<html> <head> <meta content="text/html; charset=ISO-8859-1" http-equiv="Content-Type"> </head> <body bgcolor="#FFFFFF" text="#000000"> <div class="moz-cite-prefix">On 06/24/2015 10:02 AM, Christoph Kaminski wrote: </div> <blockquote cite="mid:OF6C873102.B1069DA9-ONC1257E6E.002C1954-C1257E6E.002C2FE2@biotronik.com" type="cite"><tt><a class="moz-txt-link-abbreviated" href="mailto:freeipa-users-bounces@redhat.com">freeipa-users-bounces@redhat.com</a> schrieb am 29.04.2015 17:51:46: > Am 29.04.2015 um 15:43 schrieb Ludwig Krispenz: > > > > On 04/29/2015 03:17 PM, Martin (Lists) wrote: > >> Am 27.04.2015 um 09:45 schrieb Ludwig Krispenz: > >>> On 04/26/2015 10:49 AM, Martin (Lists) wrote: > >>>> Hallo > >>>> > >>>> after a reboot I get almost thousand of the following messages: > >>>> > >>>> DSRetroclPlugin - delete_changerecord: could not delete change record > >>>> 128755 (rc: 32) > >>> this message comes from changeglog trimming and means that an entry, > >>> which should be purged does not exist (any more). > >>> the retrocl maintains a first/lastchange and trinming starts at > >>> firstchange. if for some reason (race ?) there is an attempt to try to > >>> delete the same entry a second time this message should be logged. > >>> since the changenumbers in the error message increases, I think > >>> changelog trimming moves forward. you could do searches on > >>> "cn=changelog" to verify that trimming works. > >> changelog is part of the ldbm database plugin and contains several > >> informations I don't understand (or understand partially). What kind of > >> information should I look for? > > the changelog keeps track of the changes applied to the database, a > > typical entry looks like: > > dn: changenumber=4,cn=changelog > > objectClass: top > > objectClass: changelogentry > > changeNumber: 4 > > targetDn: cn=tuser,ou=people,dc=example,dc=com > > changeTime: 20140411093444Z > > changeType: delete > > OK, I looked in the wrong directory. Now I have found many changelog > entries, starting with number 152926 and ending with 155512 (ldapsearch > states 2588 numEntries). Should that be that much? > > The oldest is about two days and an half old and it does not change > within the last few minutes. > > > > > each entry gets a DN made up from he changenumber, so your entries will > > be named: > > .... > > dn: changenumber=61,cn=changelog > > dn: changenumber=62,cn=changelog > > dn: changenumber=63,cn=changelog > > dn: changenumber=64,cn=changelog > > .... > > changenumbers start and are always incremented, changelog trimming > > removes old entries (depending on config). > > > > so if you do a search like: > > ldapsearch .................. -b "cn=changelog" > > the changenumber of the first entry rerurne should always increase, > > indicating that trimming works. > > As it seems my trimming is broken, at least partially. Is there > something I can adjust? > > > > > you said "thousands" of messages, how frequent are they really ? > > On every reboot I got these messages. I do not get them during normal > opperation. > > Something odd I observed after the last two reboots: ns-slapd runs my > hard disk for several minutes (about 15 minutes) after the reboot. This > is the time it takes to log all these change record messages. > > Kindly > Martin > > -- </tt> <tt>We have the same issue here on some servers... Any solutions for it?</tt> </blockquote> Hi Christoph, If you do a search on 'cn=changelog' what is the number of the first returned entry ? Also what is the first number logged in the error log 'delete_changerecord: could not delete change record xxx'. Did you see those messages after a server restart ? thanks thierry<tt> </tt> <blockquote cite="mid:OF6C873102.B1069DA9-ONC1257E6E.002C1954-C1257E6E.002C2FE2@biotronik.com" type="cite"> <tt>Greetz</tt> <fieldset class="mimeAttachmentHeader"></fieldset> </blockquote> </body> </html>