Re: [Postgres-xc-bugs] backup/restore problem using BARRIER

[Michael P. <mic...@gm...>]

I have a couple of questions.
This report lacks of details.

System Configuration:
> ---------------------
> Architecture : x86_64
>
> Operating Systems :
> A. CentOS release 6.2 x86_64
> B. RHEL5.7 x86_64
> We tried in two environments. See also "Compilers"
>
> Postgres-XC version : Postgres-XC 1.0beta2
>
> Compilers used :
> A. gcc (GCC) 4.4.6 20110731 (Red Hat 4.4.6-3)
> B. gcc (GCC) 4.1.2 20080704 (Red Hat 4.1.2-51)
>
> Description of problems:
> ==============================================
> We tested online backup/restore feature using BARRIER,
> and found anomalies in restored serial values.
>
> (Outline of procedure)
> 0. Setup all coordinators and datanodes for online backups.
>
> 1. Create a table via coordinator:
> # CREATE TABLE tbl(I serial, j int);
>
> 2. First dataload via coordinator:
> # INSERT INTO tbl(j) VALUES (generate_series(1,10));
>
> 3. Base backup of all coordinators and datanodes:
> $ pg_basebackup -h localhost -p xxxx -D /xxxx
>
> 4. Second dataload via coordinator:
> # INSERT INTO tbl(j) VALUES (generate_series(11,20));
>
> 5. Install BARRIER via coordinator:
> # CREATE BARRIER 'test';
>
> 6. Archive WALs for all coordinators and datanodes:
> # SELECT pg_start_backup('xxx');
> # SELECT pg_stop_backup();
>
Was this launched directly at each node?


>
> 7. Third dataload via coordinator:
> # INSERT INTO tbl(j) VALUES (generate_series(21, 30));
>
> 8. Stop all coordinators and datanodes.
>
> 9. Discard existing database cluster, and resore base backups.
>
> 10. create recovery.conf for all coordinators and datanodes.
>
Just a confirmation:
Did you set up recovery_target_barrier to "test" in each recovery.conf?
I think you have done so but please confirm.

11. Start datanodes and coordinators.
>
> 12. Issue SELECT * ... and observe the result:
> testdb=# SELECT * FROM tbl ORDER BY i;
> i | j
> ----+----
> 1 | 1
> 2 | 2
> 3 | 3
> 4 | 4
> 5 | 5
> 6 | 6
> 7 | 7
> 8 | 8
> 9 | 9
> 10 | 10
> 12 | 11
> 13 | 12
> 14 | 13
> 15 | 14
> 16 | 15
> 17 | 16
> 18 | 17
> 19 | 18
> 20 | 19
> 21 | 20
> (20 rows)

Just by seeing what is written in this report, the data is recovered up to
the  second insert, to the point where the barrier has been created.
So it looks correct.
Am I missing something?


-- 
Michael Paquier
http://michael.otacoo.com