stability

[mr-smit]

Hi, searched all internet to find this functionallity. Can you tell more about it? Is it on your production clusters? its stable?

i want to use it in three geo clusters

and how to install and configure it ?

[Andrei-Pozolotin]

1. stability: yes, used in production

2. install function: download and unpack:

dovecot_syncer() {
    echo "# $FUNCNAME"
    local plugin_source="https://raw.githubusercontent.com/random-archer/dovecot_plugin_syncer/master/binary.tar.gz"
    local plugin_target="/tmp/binary.tar.gz"
    curl "$plugin_source" --output "$plugin_target"
    tar xzvf "$plugin_target" -C "/"  
}

2. usage: study unit test
   https://github.com/random-archer/dovecot_plugin_syncer/blob/master/test/dovecot.sh

[mr-smit]

thanks, will try to install and configure it today, i will report later.

[mr-smit]

works stable on 3 geo clusters , as u sad first i need to create my own .sh script to read file pipe and run

doveadm sync -f -u $user_name -n inbox remote:root@serv1c1
doveadm sync -f -u $user_name -n inbox remote:root@serv1c3
doveadm sync -f -u $user_name -n inbox remote:root@serv1c3
...etc.. unlimited

thanks for good script, dovecot must have it.

[Andrei-Pozolotin]

great. re: need to create my own .sh script:
here is another source for inspiration: https://github.com/random-python/mail_serv

[mr-smit]

hi, after some time i got this error:

Feb 21 14:57:41 imap([email protected])<26532><fTH/XBWf+ul/AAAB>:

Error: syncer-plugin.v13/syncer_report_change : failed to pipe: syncer_record_eol rv=-1
Error: syncer-plugin.v13/syncer_report_change : failed to pipe: record_entry rv=-1

how to reproduce, set limits on imap process and make many connections to imap, then you will see error like:
Remote closed connection (destination service { process_limit } reached?) (

then do dovecot restart:
service dovecot restart

and you will see errors about plugin and pipe.

What do you think ?

The problem is that imap process is not shutting down when no connections are made.

[mr-smit]

As I understand, we need to timeout if we can't write to pipe

int fd_pipe = open(syncer_pipe, O_RDWR); // non-blocking fifo pipe

how to add timeout here for example 3 sec ?

[mr-smit]

fd_set fds;
FD_ZERO(&fds);
FD_SET(fd_pipe, &fds);
struct timeval tv = { 1, 0 }; // 1 secs, 0 usecs

int st = select(fd_pipe+1, NULL, &fds, NULL, &tv);
if (st < 0) {
	// select threw an error
	syncer_log_err("[err] select error");
} else if (FD_ISSET(fd_pipe, &fds)) {
	// report change to non-blocking fifo pipe
	int rv;
	rv = write(fd_pipe, record_entry, strlen(record_entry));
	if (rv < 0) {
		syncer_log_err("%s : %s rv=%d", __func__, "failed to pipe: record_entry", rv);
	}
	rv = write(fd_pipe, syncer_record_eol, strlen(syncer_record_eol));
	if (rv < 0) {
		syncer_log_err("%s : %s rv=%d", __func__, "failed to pipe: syncer_record_eol", rv);
	}
} else {
	// Writing not possible in 1 seconds, wait
	syncer_log_err("[err] write not possible in 1 sec");
}

[mr-smit]

now all works fine.

[pfak]

Is this patch still required?