#417 Fault start at gtm_standby w/o connection with GTM

[Markn]

gtm_standby started without error response although gtm_standby has failed to establish connection with gtm active from gtm.log

A issue dissusced at following thread.
[Postgres-xc-general] GTM standby behaivor when cannot connect with GTM active

[Markn]

I have tried stop immediate in this situation but could not stop gtm_standby but cannot stop gtm_standby ....
Although gtm_standby is waiting for GTM backup, I guess gtm_standby needs to be shutting down when immediate shutdown.

> (4) Stop gtm_stadby

immediate stop to gtm_standby but failed

$ gtm_ctl -D gtm/data/ -Z gtm start
server starting
$ gtm_ctl -D gtm/data/ -Z gtm stop -m immediate waiting for server to shut down................
............................................... failed
gtm_ctl: server does not shut down

[Markn]

I have also observed connection error message in gtm slave although there're no network failure and gtm master is running.

[Koichi Suzuki]

I found at present there're no difference from -m immediate and -m smart. In the case of -m immediate, gtm_ctl sends SIGQUIT to gtm, but the handling is the same as in the case of -m smart, SIGTERM case.

Because gtm_ctl sends different signals for different -m option, gtm should terminate differently as:

-m smart should be the same.
-m fast should backup the restart point and exit.
-m immediate should stop all at once.

In -m immediate, after 1.1, restart point has been backed up so there could be some (up to 2000) waste of GXID and sequence values.

Because GTM restart point backup was added in 1.1, above improvement could be done in 1.1 and later.