Re: (ITS#5073) Syncrepl spinlock

[quanah@zimbra.com]

--On August 6, 2007 5:01:08 PM +0000 quanah@zimbra.com wrote:

>
>
> --On August 6, 2007 9:17:01 AM +0000 ando@sys-net.it wrote:
>
>> quanah@OpenLDAP.org wrote:
>>> Full_Name: Quanah Gibson-Mount
>>> Version: 2.3.37
>>> OS: Linux
>>> URL:
>>> Submission from: (NULL) (71.202.148.128)
>>>
>>>
>>> If you create a syncrepl master that serves the "" base, and then set up
>>> a syncrepl consumer to pull down "dc=example,dc=com", which does *not*
>>> exist on the master, the replica will sit at 100% CPU, endlessly trying
>>> to add nothing.
>>
>> I'm not observing this with either re23 nor HEAD.  The initial search
>> correctly fails with noSuchObject, and then both slapds sit at 0% CPU.
>> Can you give further details?
>
> I'll look at reproducing it again.
>
> In this particular case, the master had a database rooted at "", with
> some  data around "cn=zimbra" contained in it.  A replica was configured
> as a  syncrepl client, but with "dc=example,dc=com" as its base, by
> mistake.  This resulted in the CPU usage shooting to 100% on the replica,
> and the  logs showing it trying to add without success.

Okay, I figured it out -- This is definitely a misconfiguration type 
scenario...

The database is something like:

database bdb
base "dc=example,dc=com"

...

Then the syncrepl config has:
    searchbase=""

I'm not sure this is something that should be allowed -- where the search 
base is broader than the database.

--Quanah

--
Quanah Gibson-Mount
Principal Software Engineer
Zimbra, Inc
--------------------
Zimbra ::  the leader in open source messaging and collaboration