[Date Prev][Date Next]
slapd segfaults in one node of master-master
- To: email@example.com, Chris Purdye <firstname.lastname@example.org>, John Benjamins <JBenjamins@ifdsgroup.com>, Sri Ampalavanar <SAmpalavanar@ifdsgroup.com>
- Subject: slapd segfaults in one node of master-master
- From: Dave Smith <email@example.com>
- Date: Tue, 29 Mar 2011 11:41:19 -0400
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:mime-version:from:date:message-id:subject:to :content-type; bh=NFiaScJrfVPuoxn6StzmCjh4H0Vjs6AxwLE3ahNXPis=; b=vCsM5iYMF555jQmTbmZApwMjKri8X1/xgotpdEcxoSI9o6xEguv/C/CSuXEIEk24kK GMCN0IkYK2Q8RtRaVKqQT6sMEaqh1r/XQ4JAdlNeMybDRbl20gP7Z7MFMJZ/FkNjxWoW 1NOzOn7DtzQSi6le9nrkn2pxgGOs5TuznjZkc=
- Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:from:date:message-id:subject:to:content-type; b=jna4ysSG+IGrcBmRLQPlbIVK2WpGs9rG/A6SHeX6kyObfVTlMRog92KVDggjTbtHDR WUTWqLpLoC3pvfvM+jETAcPt0DpVGted2Ouv6zmNQMTLtu0bxthZGIP/WIy5NCCJnBoI sp4Dyg/M2Npr/wBKlYuzWpTcOoLvbYZCJpORg=
We run version 2.4.16 on two RHEL4 platforms in master-master configuration. One of our nodes regularly reports a segfault and exits. The log is shown below. This is a production platform so we run minimal logging for performance reasons.
We have been running openldap in this configuration for 2 years without a problem. The issues started after we applied as schema change and added more users accounts to the system.
The schema changes we straight-forward. We added a few optional attributes to existing objects and added one new object to our schema. Nothing that required data fixes. Our schema provides an auxiliary user account object that we use with the standard "account" object. We have a number of additional objects that are related user authentication and authorization. We applied these same changes in various development and testing environments without problem.
The other thing we did was add a lot of users accounts. Prior to this release we had some 6000 users accounts. Now we have over 15000 user accounts. We are concerned that our problem may be related to the additional load from these users. We are planning further grow in the user-base. One thing I should mention is that our system is more write-heavy that the NIS replacement system. We change attributes on the account record on every authentication request.
We suspect this might be replication related because when we dump (using slapcat) and compare the two nodes, one user account record is missing from the node with the segfaults. This weekend we plan to rebuild this node from the slapcat output of the other. We hope this will resolve our problem.
Mar 21 00:59:02 msgeuroha slapd: slapd startup succeeded
Mar 21 01:47:29 msgeuroha kernel: slapd: segfault at 0000000000000048 rip 0000003e67507c3c rsp 0000000040e7f318 error 4
Mar 21 01:53:58 msgeuroha slapd: slapd startup succeeded
Mar 21 03:46:54 msgeuroha kernel: slapd: segfault at 0000000000000048 rip 0000003e67507c3c rsp 0000000040e7f318 error 4
Mar 21 03:51:34 msgeuroha slapd: slapd startup succeeded
Mar 21 09:10:34 msgeuroha kernel: slapd: segfault at 0000000000000048 rip 0000003e67507c3c rsp 0000000040e7f318 error 4