[389-users] 389 directory server crash

Tue Jul 16 14:49:30 UTC 2013

On 07/16/2013 01:23 AM, Mitja Mihelič wrote:
> On 07/15/2013 05:28 PM, Rich Megginson wrote:
>> On 07/15/2013 02:57 AM, Mitja Mihelič wrote:
>>> On 07/12/2013 05:55 PM, Rich Megginson wrote:
>>>> On 07/12/2013 08:22 AM, Mitja Mihelič wrote:
>>>>> On 07/09/2013 03:34 PM, Rich Megginson wrote:
>>>>>> On 07/09/2013 06:43 AM, Mitja Mihelič wrote:
>>>>>>> Hi!
>>>>>>>
>>>>>>> We are having problems with some our 389-DS instances. They 
>>>>>>> crash after receiving an update from the provider.
>>>>>>
>>>>>> After looking at the stack trace, I think this is 
>>>>>> https://fedorahosted.org/389/ticket/47391
>>> Yes, it looks like it might be it. When CONSUMER_ONE crashed for the 
>>> first time, the last thing replicated was a password change.
>>> Do you perhaps know, where I could get a 389DS version for Centos6 
>>> that has the patch? The ticket says it was pushed to 1.2.11, but 
>>> would seem that our 1.2.11.15-14 is still an unpatched one and the 
>>> repositories do not have any newer versions.
>>
>> Is that the 389-ds-base that is included with CentOS6?
> Yes, the 389-ds-base-1.2.11.15-14.el6_4.x86_64 and 
> 389-ds-base-libs-1.2.11.15-14.el6_4.x86_64 are from the official 
> Centos6 updates repoository.
> 389-ds-base-debuginfo is from http://debuginfo.centos.org/6/
> The rest are from epel.

Looking at the stack trace you sent earlier - there is only 1 thread?  
You ran

gdb -ex 'set confirm off' -ex 'set pagination off' -ex 'thread apply all bt full' -ex 'quit' /usr/sbin/ns-slapd `pidof ns-slapd` > stacktrace.`date +%s`.txt 2>&1

?  If so, I have no idea what's going on - I've never seen the server deadlock itself with only 1 thread . . .

>>
>>>>>>
>>>>>>> The crash happened twice after about a week of running without 
>>>>>>> problems. The crashes happened on two consumer servers but not 
>>>>>>> at the same time.
>>>>>>> The servers are running CentOS 6x with the following 389DS 
>>>>>>> packages installed:
>>>>>>> 389-ds-console-doc-1.2.6-1.el6.noarch
>>>>>>> 389-console-1.1.7-1.el6.noarch
>>>>>>> 389-adminutil-1.1.15-1.el6.x86_64
>>>>>>> 389-dsgw-1.1.10-1.el6.x86_64
>>>>>>> 389-ds-base-debuginfo-1.2.11.15-14.el6_4.x86_64
>>>>>>> 389-admin-1.1.29-1.el6.x86_64
>>>>>>> 389-ds-console-1.2.6-1.el6.noarch
>>>>>>> 389-admin-console-doc-1.1.8-1.el6.noarch
>>>>>>> 389-ds-1.2.2-1.el6.noarch
>>>>>>> 389-ds-base-1.2.11.15-14.el6_4.x86_64
>>>>>>> 389-ds-base-libs-1.2.11.15-14.el6_4.x86_64
>>>>>>> 389-admin-console-1.1.8-1.el6.noarch
>>>>>>>
>>>>>>> We are in the process of replacing the Centos 5x base 
>>>>>>> consumer+provider setup with a CentOS 6x base one. For the time 
>>>>>>> being, the CentOS 6 machines are acting as consumers for the old 
>>>>>>> server. They run for a while and then the replicated instances 
>>>>>>> crash though not at the same time.
>>>>>>> One of the servers did not want to start after the crash,
>>>>>>
>>>>>> Can you provide the error messages from the errors log?
>>>>> I have attached error logs from the provider 
>>>>> (2013-06-27-provider_error) and the consumer 
>>>>> (2013-06-27-server_two_error) in question.
>>>>>>
>>>>>>> so I have run db2index on its database. It's been running for 
>>>>>>> four days and it has still not finished. 
>>>>>>
>>>>>> Try exporting using db2ldif, then importing using ldif2db.
>>>>> The export process hangs. After an hour strace still shows:
>>>>> futex(0x7f5822670ed4, FUTEX_WAIT, 1, NULL
>>>>> The error log for this is attached as 
>>>>> 2013-07-10-server_two-ldif_import_hangs.
>>>>
>>>> Are you using db2ldif or db2ldif.pl?  If you are using db2ldif, is 
>>>> the server running?  If not, please try first shutting down the 
>>>> server and use db2ldif.
>>>>
>>>> If db2ldif still hangs, then please follow the instructions at 
>>>> http://port389.org/wiki/FAQ#Debugging_Hangs to get a stack trace of 
>>>> the hung process.
>>> I was using db2ldif with the server shut down. I tried it again and 
>>> it hung. The LDIF file was created but its size was zero. The 
>>> produced stack trace is attached as 
>>> server_two-db2ldif_hang-stacktrace.1373877200.txt.
>>>
>>>>
>>>>>
>>>>>>
>>>>>>> All I get from db2index now are these outputs:
>>>>>>> [09/Jul/2013:13:29:11 +0200] - reindex db: Processed 65095 
>>>>>>> entries (pass 1104) -- average rate 53686277.5/sec, recent rate 
>>>>>>> 0.0/sec, hit ratio 0%
>>>>>>
>>>>>> How many entries do you have in your database?
>>>>> The number revolves around 65400. It varies perhaps 2 user del/add 
>>>>> operations a month and 20 attribute changes per week, if that.
>>>>>>
>>>>>>>
>>>>>>> The other instance did start up, but the replication process did 
>>>>>>> not work anymore. I disabled the replication to this host and 
>>>>>>> set it up again. I chose "Initialize consumer now" and the 
>>>>>>> consumer crashed every time.
>>>>>>
>>>>>> Can provide a stack trace of the core when the server crashes?  
>>>>>> This may be different than the stack trace below.
>>>>> The last provided stack trace was produced at the last server 
>>>>> crash. I will provide another stack trace when CONSUMER_ONE 
>>>>> crashes again. Currently it refuses to crash at initialization 
>>>>> time and keeps running.
>>>>>>
>>>>>>> I have enabled full error logging and could find nothing.
>>>>>>> I have read a few threads (not all, I admit) on this list and 
>>>>>>> http://directory.fedoraproject.org/wiki/FAQ#Debugging_Crashes 
>>>>>>> and tried to troubleshoot.
>>>>>>>
>>>>>>> The crash produced the attached core dump and I could use your 
>>>>>>> help with understanding it. As well as any help with the crash. 
>>>>>>> If more info is needed I will gladly provide it.
>>>>>>>
>>>>>>> Regards, Mitja
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> --
>>>>>>> 389 users mailing list
>>>>>>> 389-users at lists.fedoraproject.org
>>>>>>> https://admin.fedoraproject.org/mailman/listinfo/389-users
>>>>>>
>>>>>
>>>>
>>>
>>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.fedoraproject.org/pipermail/389-users/attachments/20130716/2d01a9db/attachment.html>