Re: slapd crash on replicate attempt

[Richard Megginson <rmeggins@xxxxxxxxxx>]

Justin Crawford wrote:
> > > One of the slapd processes just crashed again, and the logs on the 
> > > second server show only the same replication error.  I take 
> > >       
> > it to mean 
> >     
> > > the replication can't continue, because the slapd process 
> > >       
> > on the other 
> >     
> > > server has crashed.
> > >  
> > > Anyway, today is the first time in 5 months that either of these 
> > > servers has had any issue whatsoever.
> > >       
> > Something must have changed.  Even something that may appear at first 
> > glance to be innocuous.
> >     
>
> Agreed.  The servers themselves were not patched immediately prior to
> the change; they did get some patches in the week before, but nothing
> that appears to have a direct connection.
>
> We did implement a new authenticating client application during the
> previous week.  But I keep thinking that no client should be able to
> cause a slapd server crash (at least not without some evidence of
> intense load); therefore, a change to client applications does not
> strike me as a likely culprit.
>   

>   
> > Are you using VLV (browsing index in the console)?
> >     
>
> Yes, on a few subtrees, but these tend to be created on the spot by
> administrators, and not managed strictly (should they be?).  I suppose
> one of those could've been created on a subtree at any time, but I don't
> believe one was created in the days immediately prior to the crash.  Why
> do you ask?
>   
We've seen problems with VLV before.
> FYI, we rebooted the host machines last night and restared slapd with
> debug 1.  Neither slapd process crashed today.
>   
Please let us know if you can reproduce the crash.  Running with debug 1 
should help immensly, if you can afford the slowdown - running in debug 
mode can really slow down production machines.  
http://directory.fedora.redhat.com/wiki/FAQ#Troubleshooting

However, we do have a tool available that would allow you to run with 
full debugging in a production environment.  It buffers the log output 
into a circular buffer of the last N lines, and you tell it how many N 
is (e.g. 10,000).  However, the server may perform well enough even with 
debug 1, in which case you probably don't need it.
> Justin
>
> --
> Fedora-directory-users mailing list
> Fedora-directory-users@xxxxxxxxxx
> https://www.redhat.com/mailman/listinfo/fedora-directory-users
>

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature

--
Fedora-directory-users mailing list
Fedora-directory-users@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/fedora-directory-users