Postgres HA (com base no wal-toping) falha

Question

Espero que alguém possa me ajudar uma questão de fazer Wal-Thipping e Warm Standby. Meu sistema de espera funciona alegremente por semanas; de repente, começa a procurar arquivos de história que não existem. Em seguida, ele se afasta e eu não posso reiniciá -lo com sucesso sem reconstruir o modo de espera.

Ambos os sistemas estão executando o CentOS 4.5 e o Postgres 8.4.1. Eles usam o NFS para armazenar arquivos WAL da produção no modo de espera.

Um pedaço relevante do tronco, com meus comentários:

[** Recovery is running normally **]

Trigger file            : /tmp/pgsql.trigger
Waiting for WAL file    : 00000001000000830000005B
WAL file path           : /var/tafkan_backup_from_db1/00000001000000830000005B
Restoring to            : pg_xlog/RECOVERYXLOG
Sleep interval          : 2 seconds
Max wait interval       : 0 forever
Command for restore     : cp "/var/tafkan_backup_from_db1/00000001000000830000005B" "pg_xlog/RECOVERYXLOG"
Keep archive history    : 00000001000000830000004D and later
WAL file not present yet. Checking for trigger file...
WAL file not present yet. Checking for trigger file...
WAL file not present yet. Checking for trigger file...
running restore         : OK

Trigger file            : /tmp/pgsql.trigger
Waiting for WAL file    : 00000001000000830000005B
WAL file path           : /var/tafkan_backup_from_db1/00000001000000830000005B
Restoring to            : pg_xlog/RECOVERYXLOG
Sleep interval          : 2 seconds
Max wait interval       : 0 forever
Command for restore     : cp "/var/tafkan_backup_from_db1/00000001000000830000005B" "pg_xlog/RECOVERYXLOG"
Keep archive history    : 000000000000000000000000 and later
running restore         : OK

[** All of a sudden it starts looks for .history files **]

Trigger file            : /tmp/pgsql.trigger
Waiting for WAL file    : 00000002.history
WAL file path           : /var/tafkan_backup_from_db1/00000002.history
Restoring to            : pg_xlog/RECOVERYHISTORY
Sleep interval          : 2 seconds
Max wait interval       : 0 forever
Command for restore     : cp "/var/tafkan_backup_from_db1/00000002.history" "pg_xlog/RECOVERYHISTORY"
Keep archive history    : 000000000000000000000000 and later
running restore         :cp: cannot stat `/var/tafkan_backup_from_db1/00000002.history': No such file or directory
cp: cannot stat `/var/tafkan_backup_from_db1/00000002.history': No such file or directory
cp: cannot stat `/var/tafkan_backup_from_db1/00000002.history': No such file or directory
cp: cannot stat `/var/tafkan_backup_from_db1/00000002.history': No such file or directory
not restored
history file not found
Trigger file            : /tmp/pgsql.trigger
Waiting for WAL file    : 00000001.history
WAL file path           : /var/tafkan_backup_from_db1/00000001.history
Restoring to            : pg_xlog/RECOVERYHISTORY
Sleep interval          : 2 seconds
Max wait interval       : 0 forever
Command for restore     : cp "/var/tafkan_backup_from_db1/00000001.history" "pg_xlog/RECOVERYHISTORY"
Keep archive history    : 000000000000000000000000 and later
running restore         :cp: cannot stat `/var/tafkan_backup_from_db1/00000001.history': No such file or directory
cp: cannot stat `/var/tafkan_backup_from_db1/00000001.history': No such file or directory
cp: cannot stat `/var/tafkan_backup_from_db1/00000001.history': No such file or directory
cp: cannot stat `/var/tafkan_backup_from_db1/00000001.history': No such file or directory
not restored
history file not found

[** I stopped Postgres, renamed recovery.done to recovery.conf, and restarted it. **]

Trigger file            : /tmp/pgsql.trigger
Waiting for WAL file    : 00000002.history
WAL file path           : /var/tafkan_backup_from_db1/00000002.history
Restoring to            : pg_xlog/RECOVERYHISTORY
Sleep interval          : 2 seconds
Max wait interval       : 0 forever
Command for restore     : cp "/var/tafkan_backup_from_db1/00000002.history" "pg_xlog/RECOVERYHISTORY"
Keep archive history    : 000000000000000000000000 and later
running restore         :cp: cannot stat `/var/tafkan_backup_from_db1/00000002.history': No such file or directory
cp: cannot stat `/var/tafkan_backup_from_db1/00000002.history': No such file or directory
cp: cannot stat `/var/tafkan_backup_from_db1/00000002.history': No such file or directory
cp: cannot stat `/var/tafkan_backup_from_db1/00000002.history': No such file or directory
not restored
history file not found
Trigger file            : /tmp/pgsql.trigger
Waiting for WAL file    : 0000000200000083000000A2
WAL file path           : /var/tafkan_backup_from_db1/0000000200000083000000A2
Restoring to            : pg_xlog/RECOVERYXLOG
Sleep interval          : 2 seconds
Max wait interval       : 0 forever
Command for restore     : cp "/var/tafkan_backup_from_db1/0000000200000083000000A2" "pg_xlog/RECOVERYXLOG"
Keep archive history    : 000000000000000000000000 and later
WAL file not present yet. Checking for trigger file...
WAL file not present yet. Checking for trigger file...
WAL file not present yet. Checking for trigger file...
WAL file not present yet. Checking for trigger file...

[** This file is not present. All WAL files start with 00000001. **] 

Alguma ideia? Eu nem sei o que são os arquivos da história e a documentação (principalmente excelente) não é muito clara sobre isso.

Ps. Eu gostaria de estar executando VMs para que eu pudesse usar Texto do link E não precisa se preocupar com nada disso no nível do aplicativo HA NOnsense :-)

ATUALIZAÇÃO: Aqui estão alguns dos logs do servidor em espera sobre esse momento. Parece que algo fez o servidor parar de se recuperar e ficar online, mas não sei o quê. Tenho certeza de que nada poderia ter criado o arquivo de gatilho.

2010-01-20 03:30:15 EST 4b3a5c63.401b LOG:  restored log file "00000001000000830000005A" from archive
2010-01-20 03:30:23 EST 4b3a5c63.401b LOG:  restored log file "00000001000000830000005B" from archive
2010-01-20 03:30:23 EST 4b3a5c63.401b LOG:  record with zero length at 83/5BFA2FF8
2010-01-20 03:30:23 EST 4b3a5c63.401b LOG:  redo done at 83/5BFA2FAC
2010-01-20 03:30:23 EST 4b3a5c63.401b LOG:  last completed transaction was at log time 2010-01-20 03:28:04.594399-05
2010-01-20 03:30:25 EST 4b3a5c63.401b LOG:  restored log file "00000001000000830000005B" from archive
2010-01-20 03:30:37 EST 4b3a5c63.401b LOG:  selected new timeline ID: 2
2010-01-20 03:30:49 EST 4b3a5c63.401b LOG:  archive recovery complete
2010-01-20 03:30:59 EST 4b3a5c62.4019 LOG:  database system is ready to accept connections

Answer 1

Consegui resolver esse problema atualizando os sistemas operacionais do CentOS nos meus dois servidores PostGresql. Portanto, acho que isso foi um sintoma de um bug de rede subjacente de algum tipo.

Answer 2

Uma abordagem totalmente diferente para HA pode ser hospedar o banco de dados PG em um Drbd dispositivo compartilhado entre as duas máquinas.

Answer 3

Você usou seu próprio script/programa de recuperação? Se sim - por favor, não faça isso. Use PG_STANDBY do PostgreSQL Contrib.

Caso contrário - basta ignorar os arquivos .history.

Answer 4

Sua cópia replicada ficou online em algum momento. "00000002.History" está procurando um arquivo de histórico para a linha do tempo 00000002, enquanto o restante dos seus logs inicia com 00000001, que é a linha do tempo original.

Eu verificaria seus logs PostGresql logo antes de começar a procurar o arquivo de histórico para ver se há alguma indicação de que o banco de dados fique online, mesmo por um momento.