[Info-ingres] Weird problem in Ingres 10

Laframboise, André (BAC/LAC) andre.laframboise at canada.ca
Thu May 17 14:33:55 UTC 2018 

Shutdown a client process doesn't always remove the associated DBMS session.

A quick recycle (ingstop/ingstart) should clean this up and rerun the ckpdb as soon as the DBMS is available..


André Laframboise

Conseiller Principal Base de Données, Direction générale de l'Innovation et du Dirigeant principal de l'information
Bibliothèque et Archives Canada / Gouvernement du Canada
andre.laframboise at canada.ca / Tél. : 613-298-1346

Senior Database Advisor, Innovation and Chief Information Officer Branch
Library and Archives Canada / Government of Canada
andre.laframboise at canada.ca / Tel: 613-298-1346

-----Original Message-----
From: info-ingres-bounces at lists.planetingres.org <info-ingres-bounces at lists.planetingres.org> On Behalf Of Mark
Sent: Thursday, May 17, 2018 10:28 AM
To: info-ingres at lists.planetingres.org
Subject: Re: [Info-ingres] Weird problem in Ingres 10

On Thu, 17 May 2018 14:00:02 +0000, Martin Bowes <martin.bowes at ndph.ox.ac.uk> wrote:

>Please remove it. You may get a swathe of data but in this case more may shed some light on the issue.
>
>Are there any other processes you know of which are monkeying with the data?

No.

>Is there a good reason why there is nojournaling on this table? Can we turn it on ... which will entail backing up the database as well.

sw2igt at malfoy 288 % ckpdb +j sw2igt
Thu May 17 15:13:23 2018 CPP: Preparing to checkpoint database: sw2igt
Thu May 17 15:13:28 2018 E_DM110D_CPP_DB_BUSY   Database is currently
being used. Database lock conflict.

Don't know what's accessing the database right now.  All the processes I know about are shut down.

>Marty
>
>-----Original Message-----
>From: Mark [mailto:i at dontgetlotsofspamanymore.net]
>Sent: 17 May 2018 14:54
>To: info-ingres at lists.planetingres.org
>Subject: Re: [Info-ingres] Weird problem in Ingres 10
>
>On Thu, 17 May 2018 13:43:35 +0000, Martin Bowes 
><martin.bowes at ndph.ox.ac.uk> wrote:
>
>>Hi Mark,
>>
>>I note that you are turning off the sc930 trace with a SET TRACE POINT SC930 0 which is still presumably in your ESQLC script. Sadly I think that is masking what is happening here as on the first occasion we lost nearly two minutes of activity, then over a minute on the second and over 4minutes on the third occasion. Could you please remove that from the script and try again.
>
>I wanted to disable the tracing after the problematic statement had 
>finished.  I can remove this if there may be useful data shown 
>otherwise.
>
>>If this table is journaled could you please also check with auditdb 
>>for an insert of the message_id = ' 100055700000XDBW' record after 
>>10:34:06
>
>The table is not journalled.
>
>>Marty
>>
>>-----Original Message-----
>>From: Martin Bowes [mailto:martin.bowes at ndph.ox.ac.uk]
>>Sent: 17 May 2018 14:31
>>To: info-ingres at lists.planetingres.org
>>Subject: Re: [Info-ingres] Weird problem in Ingres 10
>>
>>Whoops sorry, I'll take that back there are 4 queries.
>>
>>The first and the fourth queries are  seemingly identical and the fourth indicates a rowcount 1 on the update.
>>
>>I'll have a closer look.
>>
>>Marty
>>
>>-----Original Message-----
>>From: Martin Bowes [mailto:martin.bowes at ndph.ox.ac.uk]
>>Sent: 17 May 2018 14:25
>>To: Mark; info-ingres at lists.planetingres.org
>>Subject: Re: [Info-ingres] Weird problem in Ingres 10
>>
>>Hi Mark,
>>
>>That trace shows three successive update attempts each with a different message_id. The first one updates no rows the others update a single row.
>>
>>So can you do a:
>>select count(*) from db_msg where message_id = ' 100055700000XDBW' and 
>>queue <> ' log_20150325' /* rowcount = 0 */
>>
>>select count(*) from db_msg where message_id = ' 400055700000XDBW' and 
>>queue <> ' log_20150325' /* rowcount = 1 */
>>
>>select count(*) from db_msg where message_id = ' 500055700000XDBW' and 
>>queue <> ' log_20150325' /* rowcount = 1 */
>>
>>Marty
>>
>>-----Original Message-----
>>From: Mark [mailto:i at dontgetlotsofspamanymore.net]
>>Sent: 17 May 2018 14:17
>>To: info-ingres at lists.planetingres.org
>>Subject: Re: [Info-ingres] Weird problem in Ingres 10
>>
>>On Thu, 17 May 2018 13:03:16 +0000, Martin Bowes 
>><martin.bowes at ndph.ox.ac.uk> wrote:
>>
>>>Sorry, my bad that should have been a line starting: SESSION BEGINS
>>>
>>>Can you send the update queries and parameters (if any) from the good and bad cases. Those won't show any private data, so it should be cool.
>>
>>Cheers.  I have emailed you the logs.
>>
>>>Marty
>>>
>>>-----Original Message-----
>>>From: Mark [mailto:i at dontgetlotsofspamanymore.net]
>>>Sent: 17 May 2018 13:51
>>>To: info-ingres at lists.planetingres.org
>>>Subject: Re: [Info-ingres] Weird problem in Ingres 10
>>>
>>>On Thu, 17 May 2018 12:35:38 +0000, Martin Bowes 
>>><martin.bowes at ndph.ox.ac.uk> wrote:
>>>
>>>>Going out on a limb here...
>>>>
>>>>Are we sure the two sessions are working on the same table owned by the same user in the same database as each other?
>>>
>>>I am only using a single session for this.
>>>
>>>>In the sc930 trace file can you find the connection information immediately prior to the update code. You are looking for a line like:
>>>>BEG 27/03/2018 13:26:07.598545, Database: iidbdb, User: ingres
>>>
>>>There's nothing like this in the trace files.
>>>
>>>>You could also check your database and see if there are any duplicates of the table with the query:
>>>>Select table_owner from iitables where table_name='YOURTABLE';
>>>>
>>>>Do that for each table involved in the update.
>>>
>>>Only 1 row (& only 1 table in the update).
>>>
>>>>Marty
>>>>
>>>>-----Original Message-----
>>>>From: Mark [mailto:i at dontgetlotsofspamanymore.net]
>>>>Sent: 17 May 2018 13:21
>>>>To: info-ingres at lists.planetingres.org
>>>>Subject: Re: [Info-ingres] Weird problem in Ingres 10
>>>>
>>>>On Thu, 17 May 2018 12:06:48 +0000, Martin Bowes 
>>>><martin.bowes at ndph.ox.ac.uk> wrote:
>>>>
>>>>>No error code and with a rowcount of zero means the update is successful, but had no work to do.
>>>>>
>>>>>I presume there is a where clause in the update and that it looks OK.
>>>>
>>>>There is a where clause and it does match a record so there is "work 
>>>>to do".
>>>>
>>>>>Can you compare the code which worked and the code which didn't work for discrepancies?
>>>>
>>>>Yes.  They are exactly the same!  And it only fails the first time 
>>>>it is executed, all other times work fine.
>>>>
>>>>>Marty
>>>>>
>>>>>-----Original Message-----
>>>>>From: Mark [mailto:i at dontgetlotsofspamanymore.net]
>>>>>Sent: 17 May 2018 12:57
>>>>>To: info-ingres at lists.planetingres.org
>>>>>Subject: Re: [Info-ingres] Weird problem in Ingres 10
>>>>>
>>>>>On Thu, 17 May 2018 11:21:37 +0000, Martin Bowes 
>>>>><martin.bowes at ndph.ox.ac.uk> wrote:
>>>>>
>>>>>>Twice...
>>>>>>
>>>>>>Once where it updates the data and once where it puts it back the way it was?
>>>>>
>>>>>No.  The data is the same for each.
>>>>>
>>>>>>What do the Rowcount figures say on each statement?
>>>>>
>>>>>0.
>>>>>
>>>>>>Marty
>>>>>>
>>>>>>-----Original Message-----
>>>>>>From: Mark [mailto:i at dontgetlotsofspamanymore.net]
>>>>>>Sent: 17 May 2018 12:12
>>>>>>To: info-ingres at lists.planetingres.org
>>>>>>Subject: Re: [Info-ingres] Weird problem in Ingres 10
>>>>>>
>>>>>>On Wed, 16 May 2018 12:40:48 -0400, Karl and Betty Schendel 
>>>>>><schendel at kbcomputer.com> wrote:
>>>>>>
>>>>>>>SC930 tracing is server wide and runs until you stop it with set notrace sc930.
>>>>>>>So you can connect to any database, do the set trace record '/some/ingres-writable/dir'
>>>>>>>and the set trace point sc930, go to the trace record directory 
>>>>>>>you gave it and ensure that there's at least something there 
>>>>>>>(should be one or more sessNNNNN files), and then you can run your esql program until the fault happens.
>>>>>>>
>>>>>>>You don't actually need to put the sc930 into the esql program 
>>>>>>>unless you want to do it that way.
>>>>>>>
>>>>>>>Once you capture what you need to capture, set notrace point sc930 turns it off.
>>>>>>>
>>>>>>>You'll need the trace point privilege, simplest is to do it as 
>>>>>>>the installation owner (user ingres, traditionally).
>>>>>>
>>>>>>Thanks again.  We now have traces but they don't help me.  The 
>>>>>>traces include the SQL statement and the data but, pretty much, nothing else.
>>>>>>
>>>>>>The only notable thing I observed is that the data is traced twice 
>>>>>>for the failed transaction and only once for the successful ones.
>>>>>>
>>>>>>There are no error messages.
>>>>>>
>>>>>>>Karl
>>>>>>>
>>>>>>>> On May 16, 2018, at 11:12 AM, Mark <i at dontgetlotsofspamanymore.net> wrote:
>>>>>>>> 
>>>>>>>> On Wed, 16 May 2018 12:11:08 +0000, Martin Bowes 
>>>>>>>> <martin.bowes at ndph.ox.ac.uk> wrote:
>>>>>>>> 
>>>>>>>>> Just to expand on Karl's sc930...You may already know this...
>>>>>>>>> 
>>>>>>>>> To turn it on, make a recording directory...mkdir /full/path/to/directory'
>>>>>>>>> 
>>>>>>>>> And then...
>>>>>>>>> sql iidbdb << SQL_END
>>>>>>>>> set trace record '/full/path/to/directory'; set trace point 
>>>>>>>>> sc930 1; \p\g \q SQL_END
>>>>>>>> 
>>>>>>>> Do I need to run this on the server account?  I have 
>>>>>>>> implemented it currently in the embedded SQL of my program.
>>>>>>>> 
>>>>>>>>> FYI. The digit after the sc930 indicates a tracing level, 1 should be sufficient.
>>>>>>>>> 
>>>>>>>>> Run the errant query.
>>>>>>>>> 
>>>>>>>>> And turn off the sc930.
>>>>>>>>> sql iidbdb << SQL_END
>>>>>>>>> set trace point sc930 0;
>>>>>>>>> \p\g
>>>>>>>>> \q
>>>>>>>>> SQL_END
>>>>>>>>> 
>>>>>>>>> You can now access the recording directory and start scanning the files for any sign of life from your query.
>>>>>>>>> 
>>>>>>>>> Marty
>>>>>>>>> 
>>>>>>>>> -----Original Message-----
>>>>>>>>> From: Karl and Betty Schendel [mailto:schendel at kbcomputer.com]
>>>>>>>>> Sent: 16 May 2018 12:54
>>>>>>>>> To: Ingres and related product discussion forum
>>>>>>>>> Subject: Re: [Info-ingres] Weird problem in Ingres 10
>>>>>>>>> 
>>>>>>>>> On May 16, 2018, at 7:44 AM, Mark <i at dontgetlotsofspamanymore.net> wrote:
>>>>>>>>>> 
>>>>>>>>>> On Tue, 15 May 2018 12:27:09 -0400, Karl and Betty Schendel 
>>>>>>>>>> <schendel at kbcomputer.com> wrote:
>>>>>>>>>> 
>>>>>>>>>>> It's certainly not something I have heard of or seen before.  
>>>>>>>>>>> Do you have any rules defined on the relevant tables?  Try 
>>>>>>>>>>> enabling LOG_TRACE if the problem is sufficiently 
>>>>>>>>>>> predictable, or do a logdump after the problem occurs if 
>>>>>>>>>>> it's not; the idea being to try to see whether you actually 
>>>>>>>>>>> got any PUT (insert) or REP (replace) log records that were then rolled back, or whether the insert / update was never executed at all.
>>>>>>>>>> 
>>>>>>>>>> After enabling log_trace all I got was:
>>>>>>>>>> 
>>>>>>>>>>   LOG: SAVEPOINT   Size written/reserved:      0/     0  Flags:
>>>>>>>>>> -------------------------------------------------------------
>>>>>>>>>> ------
>>>>>>>>> 
>>>>>>>>> So the insert/update isn't ever being executed.  Either it's 
>>>>>>>>> failing with some sort of silent error, which would seem odd, 
>>>>>>>>> or it's not reaching the backend at all, or it's being 
>>>>>>>>> pre-empted by a before rule.  I think the next step would be 
>>>>>>>>> to enable
>>>>>>>>> sc930 tracing and see if the backend is getting the 
>>>>>>>>> insert-update, and what end-of-query status it's recording.  
>>>>>>>>> There should be KB articles on enabling
>>>>>>>>> SC930 tracing.
>>>>>>>>> 
>>>>>>>>> Karl
>>>>>>>>> 
>>>>>>>>> _______________________________________________
>>>>>>>>> Info-ingres mailing list
>>>>>>>>> Info-ingres at lists.planetingres.org
>>>>>>>>> http://lists.planetingres.org/mailman/listinfo/info-ingres
>>>>>>>> 
>>>>>>>> --
>>>>>>>> <insert witty sig here>
>>>>>>>> _______________________________________________
>>>>>>>> Info-ingres mailing list
>>>>>>>> Info-ingres at lists.planetingres.org
>>>>>>>> http://lists.planetingres.org/mailman/listinfo/info-ingres

--
<insert witty sig here>
_______________________________________________
Info-ingres mailing list
Info-ingres at lists.planetingres.org
http://lists.planetingres.org/mailman/listinfo/info-ingres

More information about the Info-ingres mailing list