[pgpool-hackers: 3210] Re: Implement new feature auto-recovery

[Asaf Ohayon]

I can see now that this design is no good, going to throw away the ping
stuff, instead going to implement: recover_false_primary,
That is, if detach_false_primary is true, And if we detect and mark false
primary, we might as well initiate start_recovery on it (making it valid
standby)

‫בתאריך יום ה׳, 3 בינו׳ 2019 ב-23:59 מאת ‪Asaf Ohayon‬‏ <‪asaf at sysbind.co.il
‬‏>:‬

> Hello
>
> I've started to implement auto-recovery here:
>
> https://github.com/SysBind/pgpool2/commit/bc5c1b4029c569470917b72386623146fdf19415
>
> Seeking for a little advice, done so far:
>
> In health_check main loop, if the node is in degenerated state,
> try to ping it, when it repsonses, (like coming up after reboot),
> execute online-recovery on it.
>
> Issues:
> 1. Altough degenerated backend seems down (e.g: in pcp_node_info),
> in recovery() I get:
> node recovery failed, node id: %d is alive,
>
> 2. Still need some code to mark the node as candidate for auto-recovery,
> and not try to
> establish_persistent_connection to it
>
> Any remarks are welcome
>
> Thanks
>
>
>
> --
>
> Asaf Ohayon
> SRE | SysBind
> M: +972 547920828
> P: + 972 3 5096219
> E: asaf at sysbind.co.il
> Rothschild 60, Bat Yam
> www.sysbind.co.il
>


-- 

Asaf Ohayon
SRE | SysBind
M: +972 547920828
P: + 972 3 5096219
E: asaf at sysbind.co.il
Rothschild 60, Bat Yam
www.sysbind.co.il
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.sraoss.jp/pipermail/pgpool-hackers/attachments/20190106/83be08a3/attachment.html>