Stephan Bosch <stephan@rename-it.nl> wrote:
On 5/15/2015 1:03 PM, Stephan Bosch wrote:
On 5/14/2015 9:32 PM, Michael Grimm wrote:
On 14.05.2015, at 19:21, Stephan Bosch <stephan@rename-it.nl> wrote: On 5/14/2015 2:46 PM, Michael Grimm wrote:
I do have replication between two servers running, and mail is replicated without any issues. I do have sieve running at both servers without any issues.
But replication of sieve scripts doesn't work as "expected":
Whenever I do modify my default script und upload it via "doveadm sieve put -u abc -a default < default" it becomes uploaded to the home directory of user abc as expected, but it will not be replicated to the partner server.
Whenever I do deactivate the default script by "doveadm sieve deactivate -u abc", followed by a deletion "doveadm sieve delete -u abc default" plus subsequent upload of the script "doveadm sieve put -u abc -a default < default", replication takes place instantly.
Questions: Is a modification of a sieve script supposed to be triggered the latter way? Or am I doing it the wrong way? That smells like a bug. What is your configuration (dovecot -n)? This should also show the versions of the packages you have installed. Sorry, I should thought about providing more before :-( Sorry, again.
Feel free to ask for potential tests from my site. I had to setup my own replication (I usually only test with manual dsync invocation). Still, I cannot reproduce your problem: replication occurs instantly. Our setups aren't very different, with the exception that I installed two Dovecot instances on one server with replication configured between them.
Do your logs show anything strange?
No, I never found something strange. That why I began thinking about my expectations were wrong regarding how sieve replication is supposed to work.
What do your logs show when you enable mail_debug=yes?
Nothing strange as well.
What do the doveadm commands show when the -D option is added?
This I didn't try :-(
Is your configuration completely equal at both servers (except of course for the configured replication peer)?
Yes, absolutely identical. Same config, same hg revisions, same OS, etc.
This may be helpful:
http://hg.dovecot.org/dovecot-2.2/rev/7cf62f28fd62 (now included in 2.2.18).
Well, I do feel like an idiot! I swear, sieve replication never ran as expected, I always had to do it the way mentioned above. Today, after compiling brand new dovecot, sieve replication works as expected and as you confirmed it should work. I really tried hard to reproduce my issue, but: to no avail ;-) No idea whether those last minute revisions made the trick or something else …
So I feel very sorry having bothered you without real cause :-(
Thanks and with kind regards, Michael
@Timo: There's a typo in src/plugins/replication/replication-plugin.c: s/transction commit/transaction commit/