Date: Wed, 11 Nov 2009 16:59:06 +0100 From: Miroslav Lachman <000.fbsd@quip.cz> To: Ivan Voras <ivoras@freebsd.org> Cc: freebsd-geom@freebsd.org Subject: Re: gjournal and calculation of the size of journal provider Message-ID: <4AFADF4A.80404@quip.cz> In-Reply-To: <hdbf04$21k$1@ger.gmane.org> References: <4AF84245.7070108@quip.cz> <hdbf04$21k$1@ger.gmane.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Ivan Voras wrote: > Miroslav Lachman wrote: >> What is the right rule for journal size calculation? >> There are two sources stating different things. >> >> 1] journal size depends on disk write speed >> http://lists.freebsd.org/pipermail/freebsd-fs/2006-June/002016.html >> >> "For example your disk can write >> at 60MB/s. Journal switch time is 10 seconds. The journal provider has >> to have place to keep two journals (active and inactive). So bascially >> you need 60*10*2MB + gjournal headers." >> >> >> 2] journal size depends on RAM size >> http://www.freebsd.org/doc/en/articles/gjournal-desktop/article.html#UNDERSTANDING-JOURNALING >> >> >> "Your RAM size should fit in 30% of the journal provider's space. For >> example, if your system has 1 GB RAM, create an approximately 3.3 GB >> journal provider. (Multiply your RAM size with 3.3 to obtain the size >> of the journal)." >> >> >> What's the right size for journal on 143GB 15k rpm SAS disks on >> machine with 16GB of RAM? Based on second case, it will be more than >> 50 GB - one third of the size of disk. This is insane vasting. > > It really does depend on the speed of drives but it could be > approximated by saying there will not be more data to write than the > size of memory (which is probably wrong since you can write from > /dev/zero indefinitely). The first advice is sufficient, but you should > probably extend the result by 20% to be safer. So is it safe to use 4GB on PERC6 array, which is capable of 150MB/s write speed by dd test? dd if=/dev/zero of=/dev/mfid0s2e bs=1m count=10000 (150 * 10 * 2 * 1.2) = 3600 150 is write speed in MB/s 10 is journal switch time 2 is active + inactive journal 1.2 is +20% to base safer And next question about journal. I saw following message in log after reboot: GEOM_JOURNAL: Journal 1933335573: mfid0s2d contains journal. GEOM_JOURNAL: Journal 1933335573: mfid0s2e contains data. GEOM_JOURNAL: Journal mfid0s2e clean. GEOM_JOURNAL: BIO_FLUSH not supported by mfid0s2d. GEOM_JOURNAL: BIO_FLUSH not supported by mfid0s2e. "BIO_FLUSH not supported" - is it OK to use gjournal on top of the Dell PERC (LSI MegaRAID) with battery backup unit? I think so, but rather ask somebody... :) Miroslav Lachman
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4AFADF4A.80404>