Watch, Follow, &
Connect with Us

Please visit our new home
community.embarcadero.com.

Public Report
Report From: InterBase/Server/Internal/Journal/Short Term Recovery    [ Add a report in this area ]  
Report #:  138129   Status: Open
journal + shutdown causes abnormal termination
Project:  InterBase Build #:  12.0.5.366
Version:    12.0 Submitted By:   Dmitry Kuzmenko
Report Type:  Crash / Data loss / Total failure Date Reported:  8/22/2016 11:18:19 PM
Severity:    Critical / Show Stopper Last Updated: 9/6/2016 3:48:12 PM
Platform:    All platforms Internal Tracking #:  
Resolution: None (Resolution Comments) Resolved in Build: : None
Duplicate of:  None
Voting and Rating
Overall Rating: No Ratings Yet
0.00 out of 5
Total Votes: None
Description

InterBase XE7, update 5, 64bit (tested also with update 6).
Windows 2012 R2 Server 64bit.
Database 40gb. Journaling is on, journal archive is also on.
At some moment, after
gfix -shut -force 0
Interbase crashed, and after that started to crash each
time on connect.

Log contain following entries:
LV-DB1 (Server)
Mon Aug 22 07:33:30 2016

Database: E:\IBASE\INSURANCE.GDB

Short term recovery begin

LV-DB1 (Server)
Mon Aug 22 07:33:30 2016

Database: E:\IBASE\INSURANCE.GDB

Applying journal file: J:\JOURNAL\INSURANCEJR\INSURANCEJOURNAL.2016-08-21T14-43-13Z.3.JOURNAL

LV-DB1 (Server)
Mon Aug 22 07:33:30 2016

Database: E:\IBASE\INSURANCE.GDB

Short term recovery end

LV-DB1 (Server)
Mon Aug 22 07:33:30 2016

process_packet() servicing opcode=19
   An exception occurred that does


not have a description.  Exception number C0000028.

This exception will cause the InterBase server

to terminate abnormally.

LV-DB1 (Client)
Mon Aug 22 07:33:30 2016

C:\Program Files\Embarcadero\InterBase\bin\ibserver.exe: terminated abnormally (-1)

The only way to continue work was
gfix -disable wal dbfile
after that database validation has not shown any error.

Database was moved (backup/restore) from InterBase 2009 server, which had not
this problem.
Steps to Reproduce:
Hard to reproduce. During the following scenario Interbase sometimes crashes,
sometimes not:
- create journal for the database
- create archive journal
- make some changes in db to fill journal
- make gbak -archive_database
- after some time make gfix -shut force 0
InterBase may crash (see description above)
Workarounds
disable journal, which is not acceptable
Attachment
None
Comments

Dmitry Kuzmenko at 9/2/2016 2:34:22 AM -
temporary workaround was used - return (backup/restore) to the ODS 13.2 format. Seems that this problem happens only with ODS 16.

Dmitry Kuzmenko at 9/2/2016 5:36:29 AM -
unfortunately, using ODS 13.2 does not help. It does not crash after shutdown and connect, but the same problem with journal happens after Interbase stop and start.

Dmitry Kuzmenko at 9/5/2016 5:15:50 AM -
Important addition - these databases have LINGER set to 300.
So, the bug happens while linger is working or connections are active - shutdown does not detach connection, linger also keeps cache open, so, stopping InterBase server causes journal stops working on the next connect, as described in initial post.

Sriram Balasubramanian at 9/6/2016 12:21:58 PM -
Dmitry,
Thanks for the report. We are investigating further.

I agree turning off Journaling is not an appealing workaround in this instance.

In the meantime, would it be possible to turn off Linger database. It would help in shutting down the in-memory cache when the database is shutdown. Till we have a fix that can be delivered, I hope turning off Linger would help the customer to continue using Journaling with database shutdown operation.

Dmitry Kuzmenko at 9/6/2016 2:46:18 AM -
tested updates down - 4 and 3, the problem exists in all updates 3-6.

Sriram Balasubramanian at 9/6/2016 3:50:05 PM -
Please do not use QC to report any issues with InterBase anymore. Please use Embarcadero Quality portal for bugs and feature requests, at https://quality.embarcadero.com/browse/IBP

Dmitry Kuzmenko at 9/8/2016 2:04:25 AM -
Sriram, thank you for you comment.
I may recreate report at quality.embarcadero.com, with combined details. And, there (maybe) will be link to the database ready to reproduce the problem.

Server Response from: ETNACODE01