Re: 2.3 to 2.4 Murder upgrade
Andrew Morgan <morgan <at> orst.edu>
2011-12-08 18:00:26 GMT
On Tue, 6 Dec 2011, Dave McMurtrie wrote:
> On 12/06/2011 05:48 PM, Andrew Morgan wrote:
>> On Fri, 14 Oct 2011, Dave McMurtrie wrote:
>>>> Upgrading from 2.4.5
>>>> New config option: suppress_capabilities, which takes a space
>>>> separated list of capabilities which will NOT be given in any imap
>>>> capability response. This can be used on frontends to not display
>>>> capabilities that older backends don't have, so clients don't get
>>>> This leads me to believe that I should upgrade the backends to 2.4
>>>> Is that true? Has anyone done a Murder upgrade from 2.3 to 2.4 and can
>>>> tell me what order they upgraded the hosts?
>>> Yes, you'll want to upgrade the frontends first and suppress the new
>>> capabilities on the frontends that the old backends aren't yet aware of.
> I should have mentioned the reason why we did the frontends first for the
> benefit of anyone else who is searching for this information. You need to do
> the frontends first because as of 2.4, the backends will (correctly) return
> the NoSelect flag for any remote folders that you're subscribed to in an LSUB
> response. This is most common with shared folders, where the shared folder
> may not be on the same backend as the user's Inbox. The 2.4 frontends will
> know to scrub the NoSelect flag in this case.
>> Hmmm, how do I know which capabilities to suppress? Should I just
>> compare the CAPABILITY response from a 2.3 and a 2.4 server?
> We used:
> suppress_capabilities: ESEARCH QRESYNC WITHIN
> but your idea is perfectly cromulent, also.
I compared the CAPABILITY response between 2.3 and 2.4 frontends after TLS
was negotiated and found the following differences:
So, I tried adding all the new 2.4 capabilities to the
suppress_capabilities setting. I found that if I added URLAUTH=BINARY to
suppress_capabilities, I was unable to establish an IMAP connection to the
frontend. imtest didn't print any output, and proxyd on the frontend was
spinning at 100% cpu. If I remove URLAUTH=BINARY from
suppress_capabilities, leaving all the other new capabilities suppressed,
it works fine. I'm not sure if this is a bug... Let me know if I should
file it as a bug.
So that leaves the following capabilities suppressed:
suppress_capabilities: ESEARCH LIST-EXTENDED QRESYNC WITHIN XLIST ENABLE SORT=DISPLAY
The imapd.conf manpage says I only need:
suppress_capabilities: ESEARCH QRESYNC WITHIN XLIST LIST-EXTENDED
if I have 2.3 murder backends. Is that correct, or should I suppress the
new ENABLE and SORT=DISPLAY too?
Cyrus Home Page: http://www.cyrusimap.org/
List Archives/Info: http://lists.andrew.cmu.edu/pipermail/info-cyrus/