This is a discussion on oracle redo log transaction within the Oracle Database forums, part of the Database Server Software category; --> hi, i've read the guides from oracle but don't understand fully how redo logs works exactly. i'm coming from ...
| |||||||
| FAQ | Members List | Calendar | Search | Today's Posts | Mark Forums Read |
| ||||
| hi, i've read the guides from oracle but don't understand fully how redo logs works exactly. i'm coming from sqlserver world and i know i can't make a comparaison but .... I know that when a redolog is full, the LGWR do a log switch but what occrurs when a log are switched ? LGWR call DBWR to applies change from the current redolog (which will be switched) to the datafiles or/and the rollback segment ? then DBWR notifes LGWR that it can resume ? Now after the second log switch the first redo log will be rewritten and if i issue a rollback, from where the data are taken to rollback? from the rollback segment ? what process write to the rollback segment ? And if the DB crash, since the statement is not finished there will be a rollback ? why from a performance view, no one recommend a important redolog size ? > 50 Mo and onl 3 to 5 files ? Let's take a update statement on a big table (10 millions rows), if i have only 2 redo logs about 10 Mo each. How oracle does not return an error that redolog is full (like in mssql) ? is this the size of the rollback segment who actually limit the size of the transaction ? thx tom (apologies for my english) |
| |||
| "orauser" <test@test.fr> a écrit dans le message de news:cn21vi$3rj$1@aphrodite.grec.isp.9tel.net... > hi, > i've read the guides from oracle but don't understand fully how redo logs > works exactly. > i'm coming from sqlserver world and i know i can't make a comparaison but > ... > > I know that when a redolog is full, the LGWR do a log switch but what > occrurs when a log are switched ? > LGWR call DBWR to applies change from the current redolog (which will be > switched) to the datafiles or/and the rollback segment ? > then DBWR notifes LGWR that it can resume ? > > Now after the second log switch the first redo log will be rewritten and if > i issue a rollback, from > where the data are taken to rollback? from the rollback segment ? > what process write to the rollback segment ? > And if the DB crash, since the statement is not finished there will be a > rollback ? > > why from a performance view, no one recommend a important redolog size ? > > 50 Mo and onl 3 to 5 files ? > > > Let's take a update statement on a big table (10 millions rows), if i have > only 2 redo logs about 10 Mo each. > How oracle does not return an error that redolog is full (like in mssql) ? > is this the size of the rollback segment who actually limit the size of the > transaction ? > > > thx > tom (apologies for my english) > > Maybe you have to first read the first chapter of Concepts : Introduction to Oracle Server: http://download-west.oracle.com/docs...ntro.htm#20385 And maybe the whole book is useful. -- Regards Michel Cadot |
| |||
| > Maybe you have to first read the first chapter of Concepts : Introduction to Oracle Server: > http://download-west.oracle.com/docs...ntro.htm#20385 > > And maybe the whole book is useful. > and maybe you're kidding me next time answer me www.google.com it's shorter have a good day |
| |||
| "orauser" <test@test.fr> wrote in message news:cn2h1d$b6i$1@aphrodite.grec.isp.9tel.net... > > > Maybe you have to first read the first chapter of Concepts : Introduction > to Oracle Server: > > > http://download-west.oracle.com/docs...ntro.htm#20385 > > > > And maybe the whole book is useful. > > > and maybe you're kidding me > next time answer me www.google.com it's shorter > > have a good day > > No, he's serious. All of your questions are answered there (and the ones that are sure to follow from the answers). You won't regret it. |
| |||
| orauser wrote: > >> Maybe you have to first read the first chapter of Concepts : Introduction > to Oracle Server: >> > http://download-west.oracle.com/docs...ntro.htm#20385 >> >> And maybe the whole book is useful. >> > and maybe you're kidding me > next time answer me www.google.com it's shorter > > have a good day Any answer through google and any answer in this group would be longer (due to followup to resolve confusion), more involved (due to arguments), and likely less accurate (due to myths) than what you would get by reading the book. The advice to RTFM is, at least in this case, justified. /Hans |
| |||
| "orauser" <test@test.fr> wrote: > hi, > i've read the guides from oracle but don't understand fully how redo logs > works exactly. > i'm coming from sqlserver world and i know i can't make a comparaison but > ... > > I know that when a redolog is full, the LGWR do a log switch but what > occrurs when a log are switched ? > LGWR call DBWR to applies change from the current redolog (which will be > switched) to the datafiles or/and the rollback segment ? Yes. It is called a checkpoint. > then DBWR notifes LGWR that it can resume ? Well, kind of. LGWR tells DBWR to start a checkpoint for the log being switched out of, but LGWR doesn't wait for that checkpoint to complete. It waits for the completion of the checkpoint that was started a while ago, back when the log that is now being switched *into* was being switched out of. > > Now after the second log switch the first redo log will be rewritten and > if i issue a rollback, from > where the data are taken to rollback? from the rollback segment ? Yes, of course. > what process write to the rollback segment ? Any process that does anything that might need to be rolled back. > And if the DB crash, since the statement is not finished there will be a > rollback ? Yes. > > why from a performance view, no one recommend a important redolog size ? > > 50 Mo and onl 3 to 5 files ? > > Let's take a update statement on a big table (10 millions rows), if i > have only 2 redo logs about 10 Mo each. > How oracle does not return an error that redolog is full (like in mssql) > ? Once the checkpoint is complete, the data protected by the redo log (Both regular data and rollback data) is already stored on disk the ordinary and rollback segments, so the redo is no longer needed for instance recovery, and can be overwritten. (Of course, it is still necessary for media recovery, which is why it needs to be archived before being overwritten). > is this the size of the rollback segment who actually limit the size of > the transaction ? Yes. Xho -- -------------------- http://NewsReader.Com/ -------------------- Usenet Newsgroup Service $9.95/Month 30GB |
| ||||
| "orauser" <test@test.fr> wrote in message news:<cn2h1d$b6i$1@aphrodite.grec.isp.9tel.net>... > > Maybe you have to first read the first chapter of Concepts : Introduction > to Oracle Server: > > > http://download-west.oracle.com/docs...ntro.htm#20385 > > > > And maybe the whole book is useful. > > > and maybe you're kidding me > next time answer me www.google.com it's shorter > > have a good day REALLY, RTFM! http://download-west.oracle.com/docs...neredo.htm#273 See this, too: http://members.cox.net/oracleunix/readme-cdos.htm jg -- @home.com is bogus. http://davidguy.brinkster.net/computer/009.html |