[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: syncrepl'd MOD deleted entry (ITS#4365)

Aaron Richton wrote:
> I just deleted a different entry, and it remained on the slaves for 10
> minutes (way too long given lack of load/configuration). I'll be taking
> the next few minutes installing dbx on the master now...takeaway point is
> if nobody else is seeing this, I should be able to repro at will. (I mean,
> it happened twice, it must happen every time? :)
It may be illustrative for you to run test017 with debug -1 to see what 
is normally happens. Note that the syncrepl master has its sessionlog 
commented out in the distributed data, you'll have to uncomment it 
before running the test.

With the sessionlog in place, after a delete on a master, the syncrepl 
consumer should get an Intermediate response to its search request, 
before any other responses. This intermediate response is a SyncID set, 
containing the binary UUIDs of all entries deleted since its last 
update. So you should see a send_ldap_intermediate on the master's log 
with this information, and a corresponding ldap_parse_intermediate on 
the consumer. You should also see "do_syncrep2: LDAP_RES_INTERMEDIATE - 
SYNC_ID_SET" in the log, followed by an equality search for that 
entryUUID, and then a delete request on that entry.

After you've looked that over, you can look thru the logs on your 
provider and consumer and try to identify the corresponding points, that 
should give a better idea of what is or isn't happening.

  -- Howard Chu
  Chief Architect, Symas Corp.  http://www.symas.com
  Director, Highland Sun        http://highlandsun.com/hyc
  OpenLDAP Core Team            http://www.openldap.org/project/