[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: End of hypocrisy ?



On Thu, 7 Aug 2014 20:13:08 -0700 (PDT)
Rusi Mody <rustompmody@gmail.com> wrote:

> On Friday, August 8, 2014 5:40:01 AM UTC+5:30, Joel Rees wrote:

> > You do understand the chicken-and-egg nature of what you're asking
> > for?

[clip]

> Is the problem absent with text files?
> If one wants to write a text-log one needs a text-file.
> A text file exists on some file-system.

Oh, come on now, the preceding assertion is a little over the top,
don't you think? Every boot I've ever seen has a read/write filesystem
very early in the boot. Now let's look at what you need for a binary
system:

1) The same filesystem as you needed for the text log files.
2) The program to read the logs.
3) Any config that reading program might need.

And, if you're putting it in Postgres, you also need:

4) A functioning Postgres (not always trivial)
5) The right Postgres user and password

#5 Would be at least somewhat challenging if you were forced to boot
the computer from a rescue CD, which happens to me a lot.

> What of those messages that need to be logged before there are any
> filesystems mounted?

Several choices:

1) Keep it in RAM til you have a read-write filesystem up, then write
   it all.

2) Write it to the screen. Relatively little happens before the
   filesystem comes up, anyway.

3) Specify a bunch of CHS (Cylinder Head Sector) sectors for that
   purpose, cover them with a file whose whole purpose is as a
   placeholder, and write the info to the sectors. Later transfer them
   to the text log.

Keep in mind, the time between when you have a read/write filesystem
and the time Postgres is ready to be written is an eternity.

SteveT

Steve Litt                *  http://www.troubleshooters.com/
Troubleshooting Training  *  Human Performance


Reply to: