patches since 2.3.12-p2?

[Per olof Ljungmark]

Alain Spineux wrote:
> On Fri, Jul 18, 2008 at 6:10 PM, Per olof Ljungmark <peo at intersonic.se> wrote:
>> Simon Matter wrote:
>>>> Hi all,
>>>>
>>>> Could some kind soul recommend what patches it makes sense to apply
>>>> since the latest release? There are a number of them circulating on the
>>>> list but I cannot figure out for sure which ones are generally safe and
>>>> recommended.
>>>>
>>>> In particular, anything that addresses
>>>> DBERROR db4: 176 lockers
>>>> (We don't use Berkeley at all. Switched statuscache.db from
>>> I think there is something strange with your current config. If
>>> Cyrus-imapd doesn't use any BDB then I think there will never be any of
>>> those messages in your logs. Did you switch from BDB to skiplist and not
>>> removing some BDB stuff?
>>>
>>> To get rid of anything left you should be able to do the following:
>>> - Cleanly shutdown Cyrus
>>> - go to $configdir
>>> - mkdir db.old
>>> - mv db/* db.old/
>>> - db.backup* db.old/
>>> - Start Cyrus
>>>
>> Thanks for the advice, cannot shut server down right now however. I
>> changed statuscache.db from berkely-nosync to skiplist (to see if the
>> error went away) but also deleted statuscache.db in between. All else
>> has been skiplist or flat for a long long time.
>>
>> The error in question started to appear after upgrade from 2.3.11 to
>> 2.3.12-p2 when I used berkely-nosync for the statuscache.
>>
>> Would not the db.backup* files already be in the same stae, they are
>> recreated every 30 minutes...?
> 
> This is configured in your cyrus.conf . Do you have this error every 30min ?

No, randomly. They do not seem to harm much either.

What puzzles me is that Cyrus was using BerkelyDB even though I had 
skiplist configured in imapd.conf and deleted statuscache.db in between. 
Is this information stored anywhere else except imapd.conf?

Last night I
* Recompiled Cyrus without BDB support
* Changed line 1034 in ./lib/imapoptions to take skiplist as default
and now finally statuscache.db is using skiplist.

I wish I knew what went wrong in the first place, need to build a test 
setup to figure this one out, cannot fiddle more with the production unit.

Now, for the original question, I applied
Skiplist Readlocktracking
Skiplist Safeunlock
last night and I'm considering "Fast Rename" as well.

Thank you Simon and Alan for your comments.

--per