[Dovecot] Mixing v2.1 and 2.0 directors ?
I'm preparing to set up a new set of directors on dovecot-ee-2.1.10.3-1, but would prefer to do this a bit gradually. Will it be OK to set up a ring of directors with 2x dovecot-ee-2.1.10.3-1 and 2x dovecot-2.0.14 ?
-jf
On 27.12.2012, at 23.17, Jan-Frode Myklebust janfrode@tanso.net wrote:
I'm preparing to set up a new set of directors on dovecot-ee-2.1.10.3-1, but would prefer to do this a bit gradually. Will it be OK to set up a ring of directors with 2x dovecot-ee-2.1.10.3-1 and 2x dovecot-2.0.14 ?
The new v2.1 director code can handle running with old v2.1 directors (there were some protocol changes that improve things). I think v2.0 director is protocol compatible with the old v2.1 directors, so I think in theory it should work.. But it's definitely not ever been tested in practise, and v2.1 did fix a ton of director bugs. So if you end up testing it, I think you should be ready to quicky upgrade the other director as well if any errors show up in logs.
On Fri, Dec 28, 2012 at 2:02 AM, Timo Sirainen tss@iki.fi wrote:
The new v2.1 director code can handle running with old v2.1 directors (there were some protocol changes that improve things). I think v2.0 director is protocol compatible with the old v2.1 directors, so I think in theory it should work.. But it's definitely not ever been tested in practise, and v2.1 did fix a ton of director bugs. So if you end up testing it, I think you should be ready to quicky upgrade the other director as well if any errors show up in logs.
Ok, I don't think I want to test this -- realistic testing is too hard. I'll rather upgrade the old directors (keeping same ip-addresses), so that I quickly can rollback in case something doesn't work as well as expected.
BTW: What's the status of LMTP proxying in v2.1 (or more specifically dovecot-ee-2.1.10.3-1)? Do you know of many users of it, and has it proven itself much better than v2.0.14 ? I intend to upgrade the directors first, and leave the backend servers running v2.0.14 for a while.. that should be OK, right ?
-jf
participants (2)
-
Jan-Frode Myklebust
-
Timo Sirainen