL_OP_NO_TLSv1 | SSL_OP_NO_TLSv1_1 | SSL_OP_NO_TLSv1_2 } Is this a bug, something intentional, or has it simply not been added yet because nobody has been crazy enough to ask for it? Kind regards, Laurens

3 2

Dovecot 2.3.8 - How to force index creation for user/mailbox?
by Tomek Lutelmowski 08 Dec '19

08 Dec '19

Hello, I'm using maildirs, and I have problems with big mailboxes (~ 20000 emails, 30-50GB) - sometimes access to such mailbox takes very long (~10 minutes), I would like to be able to create index for mailbox in background, but I couldn't find any way to do it. So far I tried: *doveadm -D -vvv index -u user 'Folder' * But even if this folder have no dovecot.index* files - I always get: doveadm(user): Info: Wersje robocze: Cache is already up to date It seems that doveadm doesn't actually checks if the index is present. The index is created only when I log with IMAP client and check the folder. So i tried: *doveadm mailbox cache purge -u user 'Folder' * No joy either: doveadm(user): Error: Cache is unusable And command below doesn't actually do anything to index: *doveadm -vvv force-resync -u user 'Folder'* Is there any way to use doveadm tool to create/fix index? Shouldn't the index be updated on each email arrival? Cheers, TomekL

5 12

lmtp proxy 'Invalid FROM: Missing domain'
by Marc Roos 06 Dec '19

06 Dec '19

Looks like I have a correct working proxy on pop3. On both backend servers I can run cat /tmp/test.msg | /usr/libexec/dovecot/lmtp Giving 250 2.1.5 OK 354 OK Info: Disconnect from local: Connection closed (in DATA) However if I run the same command on the proxy, I am getting this error 'Invalid FROM: Missing domain'. How is this FROM even relevant if message just needs to be delivered to RCPT TO.

2 2

Sieve redirect does not collect the sendmail child process correctly (Dovecot 2.3.4.1, Pigeonhole 0.5.4)
by Delian Krustev 06 Dec '19

06 Dec '19

Hi, I got this simple test Sieve config: ############################################# # Sieve Filter # Generated by Ingo (http://www.horde.org/apps/ingo/) (11/01/2019, 10:46:06 PM) # Forwards if true { redirect "krustev(a)krustev.net"; } # Forward Keep Action if true { keep; stop; } ############################################# Which used to work fine with: # 2.2.27 (c0f36b0): /etc/dovecot/dovecot.conf # Pigeonhole version 0.4.16 (fed8554) # OS: Linux 4.9.0-8-amd64 x86_64 Debian 9.11 Then I've tried migrating to: # 2.3.4.1 (f79e8e7e4): /etc/dovecot/dovecot.conf # Pigeonhole version 0.5.4 () # OS: Linux 4.19.0-6-amd64 x86_64 Debian 10.1 And started getting the following errors: # cat .dovecot.sieve.log sieve: info: started log at Nov 01 22:55:50. error: msgid=<4221584.AlbrL4BjWX@home>: failed to redirect message to <krustev(a)krustev.net>: Failed to execute sendmail (temporary failure). A google search lead me to: https://bugs.archlinux.org/task/56933 which blames systemd service changes: NoNewPrivileges=true , however I'm not running dovecot via systemd but use DJB's daemontools for service supervision . The page above suggested to change from: sendmail_path = /usr/sbin/sendmail to submission_host = localhost in "conf.d/15-lda.conf", which worked fine. But since I don't have other problems with "sendmail" execution I've decided to debug this further. I have dovecot integrated with Qmail: # apt-cache policy qmail qmail: Installed: 1.06-6+b1 Candidate: 1.06-6+b1 Version table: *** 1.06-6+b1 900 900 http://deb.debian.org/debian buster/main amd64 Packages so sendmail's implementation comes from Qmail: # ls -al `which sendmail` lrwxrwxrwx 1 root root 14 Feb 23 2015 /usr/sbin/sendmail -> qmail-sendmail I've strace-d the qmail-lspawn process which executes dovecot's LDA, which processes the Sieve rules and executes "sendmail". The sendmail process executes and returns successfully, the mail is delivered, however the Sieve code does not wait for the child process correctly and tries to kill the already exited process after 30 seconds timeout. Here are the problematic Sieve syscalls: 22:55:15.283482 execve("/usr/lib/dovecot/deliver", ["/usr/lib/dovecot/deliver", "-e", "-a", "tester(a)krustev.net"], 0x5620b4330c40 /* 61 vars */) = 0 ... 22:55:15.286845 pipe([5, 6]) = 0 22:55:15.286870 fcntl(5, F_GETFL) = 0 (flags O_RDONLY) 22:55:15.286889 fcntl(5, F_SETFL, O_RDONLY|O_NONBLOCK) = 0 22:55:15.286909 fcntl(6, F_GETFL) = 0x1 (flags O_WRONLY) 22:55:15.286928 fcntl(6, F_SETFL, O_WRONLY|O_NONBLOCK) = 0 22:55:15.286947 fcntl(5, F_GETFD) = 0 22:55:15.286966 fcntl(5, F_SETFD, FD_CLOEXEC) = 0 22:55:15.286985 fcntl(6, F_GETFD) = 0 22:55:15.287003 fcntl(6, F_SETFD, FD_CLOEXEC) = 0 22:55:15.287024 epoll_create(128) = 7 22:55:15.287043 fcntl(7, F_GETFD) = 0 22:55:15.287062 fcntl(7, F_SETFD, FD_CLOEXEC) = 0 22:55:15.287084 epoll_ctl(7, EPOLL_CTL_ADD, 5, {EPOLLIN|EPOLLPRI|EPOLLERR|EPOLLHUP, {u32=47201248, u64=94699781110752}}) = 0 ... 22:55:15.294625 epoll_ctl(7, EPOLL_CTL_DEL, 5, 0x7ffd59156b8c) = 0 22:55:15.294647 epoll_create(128) = 12 22:55:15.294668 fcntl(12, F_GETFD) = 0 22:55:15.294687 fcntl(12, F_SETFD, FD_CLOEXEC) = 0 22:55:15.294708 epoll_ctl(12, EPOLL_CTL_ADD, 5, {EPOLLIN|EPOLLPRI|EPOLLERR|EPOLLHUP, {u32=47455536, u64=94699781365040}}) = 0 Dovecot/Sieve will wait for this pipe FD 5 later, but no event will arrive. 22:55:15.294740 rt_sigaction(SIGCHLD, {sa_handler=0x7f15404b3640, sa_mask=[], sa_flags=SA_RESTORER|SA_RESTART|SA_SIGINFO, sa_restorer=0x7f1540229840}, NULL, 8) = 0 22:55:15.294763 pipe([13, 14]) = 0 22:55:15.294788 clone(child_stack=NULL, flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD, child_tidptr=0x7f15401ec2d0) = 6929 The sendmail child process with PID 6929 is forked here. 22:55:15.294928 close(13) = 0 22:55:15.294969 fcntl(14, F_GETFL) = 0x1 (flags O_WRONLY) 22:55:15.295006 fcntl(14, F_SETFL, O_WRONLY|O_NONBLOCK) = 0 22:55:15.295039 lseek(14, 0, SEEK_CUR) = -1 ESPIPE (Illegal seek) 22:55:15.295065 getsockname(14, 0x7ffd59156a40, [28]) = -1 ENOTSOCK (Socket operation on non-socket) 22:55:15.295109 write(14, "EDITTED LONG MESSAGE DATA\n", 1461) = 1461 22:55:15.295150 close(14) = 0 22:55:15.295175 wait4(6929, 0x562102d4fee4, WNOHANG, NULL) = 0 22:55:15.295211 epoll_wait(12, [], 1, 30000) = 0 Here's where the 30 second's timeout is requested and waited. 22:55:45.325580 kill(6929, SIGTERM) = 0 22:55:45.325769 epoll_wait(12, [], 1, 5000) = 0 22:55:50.331108 kill(6929, SIGKILL) = 0 22:55:50.332867 wait4(6929, [{WIFEXITED(s) && WEXITSTATUS(s) == 0}], 0, NULL) = 6929 22:55:50.333020 epoll_ctl(12, EPOLL_CTL_DEL, 5, 0x7ffd59156b6c) = 0 22:55:50.333060 epoll_ctl(7, EPOLL_CTL_ADD, 5, {EPOLLIN|EPOLLPRI|EPOLLERR|EPOLLHUP, {u32=47201248, u64=94699781110752}}) = 0 22:55:50.333097 close(12) = 0 22:55:50.333240 sendto(9, "<19>Nov 1 22:55:50 dovecot: lda(krustev)<6926><IlpXEdO3vF0OGwAA4EGeoA>: Error: sieve: msgid=<4221584.AlbrL4BjWX@home>: failed to redirect message to <krustev(a)krustev.net>: Failed to execute sendmail (temporary failure)", 220, MSG_NOSIGNAL, NULL, 0) = 220 22:55:50.333321 openat(AT_FDCWD, "/home/krustev/mail/domains/krustev.net/tester/.dovecot.sieve.log", O_WRONLY|O_CREAT|O_APPEND, 0600) = 12 22:55:50.333557 fstat(12, {st_mode=S_IFREG|0600, st_size=0, ...}) = 0 22:55:50.333613 lseek(12, 0, SEEK_CUR) = 0 22:55:50.333643 fstat(12, {st_mode=S_IFREG|0600, st_size=0, ...}) = 0 22:55:50.333678 stat("/etc/localtime", {st_mode=S_IFREG|0644, st_size=127, ...}) = 0 22:55:50.333726 write(12, "sieve: info: started log at Nov 01 22:55:50.\n", 45) = 45 22:55:50.333788 write(12, "error: msgid=<4221584.AlbrL4BjWX@home>: failed to redirect message to <krustev(a)krustev.net>: Failed to execute sendmail (temporary failure).\n", 141) = 141 22:55:50.333857 sendto(9, "<19>Nov 1 22:55:50 dovecot: lda(krustev)<6926><IlpXEdO3vF0OGwAA4EGeoA>: Error: sieve: Execution of script /home/krustev/mail/domains/krustev.net/tester/.dovecot.sieve was aborted due to temporary failure (user logfile /home/krustev/mail/domains/krustev.net/tester/.dovecot.sieve.log may reveal additional details)", 317, MSG_NOSIGNAL, NULL, 0) = 317 22:55:50.333941 close(10) = 0 22:55:50.333996 close(12) = 0 22:55:50.334057 lstat("/home/krustev/mail/domains/krustev.net/tester/.dovecot.lda-dupes.lock", {st_mode=S_IFREG|0600, st_size=0, ...}) = 0 22:55:50.334095 unlink("/home/krustev/mail/domains/krustev.net/tester/.dovecot.lda-dupes.lock") = 0 22:55:50.334188 close(11) = 0 22:55:50.334245 write(2, "Execution of Sieve filters was aborted due to temporary failure\n", 64) = 64 22:55:50.334418 munmap(0x7f15401bc000, 103680) = 0 22:55:50.334481 munmap(0x7f15403b3000, 32800) = 0 22:55:50.334516 munmap(0x7f15400fb000, 788352) = 0 22:55:50.334610 epoll_ctl(7, EPOLL_CTL_DEL, 8, 0x7ffd5915706c) = 0 22:55:50.334640 close(8) = 0 22:55:50.334685 epoll_ctl(7, EPOLL_CTL_DEL, 5, 0x7ffd5915706c) = 0 22:55:50.334738 close(5) = 0 22:55:50.334785 close(6) = 0 22:55:50.334813 close(7) = 0 22:55:50.334841 close(4) = 0 22:55:50.334881 exit_group(75) = ? 22:55:50.335077 +++ exited with 75 +++ Here's the interesting part of sendmail's (PID 6929) execution with timestamps: .. 22:55:15.295268 execve("/usr/sbin/sendmail", ["/usr/sbin/sendmail", "-i", "-f", "tester(a)krustev.net", "--", "krustev(a)krustev.net"], NULL) = 0 ... 22:55:15.296024 chdir("/var/lib/qmail") = 0 22:55:15.296055 execve("bin/qmail-inject", ["bin/qmail-inject", "-a", "-f", "tester(a)krustev.net", "--", "krustev(a)krustev.net"], 0x55741b1890e0 /* 2 vars */) = 0 ... 22:55:15.297264 vfork() = 6930 .. 22:55:15.297930 wait4(6930, [{WIFEXITED(s) && WEXITSTATUS(s) == 0}], 0, NULL) = 6930 22:55:15.305391 exit_group(0) = ? 22:55:15.305472 +++ exited with 0 +++ The PID 6930 is a call to the setuid qmail-queue, which also executes and returns normally. With all the above stated - It seems that the problem is in the "epoll" calls. Either it is not set-up correctly or dovecot-LDA does not receive the information it waits for. I've tried to make my way trough Dovecot's source code but have concluded that someone more familiar should be able to spot the problem much quicker than me. Let me know if you need some additional information. Sorry for the long mail. Dovecot config follows: # 2.3.4.1 (f79e8e7e4): /etc/dovecot/dovecot.conf # Pigeonhole version 0.5.4 () # OS: Linux 4.19.0-6-amd64 x86_64 Debian 10.1 # Hostname: pinko.livehostserver.com auth_mechanisms = plain login disable_plaintext_auth = no lda_mailbox_autocreate = yes lda_mailbox_autosubscribe = yes log_path = /dev/stderr log_timestamp = mail_plugins = " quota" 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 namespace inbox { inbox = yes location = mailbox Archive { special_use = \Archive } mailbox Archives { special_use = \Archive } mailbox "Deleted Items" { special_use = \Trash } mailbox "Deleted Messages" { special_use = \Trash } mailbox Drafts { auto = subscribe special_use = \Drafts } mailbox Junk { auto = subscribe special_use = \Junk } mailbox "Junk E-mail" { special_use = \Junk } mailbox Sent { auto = subscribe special_use = \Sent } mailbox "Sent Items" { special_use = \Sent } mailbox "Sent Mail" { special_use = \Sent } mailbox "Sent Messages" { special_use = \Sent } mailbox Spam { special_use = \Junk } mailbox Trash { auto = subscribe special_use = \Trash } mailbox sent-mail { special_use = \Sent } prefix = } passdb { args = /usr/local/sbin/vchkpw-dovecot driver = checkpassword } plugin { antispam_allow_append_to_spam = yes antispam_backend = pipe antispam_pipe_program = /bin/sh antispam_pipe_program_notspam_args = -c;HOME=$PWD exec nice sa-learn --ham antispam_pipe_program_spam_args = -c;HOME=$PWD exec nice sa-learn --spam antispam_pipe_tmpdir = /tmp antispam_spam_pattern_ignorecase = spam;junk;junk * antispam_trash_pattern_ignorecase = trash;Deleted * quota = maildir:User quota sieve = ~/.dovecot.sieve sieve_dir = ~/sieve sieve_redirect_envelope_from = recipient } protocols = " imap sieve pop3" service anvil { client_limit = 1536 } service auth { client_limit = 1536 } service imap-login { process_limit = 512 } service managesieve-login { inet_listener sieve { port = 4190 } process_min_avail = 1 service_count = 1 } service managesieve { process_limit = 50 } service pop3-login { process_limit = 512 } service stats { client_limit = 1024 unix_listener stats-writer { mode = 0666 } } ssl_ca = </scm/deb-global/etc/ssl/intermediate-CAs.crt ssl_cert = </etc/ssl/default-fullchain.pem ssl_client_ca_dir = /etc/ssl/certs ssl_dh = # hidden, use -P to show it ssl_key = # hidden, use -P to show it userdb { driver = prefetch } protocol lda { info_log_path = log_path = mail_plugins = " quota sieve" } protocol imap { mail_max_userip_connections = 64 mail_plugins = " quota imap_quota antispam" } protocol sieve { mail_max_userip_connections = 10 managesieve_implementation_string = Dovecot Pigeonhole managesieve_max_line_length = 65536 } Best regards -- Delian

2 3

Problem with sievefilter forwarding
by ratatouille 06 Dec '19

06 Dec '19

Hello! dovecot --version 2.2.36 (1f10bfa63) I defined a sievefilter require ["copy"]; # rule:[weiterleitung] if header :contains "to" "user(a)bitclusive.de" { redirect :copy "otheruser(a)gmail.com"; stop; } It doesn't work and I don't know why. I can define a vacationfilter that works fine. Can I debug this somehow and can somebody here explain what is going wrong? Kind regards Andreas

2 2

ACL hide folders
by Peter Chiochetti 06 Dec '19

06 Dec '19

After updating from 2.2.22 to 2.2.33 rsp 2.3.8 my ACLs went bust. Below line in the global acl file hides all folders except INBOX for all users > * group=ALL Next reenable certain folders, eg: > Sent group=ALL lrwsi This made anyone see "Sent" in 2.2.22, but no longer so in 2.2.33 Any ideas how to have it like before? -- peter

1 3

v2.3.9 released
by Aki Tuomi 06 Dec '19

06 Dec '19

Hi all! We are pleased to release v2.3.9 of Dovecot. Please find it from locations below --- Aki Tuomi Open-Xchange oy https://dovecot.org/releases/2.3/dovecot-2.3.9.tar.gz https://dovecot.org/releases/2.3/dovecot-2.3.9.tar.gz.sig Binary packages in https://repo.dovecot.org/ Docker images in https://hub.docker.com/r/dovecot/dovecot --- * Changed several event field names for consistency and to avoid conflicts in parent-child event relationships: * SMTP server command events: Renamed "name" to "cmd_name" * Events inheriting from a mailbox: Renamed "name" to "mailbox" * Server connection events have only "remote_ip", "remote_port", "local_ip" and "local_port". * Removed duplicate "client_ip", "ip" and "port". * Mail storage events: Removed "service" field. Use "service:<name>" category instead. * HTTP client connection events: Renamed "host" to "dest_host" and "port" to "dest_port" * auth: Drop Postfix socketmap support. It hasn't been working with recent Postfix versions for a while now. * push-notification-lua: The "subject" field is now decoded to UTF8 instead of kept as MIME-encoded. + push-notification-lua: Added new "from_address", "from_display_name", "to_address" and "to_display_name" fields. The display names are decoded to UTF8. + Added various new fields to existing events. See http://doc.dovecot.net/admin_manual/list_of_events.html + Add lmtp_add_received_header setting. It can be used to prevent LMTP from adding "Received:" headers. + doveadm: Support SSL/STARTTLS for proxied doveadm connections based on doveadm_ssl setting and proxy ssl/tls settings. + Log filters support now "service:<name>", which matches all events for the given service. It can also be used as a category. + lib: Use libunwind to get abort backtraces with function names where available. + lmtp: When the LMTP proxy changes the username (from passdb lookup) add an appropriate ORCPT parameter. - lmtp: Add lmtp_client_workarounds setting to implement workarounds for clients that send MAIL and RCPT commands with additional spaces before the path and for clients that omit <> brackets around the path. See example-config/conf.d/20-lmtp.conf. - lda/lmtp: Invalid MAIL FROM addresses were rejcted too aggressively. Now mails from addresses with unicode characters are delivered, but their Return-Path header will be <> instead of the given MAIL FROM address. - lmtp: The lmtp_hdr_delivery_address setting is ignored. - imap: imap_command_finished event's "args" and "human_args" parameters were always empty. - mbox: Seeking in zlib and bzip2 compressed input streams didn't work correctly. - imap-hibernate: Process crashed when client got destroyed while it was attempted to be unhibernated, and the unhibernation fails. - *-login: Proxying may have crashed if SSL handshake to the backend failed immediately. This was unlikely to happen in normal operation. - *-login: If TLS handshake to upstream server failed during proxying, login process could crash due to invalid memory access. - *-login: v2.3 regression: Using SASL authentication without initial response may have caused SSL connections to hang. This happened often at least with PHP's IMAP library. - *-login: When login processes are flooded with authentication attempts it starts logging errors about "Authentication server sent unknown id". This is still expected. However, it also caused the login process to disconnect from auth server and potentially log some user's password in the error message. - dict-sql: SQL prepared statements were not shared between sessions. This resulted in creating a lot of prepared statements, which was especially inefficient when using Cassandra backend with a lot of Cassandra nodes. - auth: auth_request_finished event didn't have success=yes parameter set for successful authentications. - auth: userdb dict - Trying to list users crashed. - submission: Service could be configured to allow anonymous authentication mechanism and anonymous user access. - LAYOUT=index: Corrupted dovecot.list.index caused folder creation to panic. - doveadm: HTTP server crashes if request target starts with double "/". - dsync: Remote dsync started hanging if the initial doveadm "dsync-server" command was sent in the same TCP packet as the following dsync handshake. v2.3.8 regression. - lib: Several "input streams" had a bug that in some rare situations might cause it to access freed memory. This could lead to crashes or corruption. The only currently known effect of this is that using zlib plugin with external mail attachments (mail_attachment_dir) could cause fetching the mail to return a few bytes of garbage data at the beginning of the header. Note that the mail wasn't saved corrupted, but fetching it caused corrupted mail to be sent to the client. - lib-storage: If a mail only has quoted content, use the quoted text for generating message snippet (IMAP PREVIEW) instead of returning empty snippet. - lib-storage: When vsize header was rebuilt, newly calculated message sizes were added to dovecot.index.cache instead of being directly saved into vsize records in dovecot.index. - lib: JSON generator was escaping UTF-8 characters unnecessarily.

3 2

[2.3.8] possible replication issue
by Carsten Rosenberg 06 Dec '19

06 Dec '19

Hi, some of our customers have discovered a replication issue after upgraded from 2.3.7.2 to 2.3.8. Running 2.3.8 several replication connections are hanging until defined timeout. So after some seconds there are $replication_max_conns hanging connections. Other replications are running fast and successful. Also running a doveadm sync tcp:... is working fine for all users. I can't see exactly, but I haven't seen mailboxes timeouting again and again. So I would assume it's not related to the mailbox. >From the logs: server1: Oct 16 08:29:25 server1 dovecot[5715]: dsync-local(username1(a)domain.com)<FXnVDW22pl0tGAAA1cwDxA>: Error: dsync(172.16.0.1): I/O has stalled, no activity for 600 seconds (version not received) Oct 16 08:29:25 server1 dovecot[5715]: dsync-local(username1(a)domain.com)<FXnVDW22pl0tGAAA1cwDxA>: Error: Timeout during state=master_recv_handshake server2: Oct 16 08:29:25 server2 dovecot[8113]: doveadm: Error: read(server1) failed: EOF (last sent=handshake, last recv=handshake) There aren't any additional logs regarding the replication. I have tried increasing vsz_limit or reducing replication_max_conns. Nothing changed. -- Both customers have 10k+ users. Currently I couldn't reproduce this on smaller test systems. Both installation were downgraded to 2.3.7.2 to fix the issue for now -- I've attached a tcpdump showing the client showing the client stops sending any data after the mailbox_guid table headers. Any idea what could be wrong here or the debug this issue? Thanks. Carsten Rosenberg

5 5