On Tue, Jul 02, 2002 at 09:42:18AM -0700, christian void wrote:
> On Tue, 2 Jul 2002, P a u l Guth wrote:
>
> > The response from the coders has been basically "Everything
> > you want you can get from looking in the database."
>
> what if the database goes down?
Well, then our whole system is down and we have worked very hard
to prevent this condition. The database is VCS clustered, and
redundant to the gills. So unfortunately this one is hard to sell.
> code is not perfect. what if, for some unforseen reason, the data in the
> database ends up _not_ being representative of what is going on with the
> system?
Now see this is just brilliant. This is the kind of inherent distrust
between ops and development that leads to healthy systems. I have
forgotten the fundamental truth that coders (and thus code) are not
perfect, so you want to protect yourself from them.
> from an operational point of view, not logging is just silly. plus,
> logging to syslog can help identify interactions between your software and
> the system that might not otherwise be apparant.
Not quite sure I grok what you're saying here. Like because you could
put the app logs into the same place as system logs you might be able
to correlate application events with system events?
> having to interact with a database in order to determine what is going on
> with the system takes too much work, and involves writing additional code
> that in and of itself may be buggy.
Yes, I have been struggling to make this point as well, although I was
presenting it as additional "procedures" since it may just be "run
this SQL query" rather than "run this tool we wrote." But an excellent
point. Logging, due to its simplicity, is very likely to actually
represent what's happening. What's in the database is much more
suspect.
Thank you sir, that was most helpful.
--
___________________________________________________________________
P a u l
[EMAIL PROTECTED]
_______________________________________________
Bits mailing list
[EMAIL PROTECTED]
http://www.sugoi.org/mailman/listinfo/bits
