Re: syncrepl not syncing [ CSN older or equal to ctx ]

[Duncan Brannen <dbb@st-andrews.ac.uk>]

Gavin Henry wrote:
> <quote who="Duncan Brannen">
>   
> > Gavin Henry wrote:
> >     
> > > <quote who="Duncan Brannen">
> > >
> > >       
> > > > Hi,
> > > >       I've got a slave that I think I missed out the -w on the slapadd
> > > > when I rebuilt it recently and it's
> > > > not picking up changes from the master.
> > > >
> > > > What I'm seeing in the log is
> > > >
> > > > slapd[11886]: [ID 502419 local4.debug] Entry
> > > > reqStart=20080311204629.000002Z,cn=accesslog CSN
> > > > 20080311204629.095387Z#000000#000#000000 older or equal to ctx
> > > > 20080311204629.095387Z#000000#000#000000
> > > >
> > > > Every 60 seconds or so. (My retry interval) type is RefreshAndPersist
> > > >
> > > > Is there an easy way to reset the timestamp on the slave / master so
> > > > this fires into life or do I need to reload it?
> > > >
> > > > Slave is 2.4.7 and the master is 2.4.8 so it's due another upgrade soon
> > > > anyway but I'd rather get then synced now and upgrade out of hours.
> > > >
> > > >
> > > >         
> > > What size of db are we talking about?
> > >
> > >       
> > Pretty small, files are around 300Mb total, 150K entries.  It'll only
> > take 20-30min to reload, but it's
> > much longer to get it in/out of the dns round robin we use for load
> > balancing.
> >
> > Thanks,
> >              Duncan
> >
> >
> >     
>
> Hi Duncan,
>
> How did you get on? I never saw your reply until after 5, so presumed you
> did an upgrade out of hours.
>
> Cheers.
>   

Hi Gavin,
            Yes, thanks, I took the server out of the round robin and 
did an upgrade as well as the reload.  I got a similar initial error 
message (maybe this is normal)

phantom slapd[11886]: [ID 502419 local4.debug] Entry 
reqStart=20080314081607.000003Z,cn=accesslog CSN 
20080314081607.575746Z#000000#000#000000 older or equal to ctx 
20080314081607.575746Z#000000#000#000000

but then it started syncing ok.

eg. phantom slapd[11886]: [ID 653192 local4.debug] 
syncprov_search_response: 
cookie=rid=001,csn=20080314112502.347110Z#000000#000#000000


Should it have been possible to do this without an outage, or should it 
have fixed itself as it seems to have this time (the impression I got 
from the manual)?

One point, test 008-Concurrency produces a lot of Invalid Credentials 
messages but doesn't actually fail.  I don't recall this happening with 
2.3.x is this due
to the differing acl requirements?

I now have a different problem with the slave not recognising the 
master's certificate
TLS trace: SSL3 alert write:fatal:unknown CA
I'm wondering if I've a mix of ssl libraries in there someplace (debug 
looks like it's reading the correct directive and the other slaves work 
so it's not openldap)
There goes my afternoon ;)


Cheers,
          Duncan