Just looking through the 2.4.20pre changelogs, I see there's been a bunch of Ext3 patches since 2.4.19. I'm keen to take advantage of any bug fixes and performance improvements, but am nervous about using 2.4.20 until it's completed its release cycle. Is there any way to get a patch just containing the Ext3 changes? Are there any changes since 2.4.20 that are very important, either for reliability or performance, for data=journal mode? Does the htree patch already include these changes? If not, I wonder whether the reason the Ext3 developers can't replicate the htree directory corruption problem is because it's fixed by one of these patches; is that possible?
On Thu, Oct 10, 2002 at 10:48:03PM +0000, JP Howard wrote:> Just looking through the 2.4.20pre changelogs, I see there's been a bunch > of Ext3 patches since 2.4.19. I'm keen to take advantage of any bug fixes > and performance improvements, but am nervous about using 2.4.20 until > it's completed its release cycle.Same here. But then releasing a Kernel doesn't necessarily mean it's stable... -- Ralf Hildebrandt (Im Auftrag des Referat V a) Ralf.Hildebrandt@charite.de Charite Campus Mitte Tel. +49 (0)30-450 570-155 Referat V a - Kommunikationsnetze - Fax. +49 (0)30-450 570-916 Unfortunately, law makers don't believe in the laws of physics or mathematics, only their own laws. When will the emperor discover that he has no clothes?
On Friday 11 October 2002 00:48, JP Howard wrote: Hi Jp,> Is there any way to get a patch just containing the Ext3 changes? Are > there any changes since 2.4.20 that are very important, either for > reliability or performance, for data=journal mode? Does the htree patch > already include these changes? If not, I wonder whether the reason the > Ext3 developers can't replicate the htree directory corruption problem is > because it's fixed by one of these patches; is that possible?I have patches available for 2.4.18 and 2.4.19 with the latest ext3 code (2.4.20-pre10 status) + small fixes and latest htree patches from Theodore and also ext3 online resizing support. Interrested? -- Kind regards Marc-Christian Petersen http://sourceforge.net/projects/wolk PGP/GnuPG Key: 1024D/569DE2E3DB441A16 Fingerprint: 3469 0CF8 CA7E 0042 7824 080A 569D E2E3 DB44 1A16 Key available at www.keyserver.net. Encrypted e-mail preferred.
On Fri, 11 Oct 2002 09:22:50 +0200, "Marc-Christian Petersen" <m.c.p@gmx.net> said:> I have patches available for 2.4.18 and 2.4.19 with the latest ext3 code > (2.4.20-pre10 status) + small fixes and latest htree patches from > Theodore and also ext3 online resizing support. > > Interrested? >Sounds good. Is that one of the files on your SF page? If not, could you upload it some place? What are the fixes included? Ext3 online resizing is still at pretty early stages isn't it? Is it ready for production use? Thanks, Jeremy
Hi, On Thu, Oct 10, 2002 at 10:48:03PM +0000, JP Howard wrote:> Just looking through the 2.4.20pre changelogs, I see there's been a bunch > of Ext3 patches since 2.4.19. I'm keen to take advantage of any bug fixes > and performance improvements, but am nervous about using 2.4.20 until > it's completed its release cycle.I've pushed all my 2.4.20 ext3 changes to http://people.redhat.com/sct/patches/ext3-2.4/for-2.4.19/ There's a single patch at http://people.redhat.com/sct/patches/ext3-2.4/for-2.4.19/all-in-one.patch and the subdirectories contain each changeset broken down and documented. This brings 2.4.19's ext3 up to the 0.9.19 version in 2.4.20. Cheers, Stephen
On Fri, 11 Oct 2002 17:25:12 +0100, "Stephen C. Tweedie" <sct@redhat.com> said:> I've pushed all my 2.4.20 ext3 changes to > > http://people.redhat.com/sct/patches/ext3-2.4/for-2.4.19/ > > There's a single patch at > > http://people.redhat.com/sct/patches/ext3-2.4/for-2.4.19/all-in-one.patch > > and the subdirectories contain each changeset broken down and > documented. > > This brings 2.4.19's ext3 up to the 0.9.19 version in 2.4.20. >You're wonderful! Thanks so much! I managed to replicate the race condition yesterday using a short Perl script, but the filesystem was so hosed that any attempt to access the loopback mount caused the process to freeze--not even kill -9 would stop them. Running reboot also froze, so in the end we had to power cycle... We'll try this again with these additional patches.
On Wed, 16 Oct 2002 11:18:43 +0200, "Ralf Hildebrandt" <Ralf.Hildebrandt@charite.de> said:> Yesterday we upgraded from 2.4.19-ac4 to 2.4.20-pre10-ac2 and found > that this made the load on our main mailboxserver (ext3 for all > partitions, ordered mode for the Maildirs, journal for the mailspool) > go down tremendously. > > We used to be in to 10-30 range and now we're down to 3-5! >Interesting. The AC tree tends to be a bit more bleeding edge--are you comfortable using it on production mail servers? Well, I guess you must be, given that's what you're doing! Any ideas if it's one of more of the AC patches, or something else in the .20 patches, that's the source of your success? We've installed the 'all-in-one' patch to bring Ext3 up to that in .20, but haven't seen much performance impact compared to .19. We're also testing ResiserFS (with Chris Mason's data logging patches) on one of our servers. Our stress testing suggests that it performs well, but we'll have to see how it performs in a real world situation...
On Wed, 16 Oct 2002 13:42:31 +0200, "Ralf Hildebrandt" <Ralf.Hildebrandt@charite.de> said:> On Wed, Oct 16, 2002 at 11:39:01AM +0000, JP Howard wrote: > > 'all-in-one' patch to bring Ext3 up to that in .20, but haven't seen much > > performance impact compared to .19. > > > > We're also testing ResiserFS (with Chris Mason's data logging patches) on > > one of our servers. Our stress testing suggests that it performs well, > > but we'll have to see how it performs in a real world situation... > > My experiences with ReiserFS are: If you're willing to tolerate total > data loss, it's ok. > > Also, ReiserFS uses "data=writeback" (in ext3-speak), so it's faster > than "data=ordered". >Chris Mason's patches add a data=journal mode, which is what we've been testing with. I'm aware of problems with ReiserFS+NFS (now fixed), and problems that occured when the new VM went in (now fixed). I'm also aware of potential problems with inconsistent data when not using data=journal mode with the data logging patch. I'm not aware of any outstanding problems with ReiserFS that can cause corruption. But I'm no expert--are there some outstanding issues that you're aware of? I wrote a stress test tool over the weekend that simulates 500 users simultaneously using IMAP and LMTP doing a wide variety of actions. I ran it on our test server with ReiserFS for 24 hours and all was fine. But of course testing never really tells you that much about Real Life...