Firebird's 5.0.4 gbak crash when using parallel workers

68 views
Skip to first unread message

Fabiano SCI

unread,
Aug 20, 2026, 9:46:50 AM (8 days ago) Aug 20
to firebird-support
Hi folks!

I'm using gbak from Firebird 5.0.4.1812 (x64 on Windows) to backup a database with this command:

'C:\Program Files\Firebird\Firebird_5_0\gbak.exe -b "C:\SCI\banco\vsci.SDB" "C:\SCI\banco\vsci_SDB_QUI_1021.sci" -user SYSDBA -password masterkey -zip -verbint 100000 -G -D -PAR 8 -se 127.0.0.1:service_mgr'

The backup starts and runs for a while until it reaches the table TATUMUNICIPAL then it stops with ResultCode = 1. On the firebird.log file at that timestamp there is this information:
INET/inet_error: read errno = 10053, server host = 127.0.0.1, address = 127.0.0.1/3050
Which suggests a crash in the engine IMHO.

When I redo the backup withou the -PAR 8 switch backup runs fine. 

Also, when restoring this backup with -PAR 8 the firebird instance hangs. My only solution is supress -PAR.

This symtom algo occurs before a full backup/restore cycle. It occurs in SuperServer and Classic also.
A copy of this database can be found:
and here:



Fabiano SCI

unread,
Aug 20, 2026, 10:08:02 AM (8 days ago) Aug 20
to firebird-support
"This symtom algo occurs before a full backup/restore cycle" is actually   "This symtom algo occurs after a full backup/restore cycle"

Vlad Khorsun

unread,
Aug 20, 2026, 2:21:36 PM (8 days ago) Aug 20
to firebird-support
Hi folks!

I'm using gbak from Firebird 5.0.4.1812 (x64 on Windows) to backup a database with this command:

'C:\Program Files\Firebird\Firebird_5_0\gbak.exe -b "C:\SCI\banco\vsci.SDB" "C:\SCI\banco\vsci_SDB_QUI_1021.sci" -user SYSDBA -password masterkey -zip -verbint 100000 -G -D -PAR 8 -se 127.0.0.1:service_mgr'

The backup starts and runs for a while until it reaches the table TATUMUNICIPAL then it stops with ResultCode = 1.

Does you read gbak output ? It prints exact reason of failure:

gbak: ERROR:database VSCI.SDB shutdown
gbak:Exiting before completion due to errors
 
On the firebird.log file at that timestamp there is this information:
INET/inet_error: read errno = 10053, server host = 127.0.0.1, address = 127.0.0.1/3050
Which suggests a crash in the engine IMHO.

No, it is not.
 
When I redo the backup withou the -PAR 8 switch backup runs fine.  

Also, when restoring this backup with -PAR 8 the firebird instance hangs. My only solution is supress -PAR.

This symtom algo occurs before a full backup/restore cycle. It occurs in SuperServer and Classic also.
A copy of this database can be found:

The database  is in single-user shutdown mode, run gfix -h and see yourself.
It prevents anyone, including gbak, to run more than one attachment.

Please, be more careful next time ;)

Regards,
Vlad

Fabiano SCI

unread,
Aug 21, 2026, 8:07:38 AM (7 days ago) Aug 21
to firebird-support
So, really that what you saidm thanks!. For various reasons the gbak output was not present to me at that time.
But there is a reason for my database to be in single-user shutdown mode. Since a long time agora Firebird's documentation claim that connecting on a database while restoring a backup can lead to a corruption. So in my case I always put my database in single-user mode to avoid it.
With Firebird 5.0.4 I must care about corruption during restore if someone connects to my database?

Dimitry Sibiryakov

unread,
Aug 21, 2026, 8:09:50 AM (7 days ago) Aug 21
to firebird...@googlegroups.com
Fabiano SCI wrote 21.08.2026 14:07:
> With Firebird 5.0.4 I must care about corruption during restore if someone
> connects to my database?

Database is already in single-user mode during restore. There is no need to
do it explicitly.

--
WBR, SD.

Vlad Khorsun

unread,
Aug 21, 2026, 5:39:25 PM (6 days ago) Aug 21
to firebird-support
So, really that what you saidm thanks!. For various reasons the gbak output was not present to me at that time.
But there is a reason for my database to be in single-user shutdown mode. Since a long time agora Firebird's documentation claim that connecting on a database while restoring a backup can lead to a corruption. So in my case I always put my database in single-user mode to avoid it.

One more myth. 
I can imagine than one might break **gbak restore ** operations, yes (with some bad luck), but it is not about database "corruption".
 
With Firebird 5.0.4 I must care about corruption during restore if someone connects to my database?

If you afraid of such "corruption" - restore database into different name and rename later or/and change
SYSDBA password and hide it from "so evil" users.

Regards,
Vlad

Vlad Khorsun

unread,
Aug 21, 2026, 5:40:25 PM (6 days ago) Aug 21
to firebird-support
Fabiano SCI wrote 21.08.2026 14:07:
> With Firebird 5.0.4 I must care about corruption during restore if someone
> connects to my database?

Database is already in single-user mode during restore.

Not in case of parallel restore.
 
There is no need to do it explicitly.

Sure.

Regards,
Vlad

Reply all
Reply to author
Forward
0 new messages