[pgpool-general: 4904] Re: failover command return new_master_id -1

[Daniel Huhardeaux]

Other info: tooi-as-0005 who his shown as MASTER has delegate IP who is 
coonected to the read-only postresql server which is the one from the 
standby server!

My test id following: 3 sessions connected to delegate_IP port 5432, 
master server port 55432 standby server port 55432. I tried to drop a 
table in my test database on delegate IP which give me on the terminal

WARNING:  packet kind of backend 1 ['E'] does not match with 
master/majority nodes packet kind ['C']
FATAL:  failed to read kind from backend
DETAIL:  kind mismatch among backends. Possible last query was: "drop 
table moi;" kind details are: 0[C] 1[E: cannot execute DROP TABLE in a 
read-only transaction]
HINT:  check data consistency among db nodes
server closed the connection unexpectedly
         This probably means the server terminated abnormally
         before or while processing the request.
The connection to the server was lost. Attempting reset: Succeeded.

and in logs from standby server

2016-08-12 17:30:19 CEST [14434-1] postgres at test ERROR:  cannot execute 
DROP TABLE in a read-only transaction
2016-08-12 17:30:19 CEST [14434-2] postgres at test STATEMENT:  drop table 
moi3;

I did the same on both other sessions which are connected to postgresql 
server, only the one from the master server accepted to drop the table.

How is this possible and how to tell to pgpool delegate IP to always 
connect to read/write database ?

Le 12/08/2016 à 16:10, Daniel Huhardeaux a écrit :
> Hi,
>
> I finally got some script to be executed by setting value
> health_check_period (and sr_check_period in master/slave mode). I know
> put master_slave mode = off and switch back to replication mode = on
>
> Command which are executed: failover_command, wd_escaladation_command
> and wd_de_escaladation_command. To get this result I created a small
> script which is executed in each command stanza from pgpool.conf
> creating a file in /tmp containing the name of the command who executed
> this script.
>
> This means that there is no recovery mode available (stage 1 nor 2) :(
>
> There is also problem with pcp_watchdog_info and values returned by
> failover.sh. Here is an output after having restarted the 2 servers.
> tooi-as-0005 was master before the reboot, the postgresql server on it
> was stopped. tooi-as-0006 never became the new master despite the fact
> that the trigger file was created! I then restarted tooi-as-0005 then
> tooi-as-006.
>
> Hereunder the datas received from failover.sh command. How should I
> interprete this?
>
> vendredi 12 août 2016, 14:47:10 (UTC+0200)
> Failed node ID .... 1
> Failed Hostname ... tooi-as-0006
> Failed_port ....... 55432
> Failed_db_cluster . /var/lib/postgresql/9.5/main
> New_master_ID ..... -1
> Old_master_ID ..... 1
> New_master_hostname
> Old_primary_node_ID 1
>
> pcp_watchdog_info on tooi-as-0005 who took master state
>
> 2 YES Linux_TOOI-AS-0005_5432 tooi-as-0005
>
> Linux_TOOI-AS-0005_5432 tooi-as-0005 5432 9000 4 MASTER
> Linux_TOOI-AS-0006_5432 tooi-as-0006 5432 9000 7 STANDBY
>
>
> In the standby postgres log I see after reboot a lots of
>
> 2016-08-12 14:47:12 CEST [3475-1] WARNING:  archive_mode enabled, yet
> archive_command is not set
>
> This means that standby server is not in standby mode but in standalone.
> Right after I run my pg_masterSynchro.sh script which backup from master
> and create the recovery.conf file, standby server started to do the
> right job. How is this possible ?
>
> Thanks for your support
>
> Regards
>
> Daniel
>


-- 
TOOTAi Networks