From pierre@archlinux.de Tue Aug 10 00:55:33 2010 From: Pierre Schmitz To: arch-dev-public@lists.archlinux.org Subject: [arch-dev-public] FluxBB update round two Date: Tue, 10 Aug 2010 06:55:35 +0200 Message-ID: MIME-Version: 1.0 Content-Type: multipart/mixed; boundary="===============3504860849045175899==" --===============3504860849045175899== Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit Hi all, there is a new FluxBB release (1.4.2: http://fluxbb.org/forums/viewtopic.php?id=4504) I have tested the update locally (with innodb this time). They have enlarged the post.message column from text to mediumtext. Unfortunately this means rewriting the whole table for innodb; so their web update script cannot be used. (yes, I already complained and will open a bug report) My plan is to issue the following command manually before the update: alter table posts change message message mediumtext This takes about 10 minutes here and it seems the DB is not locked during that time. (So, is it correct that we can keep the forums online then?) If nobody objects I'll do the update tonight. (about now + 13h) Greetings, Pierre -- Pierre Schmitz, https://users.archlinux.de/~pierre --===============3504860849045175899==-- From dpmcgee@gmail.com Tue Aug 10 01:05:42 2010 From: Dan McGee To: arch-dev-public@lists.archlinux.org Subject: Re: [arch-dev-public] FluxBB update round two Date: Tue, 10 Aug 2010 00:05:42 -0500 Message-ID: In-Reply-To: MIME-Version: 1.0 Content-Type: multipart/mixed; boundary="===============5232695625118118633==" --===============5232695625118118633== Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: quoted-printable On Mon, Aug 9, 2010 at 11:55 PM, Pierre Schmitz wrote: > Hi all, > > there is a new FluxBB release (1.4.2: > http://fluxbb.org/forums/viewtopic.php?id=3D4504) I have tested the update > locally (with innodb this time). They have enlarged the post.message > column from text to mediumtext. Unfortunately this means rewriting the > whole table for innodb; so their web update script cannot be used. (yes, > I already complained and will open a bug report) Web update scripts are shit anyway. That has to be partly to blame for the long upgrade the last time. > My plan is to issue the following command manually before the update: > > =C2=A0alter table posts change message message mediumtext ALTER TABLE posts MODIFY message mediumtext; will probably work without needing to "rename" the column. > This takes about 10 minutes here and it seems the DB is not locked > during that time. (So, is it correct that we can keep the forums online > then?) If nobody objects I'll do the update tonight. (about now + 13h) Do what you want, but it would be good to have an error page up this time on the forum itself, rather than the HTTP password box that was there last time confusing all sorts of people. When things go longer than planned it is always a good thing to not have to hassle with it later. It is really easy, adapt accordingly (although the image is still there): Order deny,allow Deny from all Allow from 99.140.196.156 ErrorDocument 403 "Down for maintenance

The Arch Linux main site is down for maintenance.
Please try again later.

" -Dan --===============5232695625118118633==-- From jan@jgc.homeip.net Tue Aug 10 02:34:04 2010 From: Jan de Groot To: arch-dev-public@lists.archlinux.org Subject: Re: [arch-dev-public] FluxBB update round two Date: Tue, 10 Aug 2010 08:37:26 +0200 Message-ID: <1281422246.3785.1.camel@localhost.localdomain> In-Reply-To: MIME-Version: 1.0 Content-Type: multipart/mixed; boundary="===============5869186458357127827==" --===============5869186458357127827== Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit On Tue, 2010-08-10 at 00:05 -0500, Dan McGee wrote: > > alter table posts change message message mediumtext > > ALTER TABLE posts MODIFY message mediumtext; > > will probably work without needing to "rename" the column. > > > This takes about 10 minutes here and it seems the DB is not locked > > during that time. (So, is it correct that we can keep the forums > online > > then?) If nobody objects I'll do the update tonight. (about now + > 13h) I think both queries do the same. I don't know how big tables are, but you could be facing a complete table rebuild when running such a query. During this rebuild the complete table is locked for writing, as with any alter table query. --===============5869186458357127827==-- From pierre@archlinux.de Tue Aug 10 02:38:27 2010 From: Pierre Schmitz To: arch-dev-public@lists.archlinux.org Subject: Re: [arch-dev-public] FluxBB update round two Date: Tue, 10 Aug 2010 08:38:29 +0200 Message-ID: In-Reply-To: MIME-Version: 1.0 Content-Type: multipart/mixed; boundary="===============6524394106111855895==" --===============6524394106111855895== Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: quoted-printable On Tue, 10 Aug 2010 00:05:42 -0500, Dan McGee wrote: > On Mon, Aug 9, 2010 at 11:55 PM, Pierre Schmitz wro= te: >> Hi all, >> >> there is a new FluxBB release (1.4.2: >> http://fluxbb.org/forums/viewtopic.php?id=3D4504) I have tested the update >> locally (with innodb this time). They have enlarged the post.message >> column from text to mediumtext. Unfortunately this means rewriting the >> whole table for innodb; so their web update script cannot be used. (yes, >> I already complained and will open a bug report) >=20 > Web update scripts are shit anyway. That has to be partly to blame for > the long upgrade the last time. Yes, had a discussion about that with them on irc yesterday. Hopefully we'll get an cli updater in future. >> My plan is to issue the following command manually before the update: >> >> =C2=A0alter table posts change message message mediumtext >=20 > ALTER TABLE posts MODIFY message mediumtext; >=20 > will probably work without needing to "rename" the column. This takes about the same amount of time; it's probably doing the same thing internally anyway. >> This takes about 10 minutes here and it seems the DB is not locked >> during that time. (So, is it correct that we can keep the forums online >> then?) If nobody objects I'll do the update tonight. (about now + 13h) >=20 > Do what you want, but it would be good to have an error page up this > time on the forum itself, rather than the HTTP password box that was > there last time confusing all sorts of people. When things go longer > than planned it is always a good thing to not have to hassle with it > later. It is really easy, adapt accordingly (although the image is > still there): >=20 > Order deny,allow > Deny from all > Allow from 99.140.196.156 > ErrorDocument 403 "Down for > maintenance

The Arch Linux > main site is down for maintenance.
Please try again > later.

src=3D'http://dev.archlinux.org/~dan/funny-pictures-cat-wonders-if-he-is-br= oken.jpg' > />
" First of all I have correct my statement. I have to take the forums offline. While you can still read from that table during that operation, write attempts are blocked. I'll put the forum in maintenance mode during that time then. (thanks for the link to the lolcat btw :-)) I ask every admin to not touch anything while maintenance mode is enabled. It should take about 15 minutes. Btw: I tried your approach to block everyone but me last time but for some reasons it didn't work. It looked like the "allow from" directive was ignored. A few days ago I noticed that ipv6 is enabled...might that be a reason? So you need to use "Allow from ::ffff:99.140.196.156". Greetings, Pierre --=20 Pierre Schmitz, https://users.archlinux.de/~pierre --===============6524394106111855895==-- From pierre@archlinux.de Tue Aug 10 14:17:35 2010 From: Pierre Schmitz To: arch-dev-public@lists.archlinux.org Subject: Re: [arch-dev-public] FluxBB update round two Date: Tue, 10 Aug 2010 20:17:38 +0200 Message-ID: <4e41f797e6152cfe56831acc4f7fcf0a@archlinux.de> In-Reply-To: MIME-Version: 1.0 Content-Type: multipart/mixed; boundary="===============4747759907323033201==" --===============4747759907323033201== Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit On Tue, 10 Aug 2010 06:55:35 +0200, Pierre Schmitz wrote: > Hi all, > > there is a new FluxBB release (1.4.2: > http://fluxbb.org/forums/viewtopic.php?id=4504) I have tested the update > locally (with innodb this time). They have enlarged the post.message > column from text to mediumtext. Unfortunately this means rewriting the > whole table for innodb; so their web update script cannot be used. (yes, > I already complained and will open a bug report) > > My plan is to issue the following command manually before the update: > > alter table posts change message message mediumtext > > This takes about 10 minutes here and it seems the DB is not locked > during that time. (So, is it correct that we can keep the forums online > then?) If nobody objects I'll do the update tonight. (about now + 13h) > > Greetings, > > Pierre and done...this time it took even less than expected. -- Pierre Schmitz, https://users.archlinux.de/~pierre --===============4747759907323033201==--