Re: Are logs any use in a no logging database?

>If I'm wrong! the expert's will pipe it in right. Maybe I count as a piping expert, but don't call me if your drains are blocked. >Date: Thu, 23 May 1996 01:39:35 -0500 (CDT) >From: Cheryl Kendricks <cherylk@prod1.jcdc.doleta.gov> >X-Informix-List-Id: <list.9980> > >1. No! Agreed. >2. Writes always occur to the log for certain commands (mostly all >update/alters/creates). The log is used to record disk space allocations for any purpose, primarily so that if, for example, your 10,000 row 'INSERT INTO T1 SELECT * FROM T2' statement fails part way through, it can be undone. Adding a new extent to a table because of INSERT or UPDATE activity, or creating new tables will be logged. >It will write to the log, yet if your log is point to /dev/null (that's >like never never land) no real write occurs only disk access. The log never points to /dev/null (unless your rootdbs is /dev/null! :-) Actually, it is the log archive device points to /dev/null, which indicates to OnLine that you don't want the logs backed up, so the logs can be recycled for re-use without any further work when necessary. >Yet the engine does count it as an write!!!! Because it does actually write to the logical log in the dbspace where the logs are located (probably the root dbspace). >It does in other words access /dev/null. It doesn't, but it isn't critical either way. >If I'm wrong! the expert's will pipe it in right. > > >On Wed, 22 May 1996 Spetc@aol.com wrote: > >> We do not use logging on our database. >> >> 1. We perform level 0,1 archives, but we don't backup the logs, do we really >> need to? >> >> 2. What are the logs used for in a no logging database? For example, if I >> update every row in a table with 10 million rows with a single update >> statement and I interrupted it half way through would the first 5 million >> updates be rolled back? Would it use the logs for this purpose? Yours, Jonathan Leffler (johnl@informix.com) #include <disclaimer.h>