Thread: 8.4 doc bug for "Incrementally Updated Backups"
Based on Jürgen Fuchsberger's experience described on the general
list, I think the following should have been back-patched to 8.4 and
8.3 (assuming it gets a minor update before it EOL) as well, not just
9.0.
In particular, in 8.4 chapter 24.4.5 it says "it will be up to you to
determine how far back you need to keep WAL segment files to have a
recoverable backup" but it dangerously does not mention that it is
also up to you to construct a backup_label file such that those WAL
segment files will actually get used.
commit 13e6d6c5da184abcdfcfc9874ad17ef09f4ea044
Author: Bruce Momjian <bruce@momjian.us>
Date: Wed Aug 25 23:55:54 2010 +0000
Remove docs for "Incrementally Updated Backups" because it was of
questionable reliability; information moved to a wiki:
http://wiki.postgresql.org/wiki/Incrementally_Updated_Backups
Backpatch to 9.0.
Cheers,
Jeff
On Wed, Nov 21, 2012 at 10:09:15AM -0800, Jeff Janes wrote: > Based on Jürgen Fuchsberger's experience described on the general > list, I think the following should have been back-patched to 8.4 and > 8.3 (assuming it gets a minor update before it EOL) as well, not just > 9.0. > > In particular, in 8.4 chapter 24.4.5 it says "it will be up to you to > determine how far back you need to keep WAL segment files to have a > recoverable backup" but it dangerously does not mention that it is > also up to you to construct a backup_label file such that those WAL > segment files will actually get used. > > > > commit 13e6d6c5da184abcdfcfc9874ad17ef09f4ea044 > Author: Bruce Momjian <bruce@momjian.us> > Date: Wed Aug 25 23:55:54 2010 +0000 > > Remove docs for "Incrementally Updated Backups" because it was of > questionable reliability; information moved to a wiki: > > http://wiki.postgresql.org/wiki/Incrementally_Updated_Backups > > Backpatch to 9.0. We don't assume people are reading docs from very old versions. -- Bruce Momjian <bruce@momjian.us> http://momjian.us EnterpriseDB http://enterprisedb.com + It's impossible for everything to be true. +
On Fri, Nov 23, 2012 at 04:26:48PM -0800, Jeff Janes wrote: > On Fri, Nov 23, 2012 at 3:42 PM, Bruce Momjian <bruce@momjian.us> wrote: > > > > We don't assume people are reading docs from very old versions. > > > > Even if that is the version they are using? It is, after all, still > under maintenance, There are three options for doc patches: 1. patch only git head, meaning the next major release 2. do #1, plus the most recent major released version, e.g. 9.2.X 3. #1, #2, and all major supported released versions In general, #1 is normally for wording clarifications, #2 is for usage clarifications, and #3 is to correct mistakes. Not sure I follow that 100%, but that is what I normally do. Is that process good? Did I not follow it? -- Bruce Momjian <bruce@momjian.us> http://momjian.us EnterpriseDB http://enterprisedb.com + It's impossible for everything to be true. +
On Fri, Nov 23, 2012 at 3:42 PM, Bruce Momjian <bruce@momjian.us> wrote: > > We don't assume people are reading docs from very old versions. > Even if that is the version they are using? It is, after all, still under maintenance, Cheers, Jeff
On Fri, Nov 23, 2012 at 5:36 PM, Bruce Momjian <bruce@momjian.us> wrote: > On Fri, Nov 23, 2012 at 04:26:48PM -0800, Jeff Janes wrote: >> On Fri, Nov 23, 2012 at 3:42 PM, Bruce Momjian <bruce@momjian.us> wrote: >> > >> > We don't assume people are reading docs from very old versions. >> > >> >> Even if that is the version they are using? It is, after all, still >> under maintenance, > > There are three options for doc patches: > > 1. patch only git head, meaning the next major release > 2. do #1, plus the most recent major released version, e.g. 9.2.X > 3. #1, #2, and all major supported released versions > > In general, #1 is normally for wording clarifications, #2 is for usage > clarifications, and #3 is to correct mistakes. Not sure I follow that > 100%, but that is what I normally do. > > Is that process good? Did I not follow it? It was removed from 9.0 because it was considered to be unreliable. I think that unreliable advice about taking backups is a mistake, so it should have followed path #3. I guess it is also clarification, but a pretty major one. Cheers, Jeff
On Sat, Nov 24, 2012 at 10:56 PM, Jeff Janes <jeff.janes@gmail.com> wrote: > On Fri, Nov 23, 2012 at 5:36 PM, Bruce Momjian <bruce@momjian.us> wrote: >> On Fri, Nov 23, 2012 at 04:26:48PM -0800, Jeff Janes wrote: >>> On Fri, Nov 23, 2012 at 3:42 PM, Bruce Momjian <bruce@momjian.us> wrote: >>> > >>> > We don't assume people are reading docs from very old versions. >>> > >>> >>> Even if that is the version they are using? It is, after all, still >>> under maintenance, >> >> There are three options for doc patches: >> >> 1. patch only git head, meaning the next major release >> 2. do #1, plus the most recent major released version, e.g. 9.2.X >> 3. #1, #2, and all major supported released versions >> >> In general, #1 is normally for wording clarifications, #2 is for usage >> clarifications, and #3 is to correct mistakes. Not sure I follow that >> 100%, but that is what I normally do. >> >> Is that process good? Did I not follow it? > > It was removed from 9.0 because it was considered to be unreliable. I > think that unreliable advice about taking backups is a mistake, so it > should have followed path #3. I guess it is also clarification, but a > pretty major one. I think it can be considered a bugfix, and thus correcting a mistake. But in generael, I think we may want to be a bit more aggressive in backpatching these things. First because people will likely read the docs for the version that they are using, and thus miss things. But possibly even more important, because we know that Google (and others) tend to link to older versions of the docs as search hits before the newer ones for quite a long time after a release (or forever in some cases), which makes it even more visible. -- Magnus Hagander Me: http://www.hagander.net/ Work: http://www.redpill-linpro.com/
On Sat, Nov 24, 2012 at 01:56:27PM -0800, Jeff Janes wrote: > On Fri, Nov 23, 2012 at 5:36 PM, Bruce Momjian <bruce@momjian.us> wrote: > > On Fri, Nov 23, 2012 at 04:26:48PM -0800, Jeff Janes wrote: > >> On Fri, Nov 23, 2012 at 3:42 PM, Bruce Momjian <bruce@momjian.us> wrote: > >> > > >> > We don't assume people are reading docs from very old versions. > >> > > >> > >> Even if that is the version they are using? It is, after all, still > >> under maintenance, > > > > There are three options for doc patches: > > > > 1. patch only git head, meaning the next major release > > 2. do #1, plus the most recent major released version, e.g. 9.2.X > > 3. #1, #2, and all major supported released versions > > > > In general, #1 is normally for wording clarifications, #2 is for usage > > clarifications, and #3 is to correct mistakes. Not sure I follow that > > 100%, but that is what I normally do. > > > > Is that process good? Did I not follow it? > > It was removed from 9.0 because it was considered to be unreliable. I > think that unreliable advice about taking backups is a mistake, so it > should have followed path #3. I guess it is also clarification, but a > pretty major one. I checked just now and the unreliable advice does not appear in 8.4, so it possible that in 2010 I checked and found it only applied back to 9.0, or I might only have checked 9.0. I don't remember now, but based on the fact it was an incorrect recommendation, I probably would have backpatched it as far back as possible. Sorry I was unclear on this issue in previous emails. -- Bruce Momjian <bruce@momjian.us> http://momjian.us EnterpriseDB http://enterprisedb.com + It's impossible for everything to be true. +
On Mon, Nov 26, 2012 at 10:27:41AM +0100, Magnus Hagander wrote: > On Sat, Nov 24, 2012 at 10:56 PM, Jeff Janes <jeff.janes@gmail.com> wrote: > > On Fri, Nov 23, 2012 at 5:36 PM, Bruce Momjian <bruce@momjian.us> wrote: > >> On Fri, Nov 23, 2012 at 04:26:48PM -0800, Jeff Janes wrote: > >>> On Fri, Nov 23, 2012 at 3:42 PM, Bruce Momjian <bruce@momjian.us> wrote: > >>> > > >>> > We don't assume people are reading docs from very old versions. > >>> > > >>> > >>> Even if that is the version they are using? It is, after all, still > >>> under maintenance, > >> > >> There are three options for doc patches: > >> > >> 1. patch only git head, meaning the next major release > >> 2. do #1, plus the most recent major released version, e.g. 9.2.X > >> 3. #1, #2, and all major supported released versions > >> > >> In general, #1 is normally for wording clarifications, #2 is for usage > >> clarifications, and #3 is to correct mistakes. Not sure I follow that > >> 100%, but that is what I normally do. > >> > >> Is that process good? Did I not follow it? > > > > It was removed from 9.0 because it was considered to be unreliable. I > > think that unreliable advice about taking backups is a mistake, so it > > should have followed path #3. I guess it is also clarification, but a > > pretty major one. > > I think it can be considered a bugfix, and thus correcting a mistake. > > But in generael, I think we may want to be a bit more aggressive in > backpatching these things. First because people will likely read the > docs for the version that they are using, and thus miss things. But > possibly even more important, because we know that Google (and others) > tend to link to older versions of the docs as search hits before the > newer ones for quite a long time after a release (or forever in some > cases), which makes it even more visible. See my new email; it doesn't appear in 8.4, and I probably checked that at the time. -- Bruce Momjian <bruce@momjian.us> http://momjian.us EnterpriseDB http://enterprisedb.com + It's impossible for everything to be true. +
On Mon, Nov 26, 2012 at 01:17:18PM -0800, Jeff Janes wrote: > On Mon, Nov 26, 2012 at 12:01 PM, Bruce Momjian <bruce@momjian.us> wrote: > > On Sat, Nov 24, 2012 at 01:56:27PM -0800, Jeff Janes wrote: > >> On Fri, Nov 23, 2012 at 5:36 PM, Bruce Momjian <bruce@momjian.us> wrote: > >> > On Fri, Nov 23, 2012 at 04:26:48PM -0800, Jeff Janes wrote: > >> >> On Fri, Nov 23, 2012 at 3:42 PM, Bruce Momjian <bruce@momjian.us> wrote: > >> >> > > >> >> > We don't assume people are reading docs from very old versions. > >> >> > > >> >> > >> >> Even if that is the version they are using? It is, after all, still > >> >> under maintenance, > >> > > >> > There are three options for doc patches: > >> > > >> > 1. patch only git head, meaning the next major release > >> > 2. do #1, plus the most recent major released version, e.g. 9.2.X > >> > 3. #1, #2, and all major supported released versions > >> > > >> > In general, #1 is normally for wording clarifications, #2 is for usage > >> > clarifications, and #3 is to correct mistakes. Not sure I follow that > >> > 100%, but that is what I normally do. > >> > > >> > Is that process good? Did I not follow it? > >> > >> It was removed from 9.0 because it was considered to be unreliable. I > >> think that unreliable advice about taking backups is a mistake, so it > >> should have followed path #3. I guess it is also clarification, but a > >> pretty major one. > > > > I checked just now and the unreliable advice does not appear in 8.4, so > > it possible that in 2010 I checked and found it only applied back to > > 9.0, or I might only have checked 9.0. > > The docs were re-arranged, so it is present but in a different place. > > Sorry, I should have included the URL in the first place (24.4.5) : > > http://www.postgresql.org/docs/8.4/static/warm-standby.html OK, removed from 8.3 and 8.4 docs. I was looking in the 9.0+ file for that section. -- Bruce Momjian <bruce@momjian.us> http://momjian.us EnterpriseDB http://enterprisedb.com + It's impossible for everything to be true. +
On Mon, Nov 26, 2012 at 12:01 PM, Bruce Momjian <bruce@momjian.us> wrote: > On Sat, Nov 24, 2012 at 01:56:27PM -0800, Jeff Janes wrote: >> On Fri, Nov 23, 2012 at 5:36 PM, Bruce Momjian <bruce@momjian.us> wrote: >> > On Fri, Nov 23, 2012 at 04:26:48PM -0800, Jeff Janes wrote: >> >> On Fri, Nov 23, 2012 at 3:42 PM, Bruce Momjian <bruce@momjian.us> wrote: >> >> > >> >> > We don't assume people are reading docs from very old versions. >> >> > >> >> >> >> Even if that is the version they are using? It is, after all, still >> >> under maintenance, >> > >> > There are three options for doc patches: >> > >> > 1. patch only git head, meaning the next major release >> > 2. do #1, plus the most recent major released version, e.g. 9.2.X >> > 3. #1, #2, and all major supported released versions >> > >> > In general, #1 is normally for wording clarifications, #2 is for usage >> > clarifications, and #3 is to correct mistakes. Not sure I follow that >> > 100%, but that is what I normally do. >> > >> > Is that process good? Did I not follow it? >> >> It was removed from 9.0 because it was considered to be unreliable. I >> think that unreliable advice about taking backups is a mistake, so it >> should have followed path #3. I guess it is also clarification, but a >> pretty major one. > > I checked just now and the unreliable advice does not appear in 8.4, so > it possible that in 2010 I checked and found it only applied back to > 9.0, or I might only have checked 9.0. The docs were re-arranged, so it is present but in a different place. Sorry, I should have included the URL in the first place (24.4.5) : http://www.postgresql.org/docs/8.4/static/warm-standby.html Cheers, Jeff