Hi, Works much better than what?
I have a test environment on Aix 6.1 TL7. In this test environment I copy records from one file to another as long as the journal is full. This test takes almost twice as long on exactly the same environment but with Aix 6.1 TL8. I noticed that if I change sync period to very high value, the measurement result comes back to normal. I'm trying to find the reason of that behavior. Best regards, Piotr W dniu czwartek, 29 stycznia 2015 22:41:35 UTC+1 użytkownik DanE napisał: > > Yes, it is to ensure that logset on disk is within that time period. > > > > Change the time period if it degrades performance. > > > > Works much better than what? > > > > You could change the period to a very high number, I am not sure it can be > turned off. > > > > Dan Ell > > > > > *Dan Ell *Technical Support Engineer > > *jBASE *9245 Research Drive, Irvine, CA 92618 > 949-383-2429 > da...@jbase.com <javascript:> > www.jbase.com > > *From:* jb...@googlegroups.com <javascript:> [mailto: > jb...@googlegroups.com <javascript:>] *On Behalf Of *My_jBase > *Sent:* Wednesday, January 28, 2015 3:02 PM > *To:* jb...@googlegroups.com <javascript:> > *Subject:* Re: TIME BETWEEN LOG FILE SYNCS > > > > Hi, > > Thank You for the answer, > > Does this mean that the frequency and range of journal memory flushes is > determined by the O/S and the every-10-seconds synchronisation is an > additional activity, carried out "just in case" ? > > What should I do if this regular synchronisation degrades performace of > the system ? > > I changed the frequency to 10.000 seconds ( on my test environment, of > course ) so that the system works much better. However, I'm not sure if > it's safe for the data. Is this posible to switch off this sync completely > and use only flushing by O/S ? > > Piotr > > W dniu środa, 28 stycznia 2015 16:42:24 UTC+1 użytkownik DanE napisał: > > The flush is a maximum period of time, not a constant. Most O/S will > flush in milliseconds depending on system load. Unless the transaction > load is huge, this is not normally something that needs to be altered. I > believe it really is more of a safety net thing that never really comes > into play. > > > > What is responsible for flushing is either the process updating the logset > or optionally a separate process running jlogsync. > > > > Dan Ell > > > > *Dan Ell* > Technical Support Engineer > *jBASE* > 9245 Research Drive, Irvine, CA 92618 > 949-383-2429 > da...@jbase.com > www.jbase.com > > > > *From:* jb...@googlegroups.com [mailto:jb...@googlegroups.com] *On Behalf > Of *My_jBase > *Sent:* Wednesday, January 28, 2015 8:03 AM > *To:* jb...@googlegroups.com > *Subject:* TIME BETWEEN LOG FILE SYNCS > > > > Hello, > > > > Would You like to explain me how jbase journal works ? > > The jbase manual says : > > > > When journaling is running the basic order of operations for updates is as > follows: > > • The database item is updated in memory > > *• The transaction log is updated in memory*. > > • The transaction log is flushed every 10 seconds by default. However this > time period can be configured via an option on the administration command, > jlogadmin. It is also possible to configure an independent process to > execute the jlogsync command, to ensure the transaction log is continuously > flushed from memory at the specified interval, thus alleviating the flush > procedure from all of the update processes. > > • The database updates are flushed to disk by the operating system. > > • The database update to disk and the log update to disk can be forced to > be an atomic operation. > > There is a feature which specifies the number of seconds between each > synchronization of the log set with the disk - jlogadmin -s nnn ; All > memory used by the log set is force flushed to disk. Should the system > crash, the maximum amount of possible data loss is limited to the updates > which occurred since the last log set synchronization. By default it is 10 > seconds. What will be if I change this to f.e. 10.000 seconds ? - does that > mean that I can lose informations from 10.000 seconds if my system will > crash ? > > In my opinion it's not true because I can observe modify time of logset > files during a test and they are constantly updated - not in every 10 or > 10.000 seconds > > So what is responsible for logsets synchronization ? > > > > Piotr > > > > -- > -- > IMPORTANT: T24/Globus posts are no longer accepted on this forum. > > To post, send email to jb...@googlegroups.com > To unsubscribe, send email to jbase-un...@googlegroups.com > For more options, visit this group at > http://groups.google.com/group/jBASE?hl=en > > --- > You received this message because you are subscribed to the Google Groups > "jBASE" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to jbase+un...@googlegroups.com. > For more options, visit https://groups.google.com/d/optout. > > -- > -- > IMPORTANT: T24/Globus posts are no longer accepted on this forum. > > To post, send email to jb...@googlegroups.com <javascript:> > To unsubscribe, send email to jbase-un...@googlegroups.com <javascript:> > For more options, visit this group at > http://groups.google.com/group/jBASE?hl=en > > --- > You received this message because you are subscribed to the Google Groups > "jBASE" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to jbase+un...@googlegroups.com <javascript:>. > For more options, visit https://groups.google.com/d/optout. > -- -- IMPORTANT: T24/Globus posts are no longer accepted on this forum. To post, send email to jBASE@googlegroups.com To unsubscribe, send email to jbase-unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/jBASE?hl=en --- You received this message because you are subscribed to the Google Groups "jBASE" group. To unsubscribe from this group and stop receiving emails from it, send an email to jbase+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
