Re: psql: FATAL: the database system is starting up

On 6/1/19 2:30 PM, Tom K wrote:
> 
> 
> On Sat, Jun 1, 2019 at 4:52 PM Adrian Klaver <adrian.klaver@aklaver.com 
> <mailto:adrian.klaver@aklaver.com>> wrote:
> 
>     On 6/1/19 12:42 PM, Tom K wrote:
>      >
>      >
> 
>      >
>      > Of note are the characters f2W below.  I see nothing in the postgres
>      > source code to indicate this is any recognizable postgres
>     message.  A
>      > part of me suspects that the postgres binaries got corrupted.  
>     Had this
>      > case occur with glib-common and a reinstall fixed it.  However the
>      > postgres binaries csum matches a standalone install perfectly so
>     that
>      > should not be an issue.
> 
>     It comes from timeline.c:
> 
>     https://doxygen.postgresql.org/bin_2pg__rewind_2timeline_8c.html
> 
>     pg_log_error("syntax error in history file: %s", fline);
> 
>     ...
> 
>     There should be another error message after the above.
> 
> 
> Nope.  Here's the full set of lines in the postgres logs when running 
> the above line:
> 

> 2019-06-01 17:13:03.263 EDT [14909] FATAL:  syntax error in history 
> file: f2W
> 2019-06-01 17:13:03.263 EDT [14909] HINT:  Expected a numeric timeline ID.

Actually the above HINT is what I was looking for.

> ^C
> -bash-4.2$
> 
> What's interesting is that f2W isn't a string you'd expect to be printed 
> as part of the code logic   ( I mean, what is f2W? ).

As the HINT said neither was Postgres. That is probably down to file 
corruption.


> 
> The point of the POC and the LAB is to test these things across failures 
> as well as various configurations.  To that end, I'm just as curious how 
> to recover from these error conditions as I am just getting things to work.

I think what it proved was that a single point of failure is not good 
and that there needs to be steps taken to prevent or deal with it e.g. 
second location backup of some sort.

> 
> 
> 
> 
>     -- 
>     Adrian Klaver
>     adrian.klaver@aklaver.com <mailto:adrian.klaver@aklaver.com>
> 


-- 
Adrian Klaver
adrian.klaver@aklaver.com